基于LightGBM算法的短期电力负荷预测研究

Research on Short-Term Power Load Forecasting Based on LightGBM Algorithm

摘要: 电力用户侧短期负荷需求的快速准确预测是实现电网优化调度的基础，只有搞清楚用户侧负荷需求，才能更好实现负荷供给。因电力负荷需求受气象、生产生活等多重因素的影响，而现有方法预测准确度不够高，制约了电网总体运行效率的进一步提高。本文提出了一种基于LightGBM算法的短期电力负荷预测模型，特征输入参数中考虑了气象参数和用户历史用电指标，模型中使用了五折交叉验证法，以提高预测精确度及泛化性，并利用某省的实际电网母线的电力数据，对模型进行训练及验证。结论表明：不同历史数据规模大小的选择对LightGBM模型的预测精度有一定影响，预测数据与实际值相比，平均绝对误差均小于0.3，平均绝对百分比误差均小于0.01%，均方根误差均小于1%。将本文模型和多元线性回归模型及XGBoost模型的预测结果进行对比，证明本文模型具有更高的精确度和预测效果。

Abstract: The fast and accurate prediction of short-term load demand on the power user side is the basis of realizing the optimal dispatching of power grid. Only by understanding the load demand on the user side can we better realize the load supply. Because the power load demand is affected by multiple factors, including meteorology, production and life, and the prediction accuracy of the existing methods is not high enough, which restricts the further improvement of the overall operational efficiency of the power grid, in this paper, a short-term power load forecasting model based on LightGBM algorithm is proposed. The meteorological parameters and historical power consumption indexes of users are considered in the character input parameters. The 5-fold cross validation method is used in the model to improve the prediction accuracy and generalization of the model, and the model is trained and verified by using the power data of the actual power grid bus in a province. The conclusion shows that the selection of the size of different historical data has a certain impact on the prediction accuracy of LightGBM model. Compared with the actual value, the average absolute error of the prediction data is less than 0.3, the average absolute percentage error is less than 0.01%, and the root mean square error is less than 1%. Comparing the prediction results of the proposed model, multiple linear regression model and XGBoost model, it is proved that the proposed model has higher accuracy and prediction effect.