基于CNN-LSTM的综合能源系统负荷预测模型

负荷的准确预测是综合能源系统设计、运行、调度和管理的前提。现有的负荷预测模型中大都考虑了气象、日期因素，却没有考虑系统中电、冷、热负荷间的相关性，这会对模型的预测精度造成影响。使用了科普拉理论对系统中3种负荷之间的相关性进行分析。从分析结果看，它们之间具有强相关的关系。基于上述分析结果，提出了一种基于深度学习的智慧综合能源系统负荷预测模型，该模型使用卷积神经网络(convolutional neural network,CNN)来提取系统中电、冷、热负荷间的耦合特性相关的特征量。将得到的特征量转换为时间序列后，输入到长短期记忆(long short-term memory,LSTM)网络中进行负荷预测。实验显示，所提出的CNN-LSTM组合模型的预测精度更为精准，可为综合能源系统的负荷预测提供参考。

Load prediction model of integrated energy system based on CNN-LSTM

Accurate load prediction is the premise of design, operation, scheduling, and management of integrated energy system. Most existing load forecasting models consider meteorological and date factors, but do not consider the correlation between electricity, cold and heat loads in system, which will affect the prediction accuracy of the model. In this paper, Copula theory is used to analyze the correlation between three kinds of loads in the system. From the analysis results, there is a strong correlation between the three loads. To improve the prediction accuracy of the model, this paper proposes a load prediction model of integrated energy system based on deep learning. Firstly, the model uses convolutional neural network (CNN) to extract the feature quantities related to the coupling characteristics of electric, cold, and hot loads in the system. After characteristic values are converted into time series, these series were input into long short-term memory (LSTM) network for load prediction. Experimental results show that the prediction accuracy of CNN-LSTM combined model proposed in this paper is more accurate, which can provide reference for the load prediction of the system.