A CrOA-tuned cascaded (2DOF-PDf)–(LTI) controller for coordinated LFC–AVR regulation in interconnected power systems Authors: Mohamed F Elnaggar, A A El-Rahman, Fsaha Mebrahtu Gebru, Mohamed Metwally Mahmoud Published: 2026 ABSTRACT AbstractInterconnected power systems are frequently exposed to load disturbances, parameter uncertainties, nonlinear effects, and dynamic coupling between frequency and voltage regulation loops, which make it difficult to maintain stable and well-coordinated operation. Conventional control strategies may suffer from slow damping, large oscillations, and reduced robustness when applied to such highly coupled environments. To address this problem, this paper proposes a novel cascaded controller for the coordinated regulation of Load Frequency Control (LFC) and Automatic Voltage Regulation (AVR) in a two-area interconnected power system. The proposed controller combines a two-degree-of-freedom proportional–derivative controller with filtered derivative action (2DOF-PDf) and an inner leaky tilt-integral (LTI) stage to improve transient shaping, damping, and robustness while reducing sensitivity to disturbances and parameter variations. The controller parameters are optimally tuned using the Crayfish Optimization Algorithm (CrOA) based on the Integral of Time-weighted Squared Error (ITSE) criterion. The effectiveness of CrOA is first verified through comparison with several well-known optimization methods, including the Chimp Optimization Algorithm (ChOA), Dingo Optimization Algorithm (DOA), Sine Cosine Algorithm (SCA), Gorilla Troops Optimizer (GTO), and Gradient-Based Optimizer (GBO). The proposed control scheme is then assessed under various operating scenarios, including step load disturbances, stochastic load variations, practical nonlinearities such as Generation Rate Constraint (GRC) and Governor Dead Band (GDB), time-varying voltage reference tracking, and parametric uncertainty of up to ± 40%. MATLAB/Simulink results demonstrate that the proposed CrOA-tuned cascaded (2DOF-PDf)–(LTI) controller provides faster damping, smaller frequency and tie-line power deviations, and more accurate terminal-voltage regulation than several benchmark controllers, including Proportional–Integral–Derivative (PID), Tilted Integral Derivative (TID), Fractional-Order Proportional–Integral–Derivative (FOPID), and Fractional-Order Proportional–Integral with Proportional–Integral–Double Derivative Squared (FOPI–PIDD2) controllers. Overall, the proposed controller offers a robust and effective solution for coordinated frequency and voltage regulation in interconnected power systems.Keywords: Cascaded (2DOF-PDf)–(LTI) controller, Load frequency control, Renewable generators, Sustainable development, Optimization techniquesSubject terms: Energy science and technology, Engineering, Mathematics and computing Introduction Interconnected power systems are increasingly required to operate under wide and continuously varying operating conditions. Load changes, switching actions, faults, practical nonlinearities, and the growing impact of renewable-energy variability can create persistent mismatches between generation and demand, thereby exciting coupled electromechanical and electromagnetic dynamics. In practice, power-system stability is commonly discussed in terms of frequency stability, voltage stability, and rotor-angle stability. Frequency deviations are primarily associated with active-power imbalance, whereas voltage deviations are closely related to reactive-power imbalance. Therefore, maintaining acceptable frequency and voltage profiles requires fast and well-coordinated control actions that can restore system balance while limiting overshoot, oscillations, and long settling times1–5. These objectives are commonly supported through two secondary control layers: the turbine-governor path used for Load Frequency Control (LFC) and the excitation path used for Automatic Voltage Regulation (AVR). The LFC loop is responsible for suppressing frequency deviations and regulating tie-line power exchange among interconnected areas, whereas the AVR loop aims to maintain generator terminal voltage within desired limits. Although these two loops are often analyzed separately, their dynamic interaction cannot be neglected in interconnected multi-area systems. A disturbance occurring in one area may propagate through the tie-line and affect the neighboring area, while voltage-control actions may also influence real-power output and consequently affect frequency behavior. For this reason, coordinated LFC-AVR regulation has become an important research direction for improving the dynamic security and overall performance of interconnected power systems6–9. Early studies in LFC mainly focused on improving frequency regulation in conventional thermal power systems. Later, the scope of research expanded to include deregulated structures, distributed energy resources, electric vehicles, battery storage systems, and various auxiliary support mechanisms for better active-power balancing10–15. In parallel, research on AVR has concentrated on enhancing terminal-voltage regulation and improving generator dynamic response under disturbances and parameter variations16–19. Despite these advances, fully unified LFC-AVR formulations remain relatively limited compared with the large body of work devoted separately to LFC or AVR20,21. In many published studies, controller design is still carried out primarily for one loop and then extended to the other, or both loops are modeled together but not treated using a sufficiently coupling-aware control architecture. As a result, the reported performance may become less convincing when simultaneous frequency deviations, voltage excursions, and tie-line oscillations occur under realistic operating conditions6,7,20,21. On the controller-design side, fractional-order (FO) methods have attracted significant attention because they provide more tuning flexibility than classical integer-order structures and can improve transient shaping, damping, and disturbance rejection. Controllers such as TID and other FO-based variants have been widely reported in the LFC literature and have shown promising performance improvements over conventional controllers22–24. Review-oriented studies have also highlighted the growing maturity of FO-PID families and their broad applicability in modern power and energy systems, particularly where robustness and enhanced transient performance are required25. In addition, several researchers have proposed coordinated or composite FO-based strategies for multi-area regulation under practical effects such as communication delays, nonlinearities, and uncertainty26–29. These developments indicate that advanced FO control has become a major trend in power-system regulation research. More recently, cascade control structures have emerged as an effective alternative to single-stage controllers. Cascade architectures allow the control task to be distributed across multiple stages, where one stage can shape the dominant transient dynamics while another can strengthen steady-state regulation and robustness. This makes cascade control particularly attractive for complex and highly coupled applications such as integrated LFC-AVR regulation30–34. FO-based cascade controllers have also been reported to further improve damping characteristics and regulation quality, especially in systems where tight transient control is required35–37. However, although many of these controllers show good simulation performance, some of them suffer from increased structural complexity, limited practical interpretability, or insufficient emphasis on coordinated voltage-frequency interaction. The recent literature has also shown clear progress in LFC research for renewable-rich and hybrid power systems. For example, a new CFOTID controller supported by capacitive energy storage was proposed to enhance the LFC performance of a two-area RES-penetrated multi-source power system under practical nonlinear operating conditions38,39. Recent studies have further highlighted the need for advanced control structures in interconnected power systems operating under nonlinearities, renewable-energy fluctuations, and parametric uncertainty. For example, Mishra et al. proposed a dual-layered tilt fuzzy controller for an interconnected power system integrated with DFIG, showing the benefit of structurally enhanced control design in complex operating conditions. In another study, an integral-tilt integral-derivative-with-filter (I-TDN) control structure was shown to outperform conventional PID- and I-PD-based designs in a two-area hybrid power system subject to stochastic disturbances and nonlinear effects. In addition, improved FO-control strategies optimized by modern metaheuristic methods have also been reported for nonlinear interconnected power systems, further confirming that conventional low-complexity controllers may not provide sufficient robustness and dynamic quality in such environments40,41. These studies motivate the development of the present cascaded coordinated controller for integrated LFC–AVR regulation. Likewise, a novel TDF-TIDF controller was introduced for optimal dynamic frequency regulation of a renewable-energy-based hybrid power system, where improved dynamic behavior and robustness were achieved through a composite filtered tilt-derivative and tilt-integral structure42,43. These contributions clearly confirm the effectiveness of advanced cascade and FO-inspired control designs in frequency regulation. Nevertheless, their primary focus remains on the LFC problem itself. In other words, although modern controllers have become more sophisticated for frequency control in renewable-rich settings, relatively fewer studies have addressed the coordinated design of LFC and AVR within a unified multivariable framework. Besides controller structure, optimization-based tuning has become a dominant practice in modern LFC and AVR design because the resulting control problems are generally nonlinear, high-dimensional, and nonconvex. A wide range of optimization techniques, including GA, PSO, SCA, WOA, ICA, GTO, and other metaheuristic methods, have been used to tune controller parameters for improved dynamic