Research On Short-term Prediction Analysis Of Time Series Data In Smart Grid

Power load forecasting is an important part of power system scheduling, operation, panning and is the basis of power system economic operation and the safe operation. With the deepening of the smart grid, how to utilize the huge amounts of data from the smart meters for short-term load forecasting, overcoming the characteristics of uncertainty and complexity to get accurate load forecasting results, is a significant research subject. Considering many years of experience, statistical forecasting methods and artificial intelligence forecasting methods obtain good prediction effect in short-term load forecasting, but some of them overlook the temporal characteristics and behavior similarity of customers.Clustering algorithm can cluster the data or groups with similarity and the accuracy of load forecasting can be improved after clustering. Therefore, for temporal characteristics of smart meter data, this article based on regional and customer level short-term load forecasting on smart grid demand side, makes a research about methods of short-term load forecasting about time series data. Specifically, the following research is carried out like:In the face of time series data and similarity of load data in smart grid, this paper proposes fuzzy c-means algorithm with the wavelet-kernel regression approach. Due to the weather and date factors having great influence on load forecasting, this paper selects these factors to calculate the similarity with forecasting daily load curve, which can improve the regression prediction model. Aiming at the load data of similarity on customer level, this paper presents fuzzy c-means algorithm with wavelet neural network prediction algorithm.The experimental results show that fuzzy c-means algorithm with the wavelet-kernel regression approach and fuzzy c-means algorithm with wavelet neural network prediction algorithm have a high precision and adaptability of prediction when fully consider temporal characteristics and behavior similarity of customer, which provides a new idea for the electric power system short-term load forecasting.