Research On Forecasting Algorithm And Application Of Short-term Electricity Price In Electric Power Market

The key to the success of power reform is electricity price,which is the core and lever of electricity market.It is the signal carrier that transmits the information of supply and demand in electricity market and directs the energy flow in the entire power market.Compared with point prediction and interval prediction,the result of probability density prediction of short-term electricity price can not only reflect the fluctuation characteristics of future electricity price and possible fluctuation range but also provide approximate probability of possible short-term electricity price on the basis of forecasting short-term electricity price.Therefore,it's very significant to study the model and method of forecasting the probability density of short-term electricity price.It can make the best production strategy for the generator,and the optimal quotation strategy for the supplier.It also provide more and more detailed useful information for the user's dynamic control of the cost of electricity.It provides an important scientific basis for the supervision of real-time supervision department.In this paper,the method of combining support vector machine with quantile regression is used to predict the electricity price.Combining this method with wavelet analysis,the probability density of electricity price is obtained through the distribution of electricity