The Research And Implementation Of Multi-Variable Power Load Forecasting Technology

In recent years, as an important component of the rapid development of things, the smart grid is provided an unprecedented opportunity for development. As an important research content of smart grid, High-precision load analysis and load forecasting can optimize power generation, transmission and distribution process, improve the security and economy of grid operation, and improve the quality of electricity supply.At present, the research on electric power load forecasting has been greatly developed, but the law of power load diversification is so complex and stochastic that technology research related to load forecasting will be a long process. The process of power load change is affected by several factors, especially the weather. It will have a very important practical significance in research of factors associated with power load, and formats multi-variable information in order to reflect the variation of load.This paper will use the fine load data provided by smart meters, combined with meteorological data collected from sensors and information from Internet provided by third parties, to constitute multivariate data used in this article, and analysis and forecast the power. The research of power load forecasting should be done, the evaluation criteria and methodology of load forecasting results should be formulated. After obtaining raw data, data pre-processing is needed. The preprocessing data need correlation analysis. It contains the envelope-based segmentation slope similarity matching algorithm, and compares with Grayscale correlation and Pearson correlation coefficient algorithm. Thus the best multivariate sample is constituted, as a predictor input. Before the establishment of prediction algorithm, this paper uses clustering algorithm to analyze multivariate sample sets, and provides foundation for prediction model. Furthermore, in order to obtain accurate predictions, different prediction algorithm models should be established. This paper discusses the algorithms of Support vector regression and combination forecasting model, and describes the application in power load forecasting, and compare the forecast accuracy under a variety of methods by a large number of experiments.Finally, this project develops visual system platform for load analysis and forecasting, and use C++ comply the load-related model algorithm. Efficient Operation provides real-time, accurate electrical load analysis and forecasting. This paper elaborates system development technologies, and carries on system requirements analysis, overall design, design and implementation of functional module. The project will implement the algorithm model as an independent subsystem, and contact visualization platform through the data communications. It has advantages on scalability of the system.