performance44,45. Review studies continue to emphasize that optimization-assisted control design remains highly attractive for practical power-system regulation, particularly when uncertainty and realistic operating constraints must be considered46,47. Within this context, the Crayfish Optimization Algorithm (CrOA) has recently emerged as a competitive optimization tool with strong exploration and exploitation characteristics, and improved CrOA variants have also been reported for complex search spaces48–52. However, the effectiveness of any optimizer is still strongly dependent on the controller architecture being tuned. Therefore, a practically meaningful contribution should address not only the tuning method, but also the structural suitability of the controller for integrated LFC-AVR regulation. From the above discussion, the main research gap can be clearly identified. Although the literature includes many advanced controllers for LFC, many AVR-focused strategies, and several recent cascade or FO-based designs for renewable and hybrid systems53–57, integrated LFC-AVR studies still face important limitations. First, coordinated voltage-frequency regulation is still comparatively less explored than single-loop frequency regulation. Second, some available controllers are either too complex or insufficiently transparent from an implementation perspective. Third, practical issues such as derivative kick, disturbance sensitivity, windup-like behavior, tie-line interactions, nonlinear constraints, and uncertainty are not always addressed together within one unified controller design. Consequently, there remains a clear need for a coupling-aware, practically interpretable, and optimally tuned cascaded controller capable of achieving effective frequency and voltage regulation simultaneously in interconnected power systems. Motivated by these limitations, this paper develops an improved cascaded controller for a two-area interconnected LFC-AVR system. The proposed controller combines a two-degree-of-freedom proportional-derivative outer stage with filtered derivative action, denoted as 2DOF-PDf, and an inner leaky tilt-integral (LTI) stage. The outer stage introduces setpoint weighting to decouple reference tracking from disturbance rejection, thereby helping to reduce overshoot and mitigate derivative kick, while the derivative filter limits noise amplification. The inner LTI stage provides robust integral-type action through a tilt component and a leaky integral term, which improves steady-state accuracy while alleviating excessive accumulation effects under sustained disturbances. In this way, the proposed controller is intended to preserve the performance advantages of advanced cascade and FO-inspired control concepts while remaining structurally clear and practically implementable. To determine the controller parameters effectively, the CrOA is employed as the optimization engine using an Integral of Time-weighted Squared Error (ITSE) criterion. The proposed control strategy is then evaluated on a two-area interconnected system under multiple representative operating conditions, including step load disturbances, stochastic load changes, practical nonlinearities such as Generation Rate Constraint (GRC) and Governor Dead Band (GDB), time-varying voltage reference tracking, and parameter uncertainty. The obtained results are compared against those of several benchmark controllers under identical conditions to provide a transparent assessment of frequency regulation, voltage regulation, tie-line power damping, and overall robustness. The main contributions of this paper can be summarized as follows: A novel cascaded controller, denoted as (2DOF-PDf)–(LTI), is proposed for coordinated LFC–AVR regulation in a two-area interconnected power system, providing an integrated framework for simultaneous frequency and voltage control. A novel cascaded controller, denoted as (2DOF-PDf)–(LTI), is proposed for coordinated LFC–AVR regulation in a two-area interconnected power system, providing an integrated framework for simultaneous frequency and voltage control. The proposed controller combines setpoint weighting, filtered derivative action, and a leaky tilt-integral inner stage to improve transient shaping, damping characteristics, disturbance rejection, and steady-state regulation. The proposed controller combines setpoint weighting, filtered derivative action, and a leaky tilt-integral inner stage to improve transient shaping, damping characteristics, disturbance rejection, and steady-state regulation. A CrOA-based tuning framework is developed to identify near-optimal controller parameters under predefined bounds, and the convergence behavior of CrOA is benchmarked against several established metaheuristic optimization algorithms. A CrOA-based tuning framework is developed to identify near-optimal controller parameters under predefined bounds, and the convergence behavior of CrOA is benchmarked against several established metaheuristic optimization algorithms. A systematic comparative assessment is carried out against representative benchmark controllers, including PID, TID, FOPID, FOPI-PIDD2, and related cascaded structures, under identical simulation conditions. A systematic comparative assessment is carried out against representative benchmark controllers, including PID, TID, FOPID, FOPI-PIDD2, and related cascaded structures, under identical simulation conditions. The robustness and effectiveness of the proposed controller are verified through multiple operating scenarios, including step disturbances, stochastic load variations, GRC and GDB nonlinearities, time-varying voltage reference inputs, and parametric uncertainty. The robustness and effectiveness of the proposed controller are verified through multiple operating scenarios, including step disturbances, stochastic load variations, GRC and GDB nonlinearities, time-varying voltage reference inputs, and parametric uncertainty. Comprehensive performance evaluation is presented using standard time-domain indices and an ITSE-based metric, enabling transparent and quantitative comparison across different controllers and operating conditions. Comprehensive performance evaluation is presented using standard time-domain indices and an ITSE-based metric, enabling transparent and quantitative comparison across different controllers and operating conditions. The remainder of this paper is organized as follows. Section “Modeling of an interconnected two-zone power system” presents the modeling of the two-area interconnected LFC-AVR system. Section “Design of cascaded (2DOF-PDf)–(LTI) controller” describes the design of the proposed cascaded (2DOF-PDf)–(LTI) controller and formulates the optimization problem. Section “Applied CrOA technique” introduces the CrOA and discusses the simulation results under different scenarios. Finally, the main conclusions and future research directions are summarized in “Simulation results and performance discussion” section. Modeling of an interconnected two-zone power system Structure of the two-zone interconnected network To evaluate the proposed control scheme, a widely adopted two-zone interconnected power system is considered. The network is composed of two control areas with comparable generation ratings, coupled through a tie-line that facilitates mutual active power transfer. In each zone, LFC and AVR loops operate in a coordinated manner to preserve secure and stable system performance. The AVR subsystem is responsible for regulating the generator terminal voltage within predefined operational limits, whereas the LFC subsystem aims to counteract frequency excursions resulting from variations in active power demand. Due to the inter-area electrical coupling, a disturbance introduced in one zone does not remain confined locally; instead, it propagates through the tie-line and influences the frequency dynamics of the adjacent zone. Such coupling effects underline the importance of integrated control strategies that can effectively damp inter-area oscillations while maintaining coordinated dynamic responses across the entire interconnected system. Alternatively, the interconnected two-zone system can be represented using a graph-theoretic viewpoint, where each zone is modeled as a node and the tie-line is described as an edge characterized by a synchronization gain denoted as K12. Within this framework, frequency deviations constitute the system state variables, and the magnitude of K12 governs the rate and strength of dynamic synchronization between the two zones. This representation offers valuable insight into the consensus and coordination behavior of interconnected power systems. The subsequent sections present the mathematical formulations of the LFC and AVR subsystems adopted in this work. It should be noted that the adopted two-area non-reheat interconnected power-system model is used in this study as a standard benchmark framework for transparent and controlled evaluation of the proposed coordinated LFC–AVR controller. The selection of this benchmark model allows the effect of the proposed cascaded controller to be assessed more clearly, without excessive masking from plant-model complexity. Although the adopted system is simplified, the present study is not limited to an idealized assessment. To improve practical relevance, the proposed controller is further examined under stochastic load variations, practical nonlinear constraints including generation rate constraint (GRC) and governor dead band (GDB), and parametric uncertainty levels of up to ± 40%. These scenarios provide a more realistic basis for evaluating the robustness and effectiveness of the proposed control strategy58,59. Excitation and voltage control dynamics The AVR is a key component for maintaining voltage stability in power systems by compensating for variations in reactive power demand. Changes in reactive load led to deviations in the generator terminal voltage, which must be corrected to preserve power quality and reliable operation. The AVR system continuously measures the terminal voltage and compares it with a specified reference value to generate an appropriate control action. As shown in the AVR control structure, the terminal voltage is sensed