New Energy Power Load Forecasting and multi-time Scale Scheduling Optimization Based on Informer-Transformer

Abstract

To address the load uncertainty and limited dispatch flexibility inherent in power systems with high renewable energy penetration, this paper proposes an integrated architecture that combines the Informer prediction model with a multi-time-scale scheduling optimization strategy. Accurate forecasting and coordinated scheduling are increasingly important for maintaining the stability and efficiency of modern electromagnetic energy systems and smart grid infrastructures. The study first constructs a high-quality time series dataset from multi-source information, including load, renewable generation, and meteorological data, through systematic preprocessing. Leveraging the ProbSparse sparse attention mechanism, the Informer model performs high-precision rolling forecasting at minute-, hour-, and day-level time scales. A hierarchical multi-time-scale dispatch framework is then established to tightly couple prediction with control, covering day-ahead economic dispatch, hourly source–load balancing, and minute-level energy storage regulation. By integrating energy storage systems and enforcing operational constraints, the proposed framework jointly optimizes economic efficiency and system stability. Experimental results demonstrate that the proposed method achieves RMSE and MAE values of 0.152 and 0.118 for one-hour forecasting, significantly outperforming baseline approaches. In dispatch optimization, the operating cost is reduced to 4,281 yuan while the renewable energy curtailment rate decreases to 2.1%, indicating superior economy and robustness. The proposed framework provides an effective solution for intelligent operation of high-renewable power systems and offers valuable support for reliable electromagnetic energy management and sustainable grid operation.