Dynamic Pricing With Bi-LSTM Load Forecasting: A Path Towards Grid Stability

Abstract

Fluctuations in electricity demand between peak and off-peak periods create significant challenges for modern power grids, including reduced operational efficiency, decreased reliability, and an increased risk of power outages. Peak demand conditions strain the transmission and distribution systems, while off-peak underutilization leads to inefficient resource allocation. This study proposes a machine learning–based dynamic subsidy and penalty framework to minimize demand fluctuations and enhance grid stability. Historical electricity consumption data are analyzed to identify peak and off-peak periods, and machine learning techniques are applied to forecast future demand with improved accuracy. Based on these forecasts, a real-time tariff adjustment scheme is designed, wherein subsidies are introduced during off-peak hours to encourage higher consumption, and penalties are applied during peak hours to discourage excessive demand. Additionally, the proposed approach is expected to smoothen the load curve, reduce grid stress during peak periods, and improve utilization of infrastructure during the off-peak periods. By influencing consumer behavior through adaptive pricing, the framework ensures an effective demand-side management strategy that enhances system reliability, stability, and efficiency in smart grid environments compared with existing approaches.