via a potential transformer and converted into a feedback signal. This signal is compared with the reference voltage, and the resulting error is amplified to produce the exciter control voltage. By adjusting the exciter field voltage, the AVR modifies the generator field current, which directly affects the induced electromotive force (emf) and, consequently, the terminal voltage. The simplified block diagram of the AVR system is illustrated in Fig. 1, while the mathematical models and parameters of the AVR components are summarized in Table 1. These models accurately represent the excitation system dynamics and serve as a foundation for evaluating the proposed control strategy under various operating conditions60. Integrated representation of the unified LFC–AVR framework for a two-area interconnected power system. Table 1Dynamic parameters of the AVR subsystem in both areas.ModelTransfer functionParametersNominal valuesParameter descriptionAmplifier,9.5, 0.12Amplifier gain and time constantExciter,1.1, 0.45Exciter gain and time constantGenerator (field CKT), 0.95, 1.1Field circuit parametersSensor,1.05, 0.015Voltage sensor parameters Dynamic parameters of the AVR subsystem in both areas. Frequency regulation and load control mechanism The LFC scheme represents a fundamental layer in power system operation, tasked with preserving nominal frequency and ensuring balanced power sharing among interconnected generation units. As illustrated in Fig. 2, the LFC structure incorporates key components such as the turbine, governor, rotating mass of the generator, and varying load demand. The primary objective of this control mechanism is to counteract active power disturbances, stabilize system frequency, and regulate power exchange through tie-lines connecting different areas. Variations in system frequency (Δf) are directly associated with deviations in the generator rotor angle (Δδ), reflecting the imbalance between mechanical input and electrical output power. These frequency deviations are continuously monitored and processed by the control system to generate an appropriate real-power control signal (Δ). This signal acts on the prime mover, prompting adjustments in the mechanical torque supplied by the turbine to restore frequency toward its nominal value. Structure of the AVR and LFC control loops associated with a synchronous generator. The dynamic behavior of the LFC loop can be described using a set of fundamental differential equations that capture the interactions among the governor, turbine, and generator–load dynamics. In the dual-area configuration, these equations collectively define the system response to load perturbations, as summarized in Table 2. The governor dynamics, represented by the state variable Δ, describe the valve position changes in response to the control input u. As shown in Fig. 2, the governor operates within a negative feedback framework characterized by a speed regulation constant of 1/R. The governor response is influenced by the LFC control gains and the speed-droop parameter R, with the frequency deviation serving as the feedback signal. Accordingly, the controller output u modulates the system frequency by compensating for load variations occurring in each area of the interconnected power network. Table 2LFC Dynamic parameters of the two-area interconnected system.ModelTransfer functionParametersArea 1Area 2DescriptionGovernor,1, 0.0851, 0.075Governor dynamicsTurbine,1, 0.321, 0.35Turbine dynamicsLoad, 118, 22110, 27Generator-load modelFrequency bias–0.430.39Frequency bias factorSpeed regulation–R2.52.8Governor speed regulation LFC Dynamic parameters of the two-area interconnected system. Integrated coordination between voltage and frequency control loops In multi-area interconnected power systems, the coupling between the AVR and LFC loops is commonly modest. This feature often permits voltage and frequency regulation to be treated, to a first approximation, as separate design problems within each control area. Nevertheless, AVR corrective actions aimed at holding the generator terminal voltage at its setpoint can produce secondary effects on the generator’s real-power output. Consequently, voltage-control interventions may alter the frequency response and, in turn, influence the behavior of the LFC loop. To account for these interactions, the AVR pathway includes a set of coupling gains (see Fig. 2) that parameterize the cross-dependence between voltage- and frequency-related dynamics. Specifically, the coefficients quantify: (i) how small variations in the stator electromotive force (emf) affect active-power generation (α1); (ii) how rotor-angle deviations contribute to terminal-voltage changes (α2); (iii) how stator-emf variations feed into rotor-angle dynamics (α₃); and (iv) how rotor-angle changes influence the stator emf (α₄). Taken together, these terms provide a compact representation of the inherent AVR–LFC coupling. Inter-area coordination is further governed by the tie-line synchronization gain K12, which reflects the strength of dynamic coupling between the two areas and impacts the rate at which their responses align. The numerical values of α1–α₄ and K12 are reported in Table 3. A reduced-order block depiction of the combined LFC–AVR arrangement is presented in Fig. 3, emphasizing the coordinated voltage–frequency regulation framework adopted in this study. Table 3Coupling and synchronization parameters of the integrated AVR–LFC system.Areaα1α2α3α4βsK12Area 10.280.120.481.351.450.58Area 20.310.110.521.421.550.58 Coupling and synchronization parameters of the integrated AVR–LFC system. Block diagram illustrating the load frequency control and automatic voltage regulation models. AVR and LFC are considered simultaneously in this work because, although they are often associated primarily with reactive-power/voltage and active-power/frequency regulation, respectively, their dynamics are not completely independent in interconnected power systems. Excitation-driven voltage-control actions may influence the generator real-power behavior through electromechanical coupling, while frequency and rotor-angle deviations can also affect terminal-voltage dynamics. Therefore, treating the two loops separately may overlook important cross-effects, especially under disturbances, tie-line interactions, nonlinear constraints, and parameter uncertainty. To avoid this limitation, the present study adopts an integrated LFC–AVR model in which the overlap between the two loops is explicitly represented by coupling coefficients α1-α4​, while inter-area interaction is captured by the tie-line synchronization gain K12 Moreover, the proposed cascaded (2DOF-PDf)–(LTI) controller is structured to manage this coupled behavior: the outer 2DOF-PDf stage shapes the dominant transient response and limits excessive sensitivity, whereas the inner LTI stage improves robustness and steady-state correction. In this way, the overlap between AVR and LFC is not neglected, but explicitly accounted for in both the system model and the controller design. Dynamic mathematical modeling of the interconnected power system The tie-line interconnection between the two control areas enables active power exchange during normal operating conditions and contributes to maintaining system balance. Nevertheless, any mismatch between generated power and load demand in either area may trigger frequency oscillations that propagate throughout the interconnected network. To capture this behavior, the dynamic characteristics of the LFC loop are mathematically modeled using a set of differential equations, as expressed below: In these expressions, Δf, , , and denote the deviations in system frequency (Hz), generated power (p.u.), governor valve position (p.u.), and tie-line power flow (p.u.), respectively. The parameters , , R, and KB represent the time constants of the load, turbine, and governor, the speed regulation coefficient (Hz/p.u.), and the frequency bias factor (p.u./Hz), respectively. The Area Control Error (ACE), denoted by e(t), serves as the primary feedback signal to the secondary controller. It is formulated as a weighted combination of the local frequency deviation and the tie-line power deviation, as given by: This control signal is applied as the input to the proposed cascade controller structure, ensuring coordinated frequency regulation and power exchange between the interconnected areas. In parallel, the AVR subsystem consists of four main components: the amplifier, exciter, generator field circuit, and sensor. The generator terminal voltage is continuously monitored and compared with a predefined reference value. The resulting voltage error signal is processed by the controller and subsequently amplified to drive the excitation system, thereby regulating the field current of the generator. Even small variations in system frequency led to changes in the rotor angle (δ), which directly affect the real power output of the generator. The relationship between rotor angle dynamics and real power variation can therefore be expressed accordingly. The terminal voltage deviation is influenced by both the rotor angle displacement and the internal emf of the generator. Accordingly, the voltage variation is modeled as a weighted combination of these two quantities, while the generator field dynamics are characterized using a standard transfer function representation. This section presents the control strategy adopted for the integrated LFC and AVR problem in a two-area interconnected power system. Due to nonlinear characteristics and parameter uncertainties inherent in practical power networks, classical controllers with fixed structures may suffer from degraded performance under severe operating conditions. To overcome these limitations, a cascade controller is employed to combine fast dynamic response with improved steady-state accuracy. Design of cascaded (2DOF-PDf)–(LTI) controller In this study, an improved cascaded regulator is developed for the integrated LFC–AVR model a Two-Degree-of-Freedom proportional–derivative controller with filtered derivative (2DOF-PDf), while retaining the inner leaky tilt–integral stage (LTI) to provide robust integral-type action. The resulting cascaded controller is denoted as (2DOF-PDf)-(LTI). The two-degree-of-freedom structure introduces setpoint weighting to decouple reference tracking from disturbance rejection, which helps reduce overshoot and mitigate derivative kick, while the derivative filter limits noise amplification. The LTI stage preserves steady-state regulation through a fractional (tilt) term and a leaky integrator 1/(s + ), reducing windup-like accumulation under large disturbances. The proposed control scheme follows a serial cascade structure consisting of two connected control blocks. This arrangement allows the outer stage to shape the dominant transient response and damping, while the inner stage strengthens steady-state accuracy and robustness. The overall transfer function of the cascaded controller can be written as: Let r(t) denote the reference (setpoint) and y(t) denote the measured feedback signal applied to the outer stage. In AGC-oriented regulation, r(t) is typically zero (regulation to nominal), while y(t) is the area control error (ACE) or the corresponding regulation signal. The 2DOF concept introduces two weighting parameters, beta and gamma, that act on the reference in the proportional and derivative paths, respectively. Accordingly, the proportional-path error and derivative-path error are defined as: To reduce noise amplification and ensure realizability, the derivative action is implemented using a first-order filter D_f(s): The outer-stage control signal is then expressed as: Equivalently, the outer-stage controller can be written compactly as: In pure regulation tasks where R(s) = 0, the outer stage acts mainly through the feedback terms; however, beta and gamma remain useful to mitigate derivative kick and improve tracking if nonzero reference changes are considered. The use of the leaky tilt–integral (LTI) stage is motivated by both robustness and practical closed-loop stability considerations. The tilt component provides a flexible integral-type action that enhances low-frequency error attenuation and improves damping characteristics, which is beneficial for the coupled LFC–AVR regulation problem. However, in the presence of sustained disturbances, nonlinear constraints, and parameter uncertainty, a pure integral action may accumulate excessively and lead to over-aggressive recovery, larger excursions, or windup-like behavior. To mitigate this effect, the leakage term introduces a controlled forgetting mechanism through 1/(s + ε), which prevents unlimited integral build-up and improves the boundedness of the controller state during prolonged transients. Therefore, the LTI stage offers a more robust alternative to a conventional integral block by preserving steady-state regulation capability while improving practical stability, disturbance tolerance, and recovery smoothness. The inner stage provides integral-type action with improved robustness by combining a fractional (tilt) term and a leaky integrator. The leaky integrator 1/ (s + ), prevents excessive accumulation of the integral term and helps alleviate windup-like behavior under large disturbances or sustained control effort. The LTI stage is defined as: where is the tilt gain, N is the tilt order parameter (N > 0), is the leaky-integral gain, and epsilon is the leakage factor. Combining (14) and (15), the proposed cascaded controller is obtained as: To obtain the optimal controller parameters, the tuning problem is formulated as a time-domain optimization task. An Integral of Time-weighted Squared Error (ITSE) performance index is adopted to penalize large deviations and promote faster settling. The objective function is defined as: The ITSE criterion was selected in this work because the coordinated LFC–AVR problem requires a performance index that accounts for both transient severity and error persistence. In comparison with ITAE, the squared-error term in ITSE imposes a stronger penalty on large deviations, which is particularly important in interconnected power systems where excessive frequency, voltage, and tie-line power excursions are undesirable. At the same time, the time-weighting factor increases the contribution of errors that remain for longer durations, thereby encouraging faster damping and improved settling behavior. For this reason, ITSE was adopted as a suitable objective function for tuning the proposed controller. The CrOA is employed to minimize J and identify the optimal coefficients of the proposed (2DOF-PDf)-(LTI) controller. The optimized controller performance is evaluated using standard transient indices such as maximum overshoot, maximum undershoot, settling time, and rise time. The ITSE criterion was adopted because the integrated LFC–AVR problem requires simultaneous suppression of large transient deviations and fast attenuation of persistent errors. The squared-error term gives higher weight to large excursions, while the time-weighting factor increases the penalty on errors that remain for longer durations. Therefore, ITSE provides a suitable objective function for coordinated frequency, voltage, and tie-line power regulation in the present study. Moreover, the final controller assessment is complemented by standard transient indices such as overshoot, undershoot, settling time, and rise time. The decision variables include controller gains, derivative filter coefficient, setpoint weighting parameters, tilt order, and leakage factor. These parameters are constrained within predefined bounds to ensure stability and practical implementation: The corresponding numerical bounds are selected as: Accordingly, the CrOA is employed to minimize the ITSE objective function and identify the optimal coefficients of the proposed (2DOF-PDf)-(LTI)controller. The optimized controller performance is subsequently evaluated using standard transient performance indices such as maximum overshoot, maximum undershoot, settling time, and rise time (Fig. 4). Schematic configuration of the proposed control strategy. Applied CrOA technique The CrOA, proposed by Heming et al., is a bio-inspired optimization method that models the natural behaviors of crayfish, particularly their feeding activity, competition, and tendency to seek shelter during high-temperature conditions. These natural behaviors are translated into three main algorithmic stages designed to maintain an effective balance between global exploration and local exploitation. The summer refuge behavior mainly supports exploration, whereas the competition and foraging behaviors are associated with exploitation. At the beginning of the algorithm, a set of candidate solutions; referred to as a crayfish colony; is randomly generated. Each solution is represented as a position vector , where each corresponds to an individual in a multi-dimensional search space. The movement and behavior of crayfish are influenced by stochastic factors that emulate variations in environmental temperature. As the temperature increases, the algorithm shifts toward either the summer refuge stage or the competition stage. During the summer refuge phase, candidate solutions are updated based on their current positions and the location of a shaded area When environmental conditions become suitable, the algorithm switches to the foraging phase, where the current best solution is treated as a food source. In the foraging phase, each crayfish assesses its current position using a fitness evaluation and compares it with the optimal solution. The movement strategy depends on the perceived size and quality of the food source. If the food quantity is excessive, the crayfish first reduces it before continuing the feeding process. Through this adaptive mechanism, CrOA alternates between broad search behaviors and solution refinement, effectively imitating crayfish survival strategies to progressively guide the population toward the optimal solution60. Population initialization To guarantee feasible and stable solutions, the decision variables of the proposed controller are initialized within predefined lower and upper bounds. The initial population is generated randomly according to the following equation: The initial population can be represented in matrix form as: Each candidate solution is evaluated using a fitness function, and the corresponding fitness values are expressed as: Temperature modeling and food intake The activity and movement patterns of crayfish are strongly affected by changes in environmental temperature, which determine the operational phase of the optimization algorithm. The temperature value is computed using Eq. (21). When the surrounding temperature exceeds 30 °C, crayfish tend to migrate toward shaded or cooler locations, a behavior that is modeled in the algorithm as the summer refuge phase. Conversely, when the temperature falls within an acceptable range, the crayfish transition to the foraging phase, where exploitation of promising solutions takes place. In addition to controlling the behavioral phase, temperature also plays a critical role in regulating the feeding intensity of crayfish. Optimal feeding occurs when the temperature lies between 15 and 30 °C, with the most favorable condition observed around 25 °C. This dependency between temperature and food intake is mathematically described by Eq. (18)61. where Temp denotes the ambient temperature experienced by the crayfish: In this formulation, the parameters σ and C1 ​ regulate the intensity of food intake under different temperature conditions, while µ represents the optimal temperature at which crayfish exhibit maximum feeding activity. Exploration phase (summer resort stage) When the temperature exceeds 30 °C, the environment is considered unfavorable, and crayfish tend to move toward sheltered locations to reduce thermal stress. In the algorithm, this behavior is modeled by guiding individuals toward a shaded position which is computed as the midpoint between the global best solution and the best solution within the current population62: where ​ denotes the best solution obtained so far throughout the search process, while ​ represents the best solution among the individuals in the current colony. If a randomly generated number satisfies rand < 0.5, it is assumed that no direct competition exists for the shelter location. Under this case, the crayfish updates its position by moving gradually toward ​ according to: where t is the current iteration index. The coefficient ​ controls the step size and is reduced over iterations to enforce a smooth transition from wide exploration to more focused search: with T representing the maximum number of iterations. This update rule enables individuals to explore promising regions around the shaded point while maintaining a controlled movement toward better solutions. Exploitation phase (competition stage) When the temperature remains high and rand \ge 0.5, multiple crayfish are assumed to be attracted to the same shelter, leading to a competitive interaction. In this stage, each candidate solution is modified using the position of another randomly selected individual to diversify the search and avoid stagnation: where z is an index corresponding to a randomly chosen crayfish, determined by: This competitive update mechanism increases population diversity and helps the algorithm escape local optima by encouraging alternative search directions. Exploitation phase (competition stage) When the temperature becomes favorable Temp łe 30, crayfish switch to feeding behavior. In CrOA, the food source is assumed to correspond to the global best solution, and its position is defined as: The amount of food available to the Q crayfish is quantified by: where is a constant food factor (typically set to 3), fitness ​ is the fitness of the solution, and ​ is the fitness value of the food source. This quantity determines whether the feeding process is direct or requires an intermediate adjustment. If the food quantity is considered large (Q > (C3 + 1)/2) the food position is first scaled to reduce the gap between the current solution and the optimal region: After this adjustment, the crayfish position is updated using an oscillatory movement modeled by sine and cosine terms: On the other hand, if the food quantity is small (Q ≤ (C3 + 1)/2) the crayfish moves toward the food source directly as: Overall, the foraging stage refines the candidate solutions by driving them toward the best-known position, while adaptive handling of Q and controlled randomness improves convergence and prevents premature trapping. The complete CrOA procedure is summarized in the flowchart shown in Fig. 5. Flowchart of CrOA51. Simulation results and performance discussion This section investigates the dynamic performance of a two-area interconnected power system regulated by the proposed cascaded (2DOF-PDf)–(LTI) controller, whose parameters are optimally tuned using the CrOA. To ensure a rigorous and systematic evaluation, the analysis is conducted following a two-level assessment framework that separately examines the effectiveness of the optimization algorithm and the closed-loop control performance under various operating conditions. In the first level of analysis, the optimization capability of CrOA is evaluated by benchmarking its convergence behavior and solution quality against several well-established metaheuristic algorithms reported in the literature, including ChOA, DOA, SCA, GTO, and GBO. All optimization techniques are executed under identical settings to ensure fairness, with a population size of 30 individuals and a maximum of 200 iterations. A step load perturbation is applied at t = 0, corresponding to a 1.5% load increase in Area 1 and a 2% increase in Area 2. The comparative results reveal that CrOA consistently achieves the lowest objective function values among all competing algorithms, resulting in an overall improvement of approximately 52%. This indicates a superior ability to explore the search space and accurately identify high-quality parameter sets for the proposed controller. In the second level of analysis, the robustness and control effectiveness of the CrOA-tuned cascaded (2DOF-PDf)–(LTI) controller are examined through extensive comparisons with other CrOA-based controllers, including PID, TID, FOPID, FOPI–PIDD2, and structures. These comparisons are carried out under a wide range of operating scenarios to assess the controller’s adaptability and stability. The considered scenarios include step load disturbances, random load fluctuations, time-varying voltage reference signals in both areas, and parametric uncertainties affecting key power-system components. Particular attention is paid to the controller’s ability to suppress frequency deviations, maintain voltage regulation, and attenuate tie-line power oscillations (ΔPtie). Detailed analyses corresponding to each operating scenario are presented in the following subsections. Performance assessment of the CrOA algorithm The convergence characteristics of six optimization algorithms; namely ChOA, DOA, SCA, GTO, GBO, and the proposed CrOA; are illustrated in Fig. 6. It shows the evolution of the best objective function value over 200 iterations, providing insight into the exploration–exploitation balance and convergence efficiency of each algorithm. It is evident that CrOA consistently converges to the lowest objective function value compared to the other methods. Furthermore, CrOA demonstrates a notably fast convergence rate during the early stages of the optimization process (approximately within the first 50–90 iterations), reflecting strong global search capability. As the optimization progresses, CrOA maintains a smooth and stable convergence trajectory, gradually refining the solution without exhibiting oscillatory behavior or premature stagnation. In contrast, ChOA and DOA display slower convergence rates and tend to settle at relatively higher objective function values, indicating limited exploitation efficiency. SCA and GTO achieve moderate convergence performance but fail to reach the solution quality attained by CrOA. Although GBO exhibits comparatively competitive behavior among the benchmarked algorithms, it still falls short of the convergence speed and final solution accuracy achieved by CrOA. These results confirm the effectiveness of the CrOA in tuning the parameters of the proposed cascaded (2DOF-PDf)–(LTI) controller. The observed convergence behavior highlights CrOA’s balanced exploration–exploitation mechanism, making it well suited for solving the nonlinear and high-dimensional optimization problem associated with integrated LFC–AVR control. The optimal parameter sets obtained using each optimization technique are summarized in Table 4. . Evolution of the optimal fitness value over successive iterations. Table 4Optimized settings of the proposed (2DOF-PDf)-(LTI) controller tuned using different optimization algorithms.BlockParameterChOADOAGTOSCAGBOCrOALFC10.853.104.604.954.204.403.904.704.954.854.604.10180.0260.0480.0500.0210.0320.0β0.350.700.850.900.650.55γ0.100.400.550.600.350.250.602.404.204.802.103.60N1.407.801.209.501.602.404.303.202.604.704.103.50\varepsilon0.0850.0400.0200.0100.0300.018AVR10.454.602.104.900.804.301.204.100.951.100.551.35420.0495.0120.0500.0360.0460.0β0.300.750.600.850.550.50γ0.080.350.250.450.200.180.904.501.704.704.201.10N1.301.808.909.809.709.604.802.100.450.200.750.40\varepsilon0.0700.0350.0150.0080.0250.012LFC20.304.804.754.904.104.154.604.854.955.002.804.35220.0460.0110.0120.0300.0260.0β0.400.850.900.920.700.60γ0.120.500.600.650.400.280.254.904.702.404.800.95N1.609.301.409.909.603.904.103.604.804.904.704.55\varepsilon0.0900.0300.0180.0090.0220.015AVR20.200.352.004.851.204.600.951.600.801.050.621.25150.0500.0160.0500.0490.0360.0β0.280.600.550.800.500.45γ0.060.300.220.400.180.152.200.401.304.903.001.10N1.903.409.809.909.757.600.154.700.600.250.550.35\varepsilon0.0600.0450.0200.0100.0300.014FitnessITSE0.0410.0460.0300.0330.0290.026 Optimized settings of the proposed (2DOF-PDf)-(LTI) controller tuned using different optimization algorithms. The performance of the CrOA is assessed through an extensive comparative analysis with five widely used metaheuristic optimization methods, including ChOA, DOA, GTO, SCA, and GBO. To ensure statistically meaningful results, each benchmark test function is evaluated over 30 independent runs for all algorithms. A population size of 40 search agents is adopted, and the maximum number of iterations is set to 300 for every optimization run. The best fitness values obtained by the competing algorithms are summarized in Table 5, whereas the evolution of the fitness values throughout the optimization process is illustrated in Fig. 7. These convergence curves provide valuable insight into the convergence speed, stability, and search efficiency of each method. The comparative results clearly demonstrate the superiority of CrOA, which consistently delivers higher-quality solutions and exhibits faster and more stable convergence behavior than the other considered algorithms. Table 5Best fitness values obtained by different optimization strategies on benchmark functions.Test FunctionFunctionChOADOAGTOSCAGBOCrOAFn.1(z)3.10 × 10−376.40 × 10−1032.90 × 10−1056.80 × 10−1317.50 × 10−1520Fn.2(z)9.50 × 10−221.20 × 10−529.40 × 10−558.20 × 10−743.10 × 10−802.40 × 10−155Fn.3(z)1.60 × 10−259.20 × 10−283.40 × 10−997.50 × 10−1023.80 × 10−1056.20 × 10−118Fn.4(z)9.208.106.503.102.450.12Fn.5(z)9.80 × 10−44.60 × 10−42.70 × 10−52.55 × 10−51.95 × 10−41.35 × 10−4Total CPU time (s)0.6150.6300.6150.5350.7500.580 Best fitness values obtained by different optimization strategies on benchmark functions. Convergence behavior of various optimization algorithms evaluated on standard benchmark test functions: (a) Function 1; (b) Function 2; (c) Function 3; (d) Function 4; (e) Function 5. Robustness assessment of the proposed cascaded (2DOF-PDf)–(LTI) controller To rigorously evaluate the robustness of the proposed cascaded (2DOF-PDf)–(LTI) controller and its effectiveness in improving the dynamic behavior of the two-area interconnected power system, an extensive set of time-domain simulations is performed using the MATLAB/Simulink environment. The robustness analysis is designed to reflect realistic operating conditions and typical disturbances encountered in practical power systems. Accordingly, several representative test scenarios are considered to examine the controller’s ability to maintain frequency and voltage stability under different types of uncertainties and nonlinear effects. The simulation results are organized into distinct scenarios, summarized as follows: Scenario I: A sudden step increase in load demand is applied at t = 0 s, with a 1.5% increase in Area 1 and a 2% increase in Area 2. Scenario II: Random load variations are imposed to assess the controller’s response under stochastic operating conditions. Scenario III: System nonlinearities, including the Generation Rate Constraint (GRC) and Governor Dead Band (GDB), are incorporated to evaluate the controller’s robustness against practical non-idealities. Scenario IV: A time-varying reference voltage signal is introduced to examine the tracking performance of the AVR loop. Scenario V: A parametric sensitivity analysis is conducted to investigate the robustness of the closed-loop system against variations in key system parameters. Scenario I: step load disturbance in both areas In this scenario, the transient performance of the interconnected power system is analyzed following an abrupt increase in load demand applied at t = 0 s, corresponding to a 1.5% load increment in Area 1 and a 2% increment in Area 2. This test is intended to evaluate the capability of the proposed control scheme to simultaneously preserve frequency and voltage stability under severe and sudden disturbances. The simulation results demonstrate that the proposed cascaded (2DOF-PDf)–(LTI) controller, with parameters optimally tuned using the Crayfish Optimization Algorithm (CrOA), achieves superior regulation performance in both the LFC and AVR loops. Compared with other CrOA-based controllers, including PID, TID, FOPID, FOPI–PIDD2, and the structures, the proposed controller exhibits a pronounced reduction in overshoot and undershoot magnitudes. Moreover, faster damping of oscillations and improved settling behavior are clearly observed. The frequency deviation responses of Area 1 and Area 2, denoted by Δf1 and Δf2, are illustrated in Fig. 8a and b, respectively. The corresponding terminal voltage deviations ΔV1 and ΔV2 are presented in Fig. 9a and b. Furthermore, the tie-line power deviation ΔPtie, depicted in Fig. 10, confirms a substantial enhancement in damping characteristics and transient response when the proposed controller is employed. Overall, these results verify the robustness of the cascaded (2DOF-PDf)–(LTI) controller and highlight its effectiveness in improving the dynamic stability and disturbance rejection capability of the two-area interconnected power system. Frequency deviation responses under Scenario I conditions: (a) ΔF1, (b) ΔF2. Voltage deviation responses corresponding to Scenario I: (a) ΔV1, (b) ΔV2. Tie-line power deviation () obtained under Scenario I. Quantitative analysis further confirms the superior performance of the proposed cascaded (2DOF-PDf)–(LTI) controller. The controller yields the lowest undershoot levels among all compared strategies, with frequency deviations reaching approximately − 87.0 × 10−3 Hz for Δf1 and − 39.2 × 10−3 Hz for Δf2, while completely eliminating overshoot in both areas. The voltage responses exhibit similarly strong regulation characteristics, where the minimum undershoot, values are confined to nearly − 0.12 V and − 0.11 V, accompanied by negligible overshoots that do not exceed 0.05 V. Moreover, the tie-line power exchange is effectively damped, with peak deviations limited to about − 1.45 × 10−3 PU and 2.15 × 10−3 PU, indicating improved inter-area power balancing and reduced oscillatory interaction. From an optimization perspective, the proposed controller significantly outperforms the , FOPI–PIDD2, TID, FOPID, and conventional PID controllers. The relative improvements in the fitness index are approximately 42%, 60%, 78%, 83%, and 85%, respectively. The minimum achieved ITSE value is on the order of 0.03, which clearly demonstrates the robustness and effectiveness of the cascaded (2DOF-PDf)–(LTI) control structure for integrated LFC–AVR regulation. A comprehensive numerical comparison of these performance metrics is provided in Table 6. Figure 11 illustrates the convergence behavior of six different controller structures tuned using the CrOA, where the fitness evolution is plotted on a logarithmic scale versus the iteration number. The convergence profiles reveal that the proposed cascaded controller exhibits the fastest and most stable convergence among all candidates, reaching a fitness level on the order of 10−3 within approximately 45–55 iterations, followed by minimal oscillations throughout the remaining iterations. This behavior reflects both rapid convergence and strong solution stability, highlighting the high degree of compatibility between the proposed controller architecture and the CrOA optimization framework when applied to the highly coupled LFC–AVR dynamics. In contrast, the FOPI–PIDD2 and TID controllers display noticeably slower convergence and less stable optimization trajectories. The FOPI–PIDD2 controller is characterized by significant initial fluctuations and requires nearly 90–110 iterations to reach a stable solution, while the TID controller remains trapped in a relatively high fitness region around 10−1, indicating limited adaptability to the multi-objective tuning requirements. Although the FOPID, PID, and controllers demonstrate moderate convergence behavior, they tend to settle prematurely at higher fitness values compared to the proposed controller, reflecting reduced optimization efficiency despite employing the same tuning algorithm. Figure 12 presents the fitness-versus-time comparison, which further emphasizes the computational efficiency of the proposed controller. The cascaded (2DOF-PDf)–(LTI) controller achieves its optimal solution within a considerably shorter execution time, combining high accuracy with reduced computational burden. In contrast, controllers such as FOPI–PIDD2 and TID require longer computation times and exhibit larger fitness oscillations, whereas the FOPID, PID, and controllers converge more slowly toward suboptimal solutions. Collectively, these results underline the critical role of controller architecture in shaping convergence behavior, even under identical optimization conditions, and confirm that the proposed cascaded controller enables CrOA to more effectively identify optimal solutions for frequency, voltage, and tie-line power regulation in interconnected power systems. Table 6Statistical comparison of key dynamic performance indices for different controllers.ControllerΔf1Δf2ΔV1ΔV2ITSEMOSMUSSTMOSMUSSTMOS (× 10−4)MUS (× 10−4)STΔV1 MOSΔV1 STΔV2 MOSΔV2 STTID0.038 − 0.568.00.049 − 0.278.01.4 − 4.89.51.143.51.143.50.108FOPID0 − 0.306.40 − 0.356.57.6 − 3.05.01.104.01.103.70.041FOPI–PIDD20 − 0.169.60 − 0.159.50.15 − 4.27.61.053.61.063.70.0330 − 0.104.40 − 0.0953.30.75 − 4.36.01.073.51.022.70.034PID0 − 0.345.20 − 0.335.16.7 − 2.04.51.203.01.202.70.031(Proposed)0 − 0.0882.80 − 0.0432.22.10 − 1.053.81.022.01.362.00.028 Statistical comparison of key dynamic performance indices for different controllers. Variation of the best fitness value with respect to iteration count. Relative error profile as a function of time. Scenario II: stochastic load fluctuations in both areas This subsection investigates the dynamic behavior of the interconnected two-area system under random load disturbances to reflect uncertain operating conditions, as depicted in Fig. 13. The time-domain responses of the six examined controllers are illustrated in Figs. 14, 15 and 16, where the proposed controller—whose parameters are tuned using the CrOA; consistently delivers the most favorable regulation performance. For the combined LFC–AVR framework, the evaluation emphasizes suppressing transient excursions and accelerating recovery. In particular, the proposed controller achieves very small frequency overshoots on the order of 6.0 × 10−3 and 1.1 × 10−3, while limiting the minimum undershoots to approximately − 6.3 × 10−3 and − 2.1 × 10−3 for Δf1 and Δf2, respectively. In addition, the settling times are reduced to about 0.26 s (Area 1) and 0.16 s (Area 2), as shown in Fig. 14. Sudden and irregular variations applied to the system load. Frequency responses under Scenario II operating conditions: (a) ΔF1, (b) ΔF2. Voltage responses under Scenario II operating conditions: (a) ΔV1, (b) ΔV2. Tie-line power deviation () under Scenario II. A similar trend is observed for the tie-line power deviation. As indicated in Fig. 16, the response associated with the proposed controller returns to steady state faster than the competing designs, reaching a stable operating point with an approximate settling time of 2.0–2.2 s under the considered random disturbances. These results demonstrate that the proposed controller structure provides superior attenuation of both overshoot and undershoot, while terminal voltage variations remain only mildly influenced by load randomness. Overall, the recommended regulator exhibits strong stabilization capability and effectively mitigates network fluctuations. A deeper comparison confirms that the CrOA-tuned proposed controller achieves noticeable performance gains relative to other CrOA-tuned designs, with improvements of about 60% over FOPID, 58% over TID, and 56% over FOPI–PIDD2 in the adopted fitness-based assessment. For clarity, the practical nonlinear constraints were set to GRC = ± 3%/min and GDB = 0.036 Hz. Moreover, the generated power responses in both areas were added for the cases with and without nonlinearities in order to explicitly demonstrate the limiting effect of GRC on the power-generation dynamics and the resulting deterioration in frequency-regulation performance as listed in Table 7. Table 7Dynamic specifications of the investigated system represented as ITSE under Scenario II.ControllerΔF1ΔF2ΔV1ΔV2Total ITSETID0.12150.09100.07580.05420.3425FOPID0.04800.03920.03150.02360.1423FOPI-PIDD20.03150.02240.01900.01280.08570.03220.02410.01820.01190.0864PID0.02460.01720.01430.00980.0659(Proposed)0.01610.01280.01080.00690.0466 Dynamic specifications of the investigated system represented as ITSE under Scenario II. Scenario III: influence of practical system nonlinearities In this scenario, the dynamic frequency regulation performance of the interconnected two-area power system is examined in the presence of practical nonlinear constraints, namely the GRC and the GDB, as depicted in Fig. 17. These nonlinear elements are inherent to real-world power plants and impose physical limitations on the rate of change of generation and governor responsiveness. For the simulations conducted in this study, the GRC and GDB limits are set to ± 3%/min and 0.036 Hz, respectively. The introduction of these nonlinearities imposes additional dynamic restrictions on the control action, leading to a noticeable deterioration in frequency regulation performance. As observed from the simulation results presented in Fig. 18, the frequency overshoot levels increase to approximately 0.24 Hz and 0.10 Hz, while the corresponding undershoot magnitudes extend from nearly − 0.10 Hz to − 0.42 Hz for ΔF1 and from − 0.09 Hz to − 0.34 Hz for ΔF2. Moreover, the settling time of the frequency responses is adversely affected due to the delayed corrective action caused by the saturation and dead-band effects associated with GRC and GDB. LFC model incorporating nonlinear effects such as GRC and GDB. Comparison of frequency deviation obtained using the proposed method with and without nonlinear constraints. In practical power plants, the frequency control loop is affected by physical nonlinearities that are not captured by ideal linear models. Two of the most common nonlinear effects are the GDB and the GRC. The GDB refers to a small range of frequency deviation within which the governor does not respond effectively, mainly due to mechanical backlash, friction, and insensitivity in the valve-governor mechanism. As a result, very small frequency changes may not trigger an immediate corrective action. On the other hand, the GRC represents the physical limit on how fast the generator output can increase or decrease, since turbines and associated actuators cannot change their power instantaneously. Therefore, even when a strong control signal is available, the actual generation response is rate-limited. These two nonlinearities make the LFC problem more realistic and more challenging, because they delay the system response, enlarge frequency excursions, and generally worsen settling behavior compared with the ideal linear case. In contrast, the terminal voltage responses in both areas remain largely unaffected by the inclusion of these nonlinear constraints, indicating that the primary impact of GRC and GDB is concentrated on the frequency control loop rather than the voltage regulation loop. Table 8 provides a quantitative summary of the influence of GRC and GDB on the main system performance indices, including frequency deviations (ΔF1 and ΔF2), tie-line power oscillations (ΔPtie), and the overall Integral of Time-weighted Squared Error (ITSE) performance index. Table 8Effect of GRC and GDB on the proposed controller performance.ConditionΔF1 MOSΔF1 MUSSTΔF2 MOSΔF2 MUSST MOS MUSITSEWithout GRC & GDB0− 0.0932.90− 0.0942.12.1− 1.150.029With GRC & GDB0.238− 0.4134.90− 0.3326.54.03− 3.550.039 Effect of GRC and GDB on the proposed controller performance. Despite the performance degradation introduced by the nonlinear constraints, the proposed cascaded controller demonstrates strong robustness. As illustrated in Fig. 19, the CrOA-tuned cascaded (2DOF-PDf)–(LTI) controller effectively mitigates the adverse effects of GRC and GDB by maintaining frequency deviations within acceptable limits and ensuring stable tie-line power exchange between the two areas. These results clearly indicate that the proposed control strategy retains satisfactory dynamic performance even under realistic nonlinear operating conditions, thereby confirming its suitability for practical power system applications. Frequency response characteristics for Scenario III: (a) ΔF1, (b) ΔF2. Scenario IV: time-varying reference voltage tracking This scenario evaluates the robustness and tracking capability of the investigated control strategies when subjected to time-varying reference voltage profiles in both Area 1 and Area 2, as illustrated in Fig. 20. This operating condition is particularly challenging, since continuous variations in the desired terminal voltage demand fast, accurate, and well-coordinated responses from the integrated LFC–AVR control system. The time-domain responses of the six considered controllers are depicted in Figs. 21, 22 and 23, where the proposed cascaded (2DOF-PDf)–(LTI) controller, tuned using the CrOA, demonstrates superior dynamic performance. For frequency regulation, the analysis focuses on limiting overshoot, undershoot, and settling time under varying voltage references. The proposed controller achieves very small frequency overshoots of approximately 8.7 × 10−3 for both ΔF1 and ΔF2, while the maximum undershoots are confined to about − 1.3 × 10−2 for ΔF1 and − 2.1 × 10−3 for ΔF2. Furthermore, the settling times are significantly reduced to nearly 1.1 s for ΔF1 and 1.6 s for ΔF2, indicating fast and effective frequency stabilization despite the continuously changing voltage commands. The terminal voltage responses (ΔV1 and ΔV2), shown in Fig. 23, closely track the imposed time-varying reference trajectories with minimal deviation. The maximum voltage overshoots are limited to around 0.014 V and 0.021 V, while the corresponding undershoots do not exceed − 0.043 V and − 0.025 V, respectively. Specified reference voltage deviations for both control areas: (a) Area 1, (b) Area 2. Frequency responses corresponding to Scenario IV: (a) ΔF1, (b) ΔF2. Tie-line power deviation () observed under Scenario IV. Voltage response characteristics under Scenario IV: (a) ΔV1, (b) ΔV2. In addition, the proposed controller reaches steady-state conditions rapidly, with an average settling time of approximately 2 s, confirming its strong voltage tracking capability. Regarding power exchange dynamics, the ΔPtie response illustrated in Fig. 22 shows that the CrOA-tuned cascaded (2DOF-PDf)–(LTI) controller yields the smallest overshoot and undershoot levels, approximately 1.3 × 10−3 PU and − 3.2 × 10−3 PU, respectively, during voltage variations. This behavior highlights the controller’s effectiveness in suppressing frequency oscillations and maintaining stable tie-line power flow even under rapid reference voltage changes. Quantitatively, the performance improvements achieved by the proposed controller amount to approximately 79% compared with the CrOA-based TID controller, 71% compared with the CrOA-based PID controller, 57% compared with the CrOA-based FOPID controller, 22% compared with the CrOA-based FOPI–PIDD2 controller, and 31% compared with the CrOA-based controller. These results clearly confirm the strong robustness and adaptability of the proposed control structure under demanding voltage-tracking conditions as listed in Table 9. Table 9ITSE-based performance comparison under Scenario IV.ControllerΔF1ΔF2ΔV1ΔV2Total ITSETID0.10100.08320.06400.05010.2983FOPID0.04510.03650.02760.01910.1283FOPI-PIDD20.03040.02050.01800.01270.08160.01960.01500.01110.0880.0508PID0.02240.01700.01210.00930.0608(Proposed)0.01360.01150.01000.00780.0429 ITSE-based performance comparison under Scenario IV. Scenario V: sensitivity assessment under parameter uncertainty A sensitivity study is conducted to verify whether the integrated two-area LFC–AVR system can preserve acceptable dynamic performance when key model parameters deviate from their nominal values within a prescribed tolerance band. In this work, robustness is examined by systematically perturbing representative parameters of the governor–turbine–load block and the AVR block in both areas. Specifically, the parameters are varied by ± 20% and ± 40% around their nominal settings, while maintaining the same step-load disturbance profile adopted in Scenario I. The proposed cascaded (2DOF-PDf)–(LTI) controller, tuned via the CrOA, is then evaluated for each perturbed case. To provide a clear interpretation, the analysis is carried out in two stages. First, only the LFC-related parameters in Area 1 and Area 2 are perturbed while the AVR parameters are kept at their nominal values. Second, the AVR parameters are perturbed while the LFC block remains unchanged. For each test, the reported indicators include the maximum undershoot (MUS) and settling time (ST) of the frequency deviations (ΔF1, ΔF2), in addition to the maximum overshoot (MOS) and maximum undershoot (MUS) of the tie-line power deviation (ΔPtie). Overall, the results indicate that the proposed controller sustains stable operation and exhibits only modest variations in transient characteristics within the tested uncertainty ranges. As indicated in Table 10, the proposed controller maintains bounded frequency and tie-line responses for all tested LFC parameter variations. Although increasing or decreasing governor and turbine time constants tends to slow down the transient recovery, the system remains stable and the overall deviation levels stay within acceptable limits. Table 10LFC-parameter sensitivity results (Scenario I disturbance retained).ParameterVariationΔF1MUS (Hz)ΔF1ST (s)ΔF2MUS (Hz)ΔF2ST (s) MOS (× 10−4 PU) MUS (× 10−4 PU)Nominal0%− 0.0964.0− 0.0512.42.08− 1.18 + 40%− 0.0874.3− 0.0714.41.83− 1.92 + 20%− 0.0914.2− 0.0744.21.90− 1.73 − 20%− 0.1204.6− 0.1104.22.68− 1.95 − 40%− 0.1254.8− 0.1154.32.90− 2.01 1 + 40%− 0.1184.2− 0.0974.52.76− 2.96 + 20%− 0.1134.2− 0.0924.32.60− 2.79 − 20%− 0.1024.1− 0.0914.22.16− 1.48 − 40%− 0.0984.3− 0.0874.42.18− 1.58 + 40%− 0.0884.1− 0.0743.91.68− 2.00 + 20%− 0.0914.2− 0.0793.71.74− 1.93 − 20%− 0.1154.6− 0.1084.72.58− 1.54 − 40%− 0.1194.9− 0.1134.82.72− 1.62 1 + 40%− 0.1154.3− 0.0944.62.30− 2.18 + 20%− 0.1064.2− 0.0914.32.26− 2.09 − 20%− 0.0834.1− 0.0904.22.02− 0.50 − 40%− 0.0764.2− 0.0884.71.95− 0.38 1 + 40%− 0.0844.2− 0.0904.72.22− 1.36 + 20%− 0.0934.1− 0.0914.22.16− 1.30 − 20%− 0.0984.2− 0.0904.22.10− 0.98 − 40%− 0.0954.3− 0.0864.32.07− 0.85 1 + 40%− 0.0994.8− 0.0874.42.09− 1.03 + 20%− 0.0964.6− 0.0904.22.11− 1.08 − 20%− 0.0924.2− 0.0924.62.16− 1.34 − 40%− 0.0894.3− 0.0964.72.20− 1.52 2 + 40%− 0.1014.5− 0.0994.42.02− 0.83 + 20%− 0.0974.3− 0.0944.32.06− 0.89 − 20%− 0.0944.6− 0.0884.12.21− 1.56 − 40%− 0.0914.9− 0.0834.42.30− 1.70 2 + 40%− 0.0984.2− 0.1554.52.52− 0.58 + 20%− 0.0954.2− 0.1014.32.41− 0.72 − 20%− 0.0944.1− 0.0824.22.15− 1.98 − 40%− 0.0914.3− 0.0754.22.17− 2.25 2 + 40%− 0.0994.6− 0.0904.42.03− 0.86 + 20%− 0.0964.2− 0.0924.32.09− 1.05 − 20%− 0.0944.1− 0.0924.22.25− 1.48 − 40%− 0.0974.4− 0.0954.42.36− 1.69 2 + 40%− 0.1004.4− 0.0904.42.38− 1.64 + 20%− 0.0964.2− 0.0914.32.20− 1.43 − 20%− 0.0944.1− 0.0914.22.05− 0.97 − 40%− 0.1014.3− 0.0884.61.98− 0.90 LFC-parameter sensitivity results (Scenario I disturbance retained). Table 11 indicates that AVR parameter uncertainty has a limited influence on frequency and tie-line dynamics, while it can slightly affect the transient characteristics of terminal voltage tracking. Nevertheless, the proposed CrOA-tuned cascaded controller preserves stable and well-damped responses across both areas for all tested perturbation levels, confirming strong robustness against parametric variations. Table 11AVR-parameter sensitivity results (Scenario I disturbance retained).ParameterVariationΔF1ST (s)ΔF2ST (s) MOS (× 10−4 PU) MUS (× 10−4 PU)ΔV1MOS (PU)ΔV1MUS (PU)ΔV1ST (s)ΔV2MOS (PU)ΔV2MUS (PU)ΔV2ST (s)Nominal0%3.22.42.08− 1.181.060.852.41.060.892.4 + 40%4.86.92.28− 1.321.100.793.21.110.753.3 + 20%4.55.12.18− 1.251.080.822.91.080.812.9 − 20%4.64.42.02− 1.081.070.882.51.070.872.6 − 40%5.05.11.92− 1.011.060.932.01.060.941.9 + 40%4.96.72.30− 1.121.270.773.31.180.823.3 + 20%4.64.82.20− 1.141.120.813.01.090.833.0 − 20%4.34.12.05− 1.171.050.872.31.040.882.4 − 40%4.44.11.93− 1.141.080.892.01.080.911.8 + 40%4.84.72.29− 1.271.110.743.41.120.743.4 + 20%4.34.22.18− 1.201.080.813.01.080.813.0 − 20%4.24.22.02− 1.061.060.872.21.060.872.2 − 40%4.34.61.92− 1.001.040.922.11.030.921.6 + 40%4.64.72.06− 1.031.080.892.71.160.942.8 + 20%4.14.12.08− 1.111.090.872.61.100.862.6 − 20%4.44.22.10− 1.231.070.842.61.070.842.6 − 40%5.04.82.08− 1.301.040.792.91.030.723.0 + 40%4.84.72.25− 1.301.090.733.51.110.773.4 + 20%4.54.22.18− 1.231.080.813.21.090.813.1 − 20%4.14.12.02− 1.081.060.882.51.060.882.5 − 40%4.34.51.95− 1.031.030.921.91.030.941.9 + 40%4.74.82.02− 1.091.080.912.01.080.882.1 + 20%4.24.32.07− 1.141.070.862.61.080.862.6 − 20%4.04.22.14− 1.231.080.812.51.080.822.5 − 40%4.54.22.24− 1.291.070.732.91.070.752.9 AVR-parameter sensitivity results (Scenario I disturbance retained). Finally, it is worth noting that practical implementations often rely on discrete-time digital control, where sampling may alter stability margins and transient features relative to continuous-time models. Although the present study adopts a continuous-time formulation for clarity, the proposed framework can be extended to sampled-data realizations. The main numerical challenges are associated with storing state histories and solving coupled nonlinear algebraic equations at each step; however, these issues can be mitigated using efficient solvers and structured data handling, maintaining a reasonable accuracy–cost trade-off for simulation. From a computational standpoint, the proposed control framework involves two distinct aspects: offline tuning complexity and online implementation complexity. The offline burden is mainly related to the CrOA-based optimization process, where multiple controller parameters are tuned simultaneously under repeated time-domain simulations. Since the proposed cascaded (2DOF-PDf)–(LTI) controller contains more decision variables than simpler structures such as PID and TID, its tuning stage is naturally more computationally demanding. However, the obtained convergence results show that the proposed structure is more compatible with the CrOA search mechanism, reaching high-quality solutions in fewer iterations and with shorter execution time than several competing controllers. After tuning, the online implementation burden remains moderate because the controller is applied using fixed parameters and does not require real-time optimization. Therefore, the additional offline computational effort can be justified by the significant improvement achieved in dynamic performance, robustness, and regulation quality. Conclusions Coordinating the AVR with the LFC loop through appropriate coupling mechanisms provides an effective means for simultaneously enhancing voltage and frequency stability in interconnected power systems. In this study, an improved cascaded secondary control strategy based on the Cascaded (2DOF-PDf)–(LTI) structure has been developed to regulate the integrated AVR–LFC dynamics of a conventional two-area power system. The controller parameters were optimally tuned using the CrOA, which offers a favorable balance between global exploration and local exploitation during the optimization process. To ensure a fair and meaningful evaluation, the proposed CrOA-tuned cascaded controller was systematically compared with several well-established benchmark regulators, including PID, TID, FOPID, FOPI–PIDD2, and controllers, under identical operating conditions. A wide range of test scenarios was considered to thoroughly assess closed-loop performance and robustness. These scenarios encompassed step load disturbances, stochastic load variations, practical power-system nonlinearities such as the GRC and GDB, time-varying voltage reference tracking, and parametric uncertainty involving ± 20–40% deviations in key LFC and AVR parameters. Across all examined scenarios, the proposed Cascaded (2DOF-PDf)–(LTI) controller consistently demonstrated superior dynamic performance. In particular, it achieved faster damping of oscillations, reduced frequency and tie-line power deviations, and improved voltage tracking accuracy compared with the competing control structures. Quantitative assessments based on time-domain indices and ITSE performance metrics confirmed that the proposed controller yielded the lowest overall error values, indicating enhanced disturbance rejection capability and improved regulation quality. These results highlight that the cascaded configuration provides additional tuning flexibility and enables more effective coordination between the active-power (frequency) and reactive-power (voltage) control channels within the integrated AVR–LFC framework. Despite the demonstrated performance improvements, certain practical considerations remain. The proposed cascaded structure introduces a larger number of tunable parameters, which may increase tuning complexity and computational effort compared with simpler control schemes, particularly when metaheuristic optimization is employed. Furthermore, the present investigation was limited to a standard two-area benchmark system and did not explicitly incorporate communication delays, cyber-physical constraints, or hardware implementation limitations. Future research will focus on extending the proposed control strategy to multi-area and renewable-rich power systems, where higher levels of uncertainty and variability are present. Despite the promising results, the present study has some limitations. The adopted two-area non-reheat interconnected power-system model, although widely used as a benchmark, represents a simplified dynamic framework compared with practical large-scale power systems. It was intentionally selected to provide a controlled and transparent environment for evaluating the proposed coordinated LFC–AVR controller under stochastic loading, nonlinear constraints, and parameter uncertainty. However, the current model does not explicitly include reheat turbine dynamics, multi-source generation structures, renewable-energy penetration, storage participation, communication delays, or other higher-order practical nonlinear interactions. Therefore, future work will focus on extending the proposed cascaded (2DOF-PDf)–(LTI) controller to more realistic reheat, multi-source, renewable-rich, and higher-order interconnected power-system models, in addition to implementation-oriented validation using real-time simulation or hardware-in-the-loop platforms Additional studies will also consider the impact of communication delays and further nonlinear constraints, as well as experimental validation using hardware-in-the-loop platforms such as OPAL-RT or dSPACE. Moreover, hybrid tuning frameworks that combine CrOA with adaptive or learning-based mechanisms may be explored to reduce computational burden while preserving strong robustness under severe and rapidly changing operating conditions. Acknoledgments The authors extend their appreciation to Prince Sattam bin Abdulaziz University for funding this research work through the project number (PSAU/2025/01/38871) the crrected acknoedgments are: The authors extend their appreciation to Prince Sattam bin Abdulaziz University for funding this research work through the project number (PSAU/2025/01/38871). Author contributions Mohamed F. Elnaggar, A. A. El-Rahman, Fsaha Mebrahtu Gebru, and Mohamed Metwally Mahmoud: conceptualization, methodology, software, visualization, investigation, writing, original draft preparation. data curation, and validation. Mohamed Metwally Mahmoud supervision, resources, writing, review, and editing. Funding The authors extend their appreciation to Prince Sattam bin Abdulaziz University for funding this research work through the project number (PSAU/2025/01/38871). Data availability The datasets used and/ or analysed during the current study are available from the corresponding author on a reasonable request. Declarations Competing interests The authors declare no competing interests. Declaration of generative AI in scientific writing The authors acknowledge the use of AI-assisted tools, notably Grammarly, to improve linguistic accuracy, correct grammatical errors, and increase the overall coherence of this article. The content, analytical discussion, and conclusions presented in this work are distinctly the original contributions of the author, with AI tools utilized solely to enhance the presentation and clarity of the text. The use of AI complies with the journal’s requirements for transparency and ethical norms in authorship processes. Sustainable development goals Sustainable Development Goals mapped to this document, Affordable and Clean Energy Goal 7. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Fsaha Mebrahtu Gebru, Email: m24fsaha@gmail.com. Mohamed Metwally Mahmoud, Email: metwally_m@aswu.edu.eg. References Associated Data Data Availability Statement The datasets used and/ or analysed during the current study are available from the corresponding author on a reasonable request.