Research And Implementation Of Short Term Load Forecasting Under Smart Grid Environment

Short term load forecasting plays a vital role in power system’s operation and control process, which is important to ensure power system’s safety, stability and reliability. The sustainable development of environment is the base of society development, smart grid is the next generation of power system which aims at providing power frugal, efficient, eco-friendly, safe, and reliable electric power. Green energy like wind power and solar power and so on are the most important component in smart grid, which reduce the consumption of conventional energy and protect the environment. However, because of the fluctuating nature, they can be used widely, the load forecasting under smart grid become one of the most important factors which help utilize green resource fully. Short term load forecasting is constantly a research issue for experts in electric area, no matter under conventional environment, but also under smart grid environment.In this paper, we propose memetic algorithm optimized core vector regression model to apply on short term load forecasting and wind power generation problem. Firstly we get the most important factors which determine the load curve by analyzing the nature of electric load based on real load data and we find the date type and temperature are the most important ones among them. Then we construct the samples based on the load nature. Core vector regression is the proposed method to solve short term load forecasting problem in this paper which is proven an effective way in kinds of regression problem. Parameters selection determines the model complexity and forecasting accuracy when employ core vector regression model. Memetic algorithm is the method we propose to optimize parameters of core vector regression model. We analyze the two factors which affect the optimization effect, population size and local search threshold. The experimental results illustrate that proper population size is important to keep population diversity and proper local search threshold is important to avoid falling into local optimum. After the memetic algorithm optimized core vector regression model is built, we apply it under both conventional environment and smart grid environment which consider wind power generation. With lots of experiments, we can draw the conclusion that the proposed sample construction way based date type and temperature and proposed memetic algorithm optimized core vector regression forecasting model are practicable. Compared to genetic algorithm, the proposed model is better on solving parameter optimization problem of core vector regression and prediction accuracy.