The Research And Analysis Of The Ultra Short Term Load Forecasting Under The Influence Of Electrified Railway Load

This topic is the main content of the scientific and technological project of a local power company which concentrates on the analysis of the ultra short term load forecasting under the influence of the electrified railway load. The electrified railway load, as an important factor influencing that local load forecasting precision, is a huge difficulty for the local work of load forecasting. In response to this problem, this subject firstly makes a load analysis of that area, then makes a appropriate plan of data processing according to the result of load analysis. To realize the function mentioned above, it also introduces an improved intelligent method as the prediction method and programmed an operational software suitable for actual application in this process. After the test on actual data in the end, this method is proved to be an efficient way to improve the load forecasting precision under the influence of the electrified railway load.Load analysis and data processing are the important preparatory work for load forecasting. So this article firstly takes the load characteristics of the electrified railway as the starting point to make a comprehensive analysis of the load characteristics of the electrified railway and the power load variation of the power network respectively. The daily characteristic load of the power network under the influence of the electrified railway load is drawn according to that. And on the basis of that it eliminates the wrong load data with the method of load preprocessing and it solves the periodical problem of the load variation by determining the training samples through the approach to the similar problems. Finally it introduces the methods of wavelet and reconstruction which divided the data into the approximation and details and thus provides a data security to lower the forecasting error.Secondly, with the study of ultra short term load forecasting as the basis, this article adopts support vector machine, a frequently used method at present, as the prediction method. And it achieves the purpose of increasing the prediction precision by improving the kernel parameter of the model. After determining the prediction method, this article designs an applied software which integrates the data processing, index calculation on load characteristic, wavelet, reconstruction and the ultra short term load forecasting as a whole. The MATLAB supplied with GUI is used according to the practical requirement of the electricity company.Finally, this article makes a prediction with the newly-made software by treating the data in test area in2012as the original data. Also it makes a analysis about the prediction precision and its service cycle. The results turn out to be that it satisfies the related requirements of the ultra short term load forecasting of the State Grid with the average relative error lower than2.0%and the overall prediction precision reaching98.0%in the ultra short term load forecasting.The developed software introduced in this article has been appraised and checked by the related department, and has been put into practice in the local power company. The software is of certain spreading value with the feature of practicability, easy installation and easy operation.