Research On Improved Algorithm Of Support Vector Machine Based On Particle Swarm Optimization

With the development of smart grid, how to build a reasonable demand side response to guide users to participate in the management of smart grid, which has become the focus of the development of the power industry. As an important part of the demand response and the main development model of future electricity price, the real time electricity price has become an important research object of the current power industry in the world.At present, the research direction of real-time price mainly has two aspects, one is the spot price forecasting model based on optimal power flow, the other is a real-time price forecasting model based on Intelligent algorithm. Due to the research model based on optimal power flow data relates to the power supply company’s trade secrets, so data acquisition is very difficult, resulting in the most of the studies are not real data do rely on, the specific implementation effect has yet to be verified. And real-time pricing as one of the short-term marginal price, the predicted price are short period, so you can ignore grid electricity price is affected by uncertain factors, using intelligent algorithms to the spot price to make the forecast. This price forecasting model is verified by a large number of examples, is currently the main way of real-time priceThe through investigation and study at home and abroad intelligent grid and demand response development, under the environment of our country is about to open the sale side, will be the introduction of market rules to the power market proposed established a support vector machine real-time electricity price forecasting model. First of all, this paper summarizes the existing pricing rules and types of electricity in China, introduces the development process of real time pricing, and the supporting system and user’s demand response to the implementation of real time electricity price in smart grid. Secondly, introduced the model of support vector regression, and establish the real-time pricing model, simulation is carried out on the same data, confirmed by time series similarity search method of data mining, which can improve the quality of data, enhance the prediction accuracy of price model; finally in view of the randomness and inefficiency of parameter selection the traditional support vector machine model, which will cause the reason of error prediction, a prediction model of support vector electrical price based on particle swarm optimization algorithm, by using particle swarm optimization algorithm to parameter optimization of support vector machine, and using the optimal parameters to establish predictive real-time pricing model with the highest accuracy finally, the simulation examples verify the feasibility and superiority of the model.