Hybrid Model for Medium-Term Load Forecasting in Urban Power Grids

In urban power planning, it is typically necessary to predict future monthly, quarterly, and annual electricity consumption to conduct advance planning and ensure the stable operation of the power grid. Therefore, accurate medium-term load forecasting is of critical importance for urban power grid planning and operation. However, current research primarily focuses on short-term forecasting, which is largely limited to a single timescale. To address this issue, this paper proposes a combined model for medium-term load forecasting, enabling predictions of loads over multiple timescales within the next year. This can help optimize power supply planning. First, by improving the 3σ criterion and incorporating holiday corrections, the original data are processed. Combining the advantages of the Prophet algorithm in capturing linear relationships and future trends with the Random Forest algorithm in capturing nonlinear relationships, a Prophet–Random Forest combined forecasting model is constructed. This model is then applied to predict the electricity consumption of a city in southern China. The results demonstrate that the proposed model achieves high accuracy in medium-term forecasting and can predict loads across multiple timescales. Specifically, for annual, quarterly, and monthly predictions, the average prediction errors are 1.02%, 2.66%, and 3.92%, respectively, showcasing strong forecasting performance.