基于改进SLIQ算法的用户侧短期负荷时间序列预测模型

User Side Short-term Load Time Series Forecasting Model Based on Improved SLIQ Algorithm

配电网净负荷需求较大,环境敏感度高,短期负荷预测可靠性较差,由此,研究基于改进SLIQ算法的用户侧短期负荷时间序列预测模型。采用SLIQ算法,分类处理包含气候变化的海量负荷数据,分析用户侧需求响应的不确定性,将分类后的数据和用户需求响应率输入基于时间序列的组合预测模型,初步预测用户侧短期负荷结果;利用S型函数的神经网络模型,改进SLIQ算法求解过程,采用梯度搜索学习的方式,修正S型激发函数计算误差,调整用户侧需求响应的不确定性,输出用户侧短期负荷预测结果。测试结果显示,该模型结合气象条件因素后,完成历史负荷数据分类的可行性较高,日负荷预测准确率和拟合系数结果分别在0.95和0.94以上,能够可靠完成用户侧负荷预测。

The net load demand of the distribution network is large,the environmental sensitivity is high,and the reliability of short-term load forecasting is poor.Therefore,the user side short-term load time series forecasting model based on the improved SLIQ algorithm is studied.The SLIQ algorithm is adopted to classify and process the massive load data including climate change,analyze the uncertainty of demand response on the user side,input the classified data and user demand response rate into the combined forecasting model based on time series,and preliminarily forecast the short-term load results on the user side.The neural network model of S-type function is used to improve the solution process of SLIQ algorithm.The gradient search learning method is used to correct the calculation error of S-type excitation function,adjust the uncertainty of demand response on the user side,and output the short-term load forecasting results on the user side.The test results show that the model is feasible to complete the classification of historical load data after combining meteorological conditions.The accuracy and fitting coefficient of daily load forecasting are above 0.95 and 0.94,respectively,which can reliably complete the load forecasting on the user side.