基于改进LSTM的短期电力供需负荷预测方法研究

Research on Short-Term Power Supply and Demand Load Forecasting Method Based on Improved LSTM

为提升短期电力供需负荷预测过程的泛化能力,获取更准确的短期电力负荷预测结果,提出基于改进LSTM的短期电力供需负荷预测方法。先采用聚类算法检测电力供需负荷数据内的异常数据,并采用灰色理论修正数据。然后构建长短时记忆(LSTM)网络预测模型,将修正后的电力供需负荷数据内29个特征值作为模型输入。最后参考模拟退火算法概率突变理论改进粒子群算法,利用改进后的粒子群算法求解LSTM预测模型内主要参数的最优值,通过运算输出后一天24个整点电力供需负荷预测结果。实验结果表明:所研究方法可有效检测异常数据,提升短期电力供需负荷数据预测模拟的精度与泛化能力。

In order to improve the generalization ability of short-term power supply and demand load forecasting process and obtain more accurate short-term power load forecasting results,a short-term power supply and demand load forecasting method based on improved LSTM is proposed.Firstly,the clustering algorithm is used to detect the abnormal data in the power supply and demand load data,and the gray theory is used to correct the data.Then,a long short term memory(LSTM)network forecasting model is constructed,and 29 eigenvalues in the revised power supply and demand load data are used as model inputs.Finally,the probability mutation theory of simulated annealing algorithm is utilized to improve the particle swarm optimization algorithm,the improved particle swarm optimization algorithm is used to solve the optimal value of the main parameters in the LSTM forecasting model,and the 24 hourly power supply and demand load forecasting results of the next day are output through the operation.The experimental results show that the proposed method can effectively detect abnormal data and improve the accuracy and generalization ability of short-term power supply and demand load forecasting simulation.