Power System Temperature-lowering Load Estimation And Mid-long Term Forecasting Method Research

Growing number of air conditioner and frequent hot weather result in cooling equipments turn on more frequently, and temperature-lowing load increases sustainablely. Growing temperature-lowing load has become one of the most important reason that summer peak load sets new record. It also impacts on power system daily load characteristic negatively. Calculating and forecasting temperature-lowing load more accurately can help understand the characterstic of power load, increasing load forecasting accutacy and provide a reference for grid planning.This paper introduces construction of provincial and regional load characteristic analysis index system in three time dimensions including year, month and day according to load data and weather imformation of Guangdong power grid and variable cities from 2008 to 2013. Meanwhile, it analyzes load characterstics of Guangdong power grid and other citiy power grid. On the basis of load characterstics and weather conditiongs, K-MEANS algorithm is applied in cluster analysis on cities of Guangdong province.Considering that traditional temperature-lowing load estimation methods have kinds of deficiency, an estimation method based on meteorological data and entropy weight theory is proposed. The maximum value of annual maximum load curve load date subtract the benchmark curve without temperature-lowering load is annual maximum temperature-lowering load. Meteorological axis system was built based on temperature, relative humidity, precipitation and other meteorological data to determine base meteorological quadrant and filtrate benchmark week-day when calculating base load curve. According to the correlation coefficient between maximum daily load and tem-perature, relative humidity or precipitation on benchmark weekdays, weights of benchmark weekday load curves is calculated, using entropy weight theory. At last, annual maximum temperature-lowering load is estimated according to Guangdong province and cities 2009~2013 load and meteorological data. The result shows that the new method is more appropriate.Except above, An uncertain support vector machine method optimized by entropy and variable precision rough set(VPRS)is proposed in the article for mid-long term temperature-lowing load forecasting. It excavates the relationship between data and eliminating redundant information to search the key input variable of SVM. Based on variable precision rough set and entropy, the method reduces the SVM condition attribute set. According to the reduction result, SVM makes use of the corresponding variable for temperature-lowing load forecasting. The input variables are updated as the forecasting year changes, which could make power grid operating and planning staff pay more attention to key factors in different period. Last, 2012~2020 annual maximum temperature-lowing load of Guangdong province is forecasted, the result shows that the forecasting method is effective and the forecasting accuracy is stable. It is also of strong robustness on impact of various uncertainness. The method is truly a dynamic mid-long term temperature-lowing load forecasting method.