Power load control technology under the framework of deep data mining and cluster analysis
In order to improve the service level of important users in the distribution network area of the new power system,this paper studies the power load control technology.By training high-precision and high-efficiency load prediction networks,real-time guidance is provided for load control.A clustering algorithm based on Dynamic Time Warping(DTW)similarity was introduced to preprocess user data based on historical electricity consumption trends,voltage levels,access methods,and other information of users in the substation area.By pooling similar users,the efficiency of deep mining was improved.The Deep Neural Network(DNN)is used as the nonlinear prediction fitting algorithm of user load,and the Cross entropy function is used as the Loss function of the network,which improves the efficiency of DNN error direction propagation iteration.In order to evaluate the performance of the algorithm,the proposed method is tested based on the actual load data of a power supply company.The simulation results show that compared with the traditional K-means algorithm,the clustering efficiency of the DTW algorithm is better than the consistency between the change curve and the actual data,and the MAPE and RMSE of the DTW-DNN algorithm are increased by 6.9% and 3.96% respectively compared with the ordinary DNN network.
cluster analysisDNNDTWdeep learningload forecastingload control
NETL
NSTL
万方数据
