Іntelligent automation of the forecasting procedure for the operations of electric power enterprises

Abstract

The article investigates a highly relevant scientific and practical problem related to the development of intelligent automation systems for forecasting the operations of electric power enterprises under conditions of high information uncertainty and incomplete input data. Decision-makers face increasing responsibility within a competitive business environment. This necessitates the urgent implementation of novel mathematical computational models and intelligent information technologies. The research focuses on automating the forecasting process for electric power enterprises. The primary goal is to develop a software application based on formal, structural-functional, and logical models.

In this work, a comparative review of existing forecasting methods has been conducted, highlighting time-series methods, exponential smoothing, simple and moving averages. Special attention is paid to the development of the fuzzy forecasting system architecture, which accounts for both the probabilistic nature of the parameters and the specific characteristics of the electric power industry. The development cycle of the fuzzy model is described in detail, including the stages of fuzzification of input variables, the formation of a comprehensive knowledge base, the configuration of the inference engine based on fuzzy rules, and the final defuzzification process to obtain real-world values for long-term strategies. To limit the confidence interval of the parameters, various membership functions (Gaussian, triangular, and exponential) are applied. Additionally, the model incorporates reliability indicators for the components of the electric power complex. It is proposed to use the constructed knowledge base in conjunction with statistical forecasting results to refine the trend of the investigated parameter.

During the modeling process, three key regions of failure rate were identified. This distribution improved the accuracy of evaluating the equipment's technical condition and the probability of an application failure event. The study's result is a software complex that has been created and fully tested, utilizing artificial intelligence modules and a neural network-based forecasting architecture. The conducted forecasting experiments on retrospective data have confirmed the high efficiency, stability, and adequacy of the developed system and its evaluation results. Ultimately, this approach significantly minimizes potential financial risks and optimizes resource allocation.