关键词: 综合能源系统, 负荷预测, 卷积神经网络, 长短期记忆神经网络, 注意力机制, 多目标霜冰算法

Abstract:

To improve the forecast accuracy and operation economic efficiency of integrated energy systems under multi-energy coupling conditions， an electric-thermal load forecasting and optimal scheduling method was proposed based on an improved convolutional neural network and long short-term memory neural network（CNN-LSTM）. The method integrated CNN and LSTM， introduced an attention mechanism to enhance the capabilities of feature extraction and temporal response， and achieved the coordinated optimization of the forecasting model and operation strategy in combination with the multi-objective rime optimization algorithm（MORIME）. Simulation results based on typical summer and winter days of a park in Jinan showed that the improved CNN-LSTM model achieved a 32.6% reduction in the mean absolute percentage error of load forecasting compared to that of the traditional model. With MORIME， the total operation costs of the system in summer and winter reduced by 12.7% and 10.3% respectively， and the energy utilization efficiency of the system increased by 3.8% to 4.1%. The efficient coordination of the electric-thermal energy storage system and the time-shift utilization of energy significantly improve the operation economic efficiency and energy utilization efficiency of the system， thereby providing a new technical approach for the optimal scheduling of complex multi-energy systems.

Key words: integrated energy system, load forecasting, convolutional neural network, long short-term memory neural network, attention mechanism, multi-objective rime optimization algorithm