基于多种模型的电力系统负荷预测

Power System Load Forecasting Based on Multiple Models

摘要: 对于时间序列的拟合预测模型有很多，但随着实际问题的不断变化，传统的时间序列预测模型已经渐渐不满足要求，现如今更加需要的是以问题为导向的，具有高精度，高泛化能力的预测模型。本文采用了ARIMA、机器学习算法对某地区未来十天间隔十五分钟用电负荷进行预测，并通过预测值和观测值的MAE、RMSE、MAPE对精度进行了比较，研究发现精度最好的为GBDT模型，所以最终选取GBDT模型来进行后续的预测工作。

Abstract: There are many fitting prediction models for time series, but with the constant change of practical problems, the traditional time series prediction model has gradually failed to meet the require-ments, and now more problem-oriented prediction models with high precision and high generaliza-tion ability are needed. In this paper, ARIMA and machine learning algorithms are used to predict the electricity load in a region at an interval of 15 minutes in the next ten days, and the accuracy is compared by MAE, RMSE and MAPE of the predicted value and observed value. It is found that the GBDT model has the best accuracy, so GBDT model is finally selected for the subsequent prediction work.