电价预测对于国家电力市场的销售价格，电力调度和市场波动管理具有重要意义，但现有方法在电价预测的准确性上不理想。为了进一步提升电价预测的准确性，提出一种基于改进完全自适应噪声集合经验模态分解(ICEEMDAN)，贝叶斯优化(BO)和长短时记忆网络(LSTM)的短期电价预测模型。ICEEMDAN将原始数据分解为多个本征模态函数(IMF)和一个残差序列，然后将IMF分量重构为高频，中频和低频三个子序列，将子序列和残差序列分别与相关因素结合，重构为四个多维特征矩阵，输入BO-LSTM模型进行训练，最后得到预测结果。用西班牙国家电网公司Red Electric España运营数据进行算例分析，结果表明ICEEM-DAN-BO-LSTM模型具有更高的准确度，在电价跳跃点和峰值点处预测结果表现出色，与其他方法相比预测效果更好，对能源企业和国家电力市场调控策略具有实用价值。
Short-term electricity price prediction based on improved CEEMDAN-BO-LSTM
Electricity price prediction is crucial for the sales pricing,power dispatch,and mar-ket volatility management in the national electricity market.However,existing methods are often inadequate in terms of accuracy.To enhance the accuracy of electricity price predic-tions,a short-term electricity price prediction model based on Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN),Bayesian Optimization(BO),and Long Short-Term Memory networks(LSTM)is proposed.ICEEMDAN decompo-ses the original data into multiple Intrinsic Mode Functions(IMFs)and a residual sequence.The IMF components are then reconstructed into three sub-series:high-frequency,mid-fre-quency,and low-frequency.These sub-series and the residual sequence are combined with re-lated factors to form four multi-dimensional feature matrices,which are then input into the BO-LSTM model for training.The final prediction results are obtained from this process.Case studies using data from Red Eléctrica de España,the Spanish national grid operator,demonstrate that the ICEEMDAN-BO-LSTM model has higher accuracy.It performs excep-tionally well in predicting price jumps and peak points,and it outperforms other methods,making it valuable for energy companies and national electricity market regulation strategies.
万方数据
维普
