基于贝叶斯优化和长短期记忆神经网络（BO-LSTM）的短期电力负荷预测

考虑到气象因素对电力短期负荷预测精度存在影响，提出了一套贝叶斯优化和长短期记忆神经网络（BOLSTM）组合预测模型。通过贝叶斯优化算法优选出全局最优参数组合，再采用优选出的五类气象因素数据（日最高温度、日最低温度、日平均温度、日平均相对湿度、降雨量）以及实际电力负荷数据作为输入特征量对优化后的LSTM神经网络进行训练。最后对某地区的电力负荷数据进行预测分析，并与不同方法对比分析，证明了考虑气象因素后的BO-LSTM神经网络预测精度高，可以作为可靠的短期电力负荷预测工具。

Considering that meteorological factors have a great influence on the accuracy of short-term power load forecasting,Bayesian optimization and long and short-term memory neural network(BO-LSTM)combined model is proposed. The global optimal hyperparameters is obtained by Bayesian optimization algorithm and then five kinds of meteorological factors(daily maximum temperature,daily minimum temperature,daily average temperature,daily average relative humidity,rainfall)and actual power load data are selected as input features to train optimized LSTM neural network. Through the prediction analysis of daily load data in a certain area and comparative analysis with different methods,it is proved that BO-LSTM model has high prediction accuracy and it can be used as a reliable short-term power load forecasting tool.