Research Of Cleaning Model For Bad Electric Load Data Based On Neural Network

High-quality electric load data is an important guarantee to achieve a load forecast with high accuracy. It is also used as the basis for the electric utilities to analyze the operational status and make the decision about power switching. Bad data in the database may have a bad impact on the safe operation of whole Power System. So making a useful electric load data cleaning model, to identify and adjust the dirty data on time play a very important role in building a Power System with good safety and efficiency.By analyzing the data cleaning methods and procedures, the paper chooses the ART2 network which has obvious advantages in the field of Pattern Recognition and Artificial Intelligence as the basis of data cleaning and put forward an improved algorithm to correct the model drift problem of traditional ART2 network. It limits the model drift by adding a vigilance matrix and Reset System, correct the accuracy of clustering and experimental results demonstrate the effectiveness of the algorithm. On the basis of improved ART2 algorithm, an electric load data cleaning model is established by using GSA algorithm and the combination method together. The GSA algorithm can optimizes the cluster result of ART2 network and it is very suitable for work with neural network. The paper finds the combination point and put the two together to complete the model and retrofits the time-consuming defect of GSA.The proposed model uses the algorithms and steps which have been proven effectively in the field of data cleaning and improves some related features of them, enhancing the accuracy rate of the extraction of characteristic curves and the identification of bad data. and they are proved by Simulation experiment. The paper extends the research field of data cleaning, has some theoretical and practical value.