A new stochastic diffusion process to model and predict electricity production from natural gas sources in the United States
Safa’ Alsheyab · 2025
Read A new stochastic diffusion process to model and predict electricity production from natural gas sources in the United States by Safa’ Alsheyab (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01925.txt.
Introduction
Safa Alsheyab models how natural gas's share of United States power production evolves. The tool is a new stochastic diffusion process. It is a random-walk-with-drift mathematical model, for quantities that wander over time. The variable tracked is gas-fired generation as a percentage of total electricity production.
The method works through trend function analysis, fitting both conditional and unconditional versions of the estimated trend. Parameters are estimated by maximum likelihood from discretely sampled paths of the annual series covering 1990 to 2021. The proposed process fits the historical data effectively and produces dependable medium-term forecasts, which the paper reports for 2022 and 2023. The contribution is one of method: a fresh diffusion model, with its estimation machinery worked out, applied to an energy series that planners watch closely as generation mixes shift.
Catalog record
| CREATOR | Safa’ Alsheyab |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-01925
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A new stochastic diffusion process to model and predict electricity production from natural gas sources in the United States Safa’ Alsheyab Abstract This paper introduces a new stochastic diffusion process to model the electricity production from natural gas sources (as a percentage of total electricity production) in the United States. The method employs trend function analysis to generate fits and forecasts with both conditional and unconditional estimated trend functions. Parameters are estimated using the maximum likelihood (ML) method, based on discrete sampling paths of the variable ”electricity production from natural gas sources in the United States” with annual data from 1990 to 2021. The results show that the proposed model effectively fits the data and provides dependable medium-term forecasts for 2022-2023. Math. Subject Classification: 62M86, 60H30, 65C30 Key Words and Phrases: Stochastic diffusion process; maximum likelihood estimate; fit and forecast; trend function; prediction accuracy; electricity production from natural gas 1 Introduction Electricity is a fundamental part of modern life, greatly improving our standard of living by powering homes, businesses, and technologies. However, generating electricity can also harm the environment. The extent of this damage largely depends on how electricity is produced. For instance, burning coal releases twice as much carbon dioxide as burning an equal amount of natural gas. Considering the many benefits of natural gas, such as its relatively low cost and low carbon footprint, it is considered a good source of electricity. Plus, gas plants can be built quickly, unlike nuclear facilities. The percentage of electricity produced from natural gas, \% of total, represents the share of natural gas in overall electricity generation, measured in total gigawatt hours (GWh) produced by power plants. Electricity production in the United States has undergone a structural transformation, with natural gas becoming one of the main sources of power generation. Understanding the dynamics of its share in total electricity production is essential for policy design, risk management, and long-term planning. Electricity generation from natural gas in the United States has grown steadily over the past thirty years. As reported in Table 1, production increased from 11.85815 units in 1990 to over 41.90 units in 2023, more than tripling. However, this growth has not been steady; the data show fluctuations and irregular deviations. For example, while production rose steadily during the early 2000s, it experienced a temporary dip in 2003, a sharp acceleration after 2009, and more recently, a slight decline in 2021 before rebounding in 2022–2023. Such variability indicates the influence of both long-term growth trends and short-term stochastic disturbances. Modelling such behaviour requires a framework that captures both deterministic dynamics and random shocks. Stochastic differential equations (SDEs) provide a flexible approach that combines deterministic trends with random variability to meet these challenges. Previous studies have extensively used stochastic models in finance and natural sciences, with growing applications in energy markets. For example, Schwartz [ 1 ] developed stochastic models for commodity prices, which were later adapted for electricity markets. Geman and Roncoroni [ 2 ] highlighted the importance of capturing seasonality and jumps in power prices, Benth et al. [ 3 ] used stochastic processes to describe