A Distributed Data Management and Machine Learning Framework for Electricity Consumption Forecasting in Power Grid Enterprises

The study addresses the development of a data management information and analytical system to support the business process of planning and forecasting electricity consumption at PJSC Rosseti Center and Volga Region. The relevance of the research is determined by the need to improve the quality, timeliness, and consistency of data processing in large power grid companies, where fragmented data sources, manual operations, and limited integration between information systems reduce the accuracy of forecasts and the efficiency of managerial decision-making. The research applies systems, process, architectural, and project approaches to analyze the current state of the enterprise and the As-Is business process, identify key limitations, formulate requirements, and design a target To-Be solution. The proposed system includes an ETL pipeline, analytical data marts, a BI dashboard, a machine learning forecasting module, and a user interface for working with forecast scenarios. A software prototype was implemented to confirm the technical feasibility of the proposed architecture. The experimental evaluation demonstrated the applicability of machine learning methods to electricity consumption forecasting, with the linear regression model achieving an R² value of 0.994 on the considered dataset. The economic assessment showed that the project is investment-attractive in the medium term, with an average annual economic effect of 7.86 million rubles, ROI of 38%, NPV of 428 thousand rubles, and a dynamic payback period of 3.78 years. The implementation of the proposed solution can improve forecasting accuracy, reduce data processing and reporting time, decrease the influence of the human factor, and enhance the analytical support of management decisions.