摘要
母线负荷基数小,波动性和不确定性大,随着光伏、风电等可再生能源的接入,母线负荷受天气等随机性因素的影响增加,母线负荷的高精度预测受到很大影响。针对小样本场景下母线负荷预测问题,提出了一种基于离散小波变换-多目标黏菌算法-支持向量机(discrete wavelet transformation-multiple objective slime mould algorithm-support vector machine, DWT-MOSMA-SVM)的多目标优化短期母线负荷预测方法。首先采用离散小波变换对母线负荷数据进行处理;然后兼顾预测的精度和稳定性两个目标函数,采用多目标黏菌算法对支持向量机的惩罚因子和核函数参数进行优化;最后在优化所得的Pareto前沿面上选择Pareto最优解,以此搭建支持向量机(support vector machine, SVM)预测模型进行训练,并将预测结果与长短期记忆网络(long short-term memory, LSTM)、未优化的SVM以及多目标黏菌算法(multi-objective slime mold algorithm, MOSSA)优化的SVM模型预测结果进行对比。实验结果表明,提出的MOSMA-SVM模型的预测精度和稳定性更佳。
The bus load has small base, high volatility and uncertainty. With the access of renewable energy such as photovoltaic and wind power, the bus load is increasingly disturbed by random factors such as weather, and high precision forecasting of bus load is greatly affected. Aiming at the problem of bus load forecasting in small-sample scenario, this paper proposes a short-term bus load forecasting method based on discrete wavelet transformation-multiple objective slime mould algorithm-support vector machine(DWT-MOSMA-SVM). Firstly, discrete wavelet transform is used to process bus load data. Then giving consideration to the accuracy and stability of the forecasting, multiple-objective slime mold algorithm(MOSMA) is used to optimize the penalty factor and kernel function parameters of SVM. Finally, selecting the Pareto optimal solution on the Pareto front, SVM forecasting model is built for training. The forecasting results are compared with those of LSTM, un-optimized SVM and MOSSA-optimized SVM model. Experimental results show that the proposed MOSMA-SVM model has better forecasting accuracy and stability.
作者
陈逸枞
张大海
李宇欣
王颖
CHEN Yicong;ZHANG Dahai;LI Yuxin;WANG Ying(School of Electrical Engineering,Beijing Jiaotong University,Beijing 100044,China)
出处
《电力建设》
CSCD
北大核心
2023年第3期49-55,共7页
Electric Power Construction
基金
国家自然科学基金青年基金项目(52107067)。