基于GRA-LSTM神经网络的区域综合能源系统多元负荷短期预测模型

冷、热、电负荷预测是发挥区域综合能源系统优势的关键技术。由此构建了基于灰色关联度分析(grey relation analysis,GRA)和长短期记忆(long short term memory,LSTM)神经网络的区域综合能源系统多元负荷短期预测模型,该模型利用LSTM神经网络在处理时间序列中间隔或延迟较长的样本和非线性数据方面的优势,采用GRA法定量分析多元负荷之间以及和各气象影响因素之间的耦合性。针对北方地区气候特点,利用DeST软件建立某写字楼建筑模型,运用动态模拟和统计方法模拟出写字楼全年逐时冷、热、电负荷。算例分析结果表明,基于GRA-LSTM神经网络的区域综合能源系统多元负荷短期预测模型具有较好的预测精度和应用价值。

Research on Multi-load Short-term Forecasting Model of Regional Integrated Energy System Based on GRA-LSTM Neural Network

Forecasting of cold, heat and electric load is the key technology to take advantage of regional integrated energy systems. This paper constructs a multi-load short-term prediction model for regional integrated energy systems based on GRA-LSTM neural network. This model makes good use of advantages of the LSTM neural network in processing samples with long interval or delay in time series and nonlinear data, and uses grey correlation analysis(GRA) quantitative analysis method to analyze coupling between multi-load and meteorological factors. Aiming at climate characteristics in the northern region, this paper uses DeST to simulate the establishment of an office building model, and uses dynamic simulation and statistical methods to simulate the annual cold,heat and electric load of the office building. The results of numerical examples show that the multi-load short-term prediction model of regional integrated energy system based on GRA-LSTM neural network has better prediction accuracy, and the model is reasonable and has certain application value.