基于深度学习的能源系统需求响应短期负荷预测

Short⁃term load forecasting of energy system demand response based on deep learning

为提升能源系统短期负荷预测精度,解决需求响应给能源系统带来的全新挑战,提出基于深度学习的能源系统需求响应短期负荷预测方法。构建栈式自编码神经网络SAE-NN深度学习模型,考虑能源系统的需求响应,并对其数据进行调整,得到用户预期收益和舒适度是影响短期负荷变化的两项重要因素。在构建的SAE-NN深度学习模型中进行样本训练,提取深层特征。利用逻辑回归(LR)模型预测能源系统的短期负荷,提升模型预测精度。仿真结果表明,所提方法的短期负荷预测结果与实际负荷基本一致,测得的相对误差为0.61%,平均绝对误差为1.88%,均方根误差为0.523,希尔不等系数为0.009,均低于对比方法,短期负荷预测精度高,预测效果好。

In order to improve the accuracy of short⁃term load forecasting of energy system and solve the new challenges brought by demand response to energy system,a method of short⁃term load forecasting of energy system demand response based on deep learning is proposed.The deep learning model of stack type self coding neural network SAE⁃NN is constructed.Considering the demand response of energy system,and adjusting its data,the expected income and comfort of users are two important factors affecting the short⁃term load change.Samples are trained in the constructed SAE⁃NN deep learning model to extract deep features.Using the Logistic Regression(LR)model to predict the short⁃term load of the energy system,improve the prediction accuracy of the model.The simulation results show that the short⁃term load forecasting results of the proposed method are basically consistent with the actual load.The measured relative error is 0.61%,the average absolute error is 1.88%,the root mean square error is 0.523,and the hill inequality coefficient is 0.009,which are lower than the comparison method.The short⁃term load forecasting accuracy is high and the forecasting effect is good.