基于FA-GRU模型的短期电力负荷预测

FA-GRU Model-based Short-term Load Forecasting

针对电力负荷随机性强、稳定性差的问题,提出一种结合因子分析(FA)和门控循环单元(GRU)的电力负荷预测方法。首先,利用因子分析探寻负荷数据的内在逻辑,找出各类条件因素(如时间、电价、温度等)对用电负荷的影响并降低数据规模和复杂程度;接着,将因子分析与门控循环单元相结合,建立负荷预测模型;最后,对江苏某地区实际负荷进行预测,结果表明与未降维的模型相比,降维后的预测模型有着更高的预测精度且训练速度更快。

Aiming at addressing strong stochasticity and poor stability of power load,a load forecasting method combining factor analysis(FA)and gated recurrent unit(GRU)was proposed.First the internal logic of load data was clarified by FA to determine the influences of various conditional factors such as time,electricity price,and temperature on power load and to reduce scale and complexity of data.Then the load forecasting was modeled by combining FA and GRU.Final a case forecasting on the regional actual load in Jiangsu was carried out,which demonstrated higher forecasting accuracy and training speed of the proposed model compared with the model without dimension reduction.