摘要
针对传统短期电力负荷预测模型特征获取能力欠佳、预测精度低的问题,提出一种SSA优化CNN+BILSTM-LSTMAttention的双通道短期电力负荷预测模型。该模型构建多层CNN与BiLSTM双通道结构,提高模型的特征获取能力;利用灰色关联分析筛选温度、湿度等气象影响因素参与模型训练,同时,利用以均方误差最优为目标函数的SSA优化算法,自适应选取迭代次数、时间步等模型超参数,提高模型的预测精度。通过实验对比表明,该模型在误差指标和拟合优度指标两方面较BiLSTM-Attention、Attention-BiLSTM-LSTM等模型均有提升。
To address the problems of poor feature acquisition ability and low prediction accuracy of traditional short-term power load forecasting models,a SSA-optimized CNN+BILSTM-LSTM-Attention dual-channel short-term power load forecasting model is proposed.A multi-layer CNN and a BiLSTM dual channel structure is constructed in order to improve the feature acquisition capability of the model.Grey correlation analysis is used to screen meteorological influences such as temperature and humidity to participate in model training.The SSA optinization algorithm with the optimal mean square error as the objective function is used,model hyperparameters such as number of iterations and time steps are selected adaptively in order to improve the model forecasting accuracy.Experimental results show that the model is better than the BiLSTM-Attention and Attention-BiLSTM-LSTM models in terms of both error metrics RMSE,MAE and goodness-of-fit metric.
作者
吴迪
段晓旋
马超
WU Di;DUAN Xiaoxuan;MA Chao(Hebei University of Engineering,School of Information and Electrical Engineering,Handan 056038,China)
出处
《河北电力技术》
2023年第4期35-42,共8页
Hebei Electric Power
基金
国家电网有限公司科技指南项目(5600-202019167A 0000)
河北省自然科学基金项目(F2020402003)。
关键词
短期电力负荷预测
双通道模型
麻雀搜索算法
多因素筛选
超参数优化
electricity load forecasting
dual-channel model
sparrow search algorithm
multi-factor filtering
hyperparameter optimization