electricity spot prices with periodic components, while Weron [ 4 ] provided a thorough review of stochastic models designed specifically for power markets, emphasizing the need to include both cyclical and random dynamics to accurately represent electricity price behavior. In the context of electricity, up to our knowledge, most of the literature focuses on prices, or consumption, or optimization models, or uses different methodological approaches such as stochastic time change, hybrid diffusions, or optimization frameworks. For example, Borovkova and Schmeck [ 5 ] have primarily focused on modelling electricity prices using a time-changed jump diffusion that adapts volatility and jump intensity over time. Mert et al. [ 6 ] develop a hybrid stochastic diffusion ensemble specifically for forecasting natural-gas prices. Leonel et al. [ 7 ] focus on optimization-based decision-making for large industrial self-producers using natural gas, aiming to minimize energy costs under uncertainty through a two-stage stochastic optimization framework. While for the natural gas production in the United States, Cai and Deng [ 8 ] propose a grid-optimized fractional grey model to forecast U.S. natural gas production; their focus is on total production forecasting using deterministic/grey system techniques. Despite these advancements, limited attention has been given to electricity production shares by fuel source within a stochastic diffusion framework. This creates a gap in the literature. Our goal is to use a new stochastic differential equation model to analyze U.S. electricity production from natural gas between 1990 and 2021 and to forecast for 2022-2023. The stochastic component captures external shocks and random variability. Building on this foundation, we consider the following SDE of the form dX(t)=\left(\frac{2}{t}-\lambda-\frac{\lambda\pi}{2t^{2}}e^{-\lambda/t}\cot(\frac{\pi}{2}e^{-\lambda/t})\right)X(t)dt+\sigma X(t)dW(t), (1) which is particularly relevant for modelling U.S. electricity production from natural gas ( \% of total), here W is the standard Brownian motion, and \lambda and \sigma are unknown indexed parameters. For more information about Brownian motion, SDEs, and their applications, we refer the reader to [ 9 ] , [ 10 ] , [ 11 ] , [ 12 ] , and [ 13 ] . The remainder of this paper is organized as follows: In the subsequent section, the proposal model is defined as the solution to a stochastic differential equation (SDE), from which the explicit expression of the process, the transition probability distribution function (TPDF), and the moments( particularly, the conditional and unconditional trends of the process) are derived. Section 3 is dedicated to estimating the model’s parameters using the Maximum Likelihood Estimation (MLE) method, based on discrete sampling of the process. In Section 4 , the application of the proposed model to specific data is presented: electricity production from natural gas sources ( \% of total) in the United States for the period 1990-2023. The accuracy of the forecast is assessed using standard performance metrics, including mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE). The performance of the proposed model is compared with that of the Gompertz model using the same dataset. Furthermore, a small Monte Carlo experiment is conducted to assess the robustness of the estimation procedure. Finally, Section 5 offers concluding remarks. 2 The model and its probabilistic characteristics In this section, we present a novel Sine-Like stochastic diffusion process. We derive and analyze the critical properties of this process, encompassing the existence of solutions, transition probability distributions, and moments. 2.1 Sine-Like (SL) stochastic diffusion model The SL process is the non-homogeneous diffusion process \{X(t),t\in[t_{1},T],t_{1}>0\} with values in (0,\infty) , and satisfies the following It \hat{\text{o}} ’s SDE dX(t)=a(t,X(t),\theta)dt+b^{1/2}(t,X(t),\theta)dW(t);\hskip 14.22636ptX(t_{1})=X_{1} (2) with infinitesimal moments given by \displaystyle a(t,X(t),\theta)=\left(\frac{2}{t}-\lambda-\frac{\lambda\pi}{2t^{2}}e^{-\lambda/t}\cot(\frac{\pi}{2}e^{-\lambda/t})\right)X(t), \displaystyle b^{1/2}(t,X(t),\theta)=\sigma X(t) (3) where \sigma>0 , and \lambda is a non-zero real constant. The solution of ( 2 )-( 2.1 ) is found, by applying It \hat{\text{o}} ’s integral, as follows \displaystyle X(t)=X_{1}+ \displaystyle\int_{t_{1}}^{t}\left(\frac{2}{s}-\lambda-\frac{\lambda\pi}{2s^{2}}e^{-\lambda/s}\cot(\frac{\pi}{2}e^{-\lambda/s})\right)X(s)ds \displaystyle+\sigma\int_{t_{1}}^{t}X(s)\text{d}W(s) 2.2 Existence and uniqueness In this part, we show the existence and uniqueness of the solution for the SL process given in ( 2 )-( 2.1 ). To achieve this goal, it is sufficient to verify uniform Lipschitz and linear growth conditions for the infinitesimal moments; see [ 14 ] , which are realized in the following theorem. Theorem