基于机器学习的源荷互动微电网优化调度

为了发挥微网电力市场的活力,实现清洁能源的优化配置,提高微电网消纳率,在融合半监督K-means聚类分析方法和支持向量机2种机器学习算法的基础上,提出了微电网源荷协调优化调度方法.首先利用改进的K-means聚类算法对源荷历史数据进行预处理.其次运用SVM对聚类后的微电网源荷数据进行预测,在预测结果中选取典型场景,以典...

Optimal Scheduling of Source-load Interactive Micro-grid Based on Machine Learning

To give full play to the vitality of the micro-grid power market,realize the optimal allocation of clean energy,and improve the absorption rate of micro-grid,based on the fusion of two machine learning algorithms such as semi-su?pervised K-means clustering analysis and support vector machine(SVM),an optimal scheduling method for source-load interactive micro-grid is proposed.First,the improved K-means clustering algorithm is used to preprocess the his?torical data of source and load.Second,SVM is used to predict the clustered source and load data of micro-grid,and the typical scenarios are selected from the prediction results to accurately obtain the probability distribution of wind pow?er output which is represented by the typical scenarios.Afterwards,the total operating cost optimization model of micro-grid is established,which guides users to respond to wind power output through the user demand response,thus in?creasing the coordination and interaction between source and load in the model and improving the matching degree be?tween wind power and electricity load.Finally,simulation results show that the proposed algorithm can reduce the error between the predicted and actual values,improve the real-time prediction accuracy of the micro-grid system,and in?crease the economic benefits of micro-grid.