关键词: 综合能源系统, 多元负荷联合预测, 光伏发电功率预测, 数据驱动建模, 运行优化

Abstract: In recent years, the rapid development of digital technologies such as the Internet of Things (IoT), big data, and artificial intelligence (AI) has introduced new method for optimizing the operation of integrated energy systems (IES). This paper proposes a data-driven approach for IES operation optimization, focusing on an industrial park with a self-contained energy station in northern China. Firstly, a deep learning long short-term memory (LSTM) neural network is employed to conduct multivariate load forecasting and photovoltaic (PV) power generation prediction, providing a precise basis for energy station optimization. Secondly, data-driven machine learning algorithms are applied to establish full working-condition models for key energy supply equipment. Finally, optimization is performed using the particle swarm optimization (PSO) algorithm under three objectives: energy efficiency, economic performance, and comprehensive benefits. The typical daily optimization results demonstrate the following: under the energy efficiency-optimal scenario, the system achieves a comprehensive energy utilization rate of 83.0% with an operating cost of 64,802 CNY; under the economic-optimal scenario, the operating cost is reduced to 64,590 CNY, while the energy utilization rate remains at 79.3%; under the comprehensive benefit-optimal scenario, compared with actual operation, the energy utilization rate improves by 7.5%, and operating costs are reduced by 6,444 CNY. The results indicate that the proposed method provides practical significance for guiding the optimization of integrated energy system operations.

Key words: integrated energy system, joint forecasting of multi-loads, photovoltaic power forecasting, data-driven modeling, operational optimization