Enhanced Short-Term Load Forecasting Using BILSTM and AdaBoost with Error Correction

Short-term load forecasting is crucial for the safety of the power grid and the operation and maintenance of electric fields. However, existing methods have limitations in capturing the complex nonlinear characteristics of loads, resulting in low prediction accuracy. To improve the predictability of load sequences, this paper proposes a load forecasting model based on Bidirectional Long Short-Term Memory (BILSTM) networks and the Adaptive Boosting (AdaBoost) algorithm, combined with an error correction strategy for short-term load forecasting. Firstly, an initial model is established using BILSTM, and the sample weights are iteratively adjusted through AdaBoost to obtain the enhanced BILSTM-AdaBoost forecasting model. Subsequently, an error prediction model is constructed to correct the errors in the initial prediction results. Experimental results show that the proposed model achieves a prediction accuracy of over 98%, verifying its effectiveness.

I. Introduction

In modern society, the importance of smart grids is increasingly prominent. By integrating advanced network technologies and energy management techniques, smart grids not only enhance energy efficiency but also improve network security and demand-side management capabilities ^[1]. Among the numerous applications of smart grids, Short-Term Load Forecasting (STLF) plays a crucial role and is widely used in power supply scheduling, electricity pricing, renewable energy integration, and reducing grid maintenance costs ^[2]. However, due to the time series and nonlinear characteristics of load data, STLF faces significant challenges ^[3].