Study On Substation Peak Load Shifting Plan Based On Load Forecasting

Since twentieth century, our country’s power grid has entered a new stage and electric power has become an essential part of the economic development and people’s life. However, driven by rapid growth in electricity demand, the country’s major cities have the power shortage, with the power rationing appearing in many cities, which not only affects people’s life and causes huge economic losses for power companies and their customers, but also brings a severe test to the reliability and stability of power grid.In this context, the implementation of power demand side management is imperative. Power demand side management refers that the Power Company and customers to configure the power market together, in order to improve the reliability of power system, reduce energy consumption, and achieve the best economic benefit. So far, as the most effective means of demand side management, the holiday plan of substations, is still formulated by experienced staff that doesn’t meet the economic efficiency requirements of smart grid era. As the premise of a reasonable arrangement of holiday plan, how to improve the accuracy of load forecasting has always been the key research on the load forecasting.In the above analysis, this paper proposes a power load forecasting method based on Gaussian process random process and a generation scheme of holiday plan of substations based on the TABU search algorithm. Programmed in MATLAB and VS2010 environment, it proves a good solution to the limitation and low economic of the traditional method. The main research contents are:(1) Based on the continuity and trend similarity of the load data,this paper proposes a load forecasting method based on the theory of Gaussian regression. Gaussian process refers to a set of random variables, in which the collection of arbitrary finite random variable obeys Gaussian distribution, determined by the mean function and the covariance function. Firstly, by means of supervised training of the historical data, the prior distribution is established based on Bayesian theory. Secondly, under the condition of the pre-forecasting data, the prior distribution is converted to the posterior distribution. The kernel functions and hyper-parameters of the Gauss distribution in the training process can be updated in real time. Finally, the case shows that the method greatly improves the accuracy of load forecasting.(2)This paper analyses the shortcomings of existing models for peak load shifting plan and describes it as a NP(NP-Complete) problem. Then establish the optimization model, and find the global the optimal solution based on the TABU search algorithm. The case shows that the method is better than the traditional manual method in terms of the economic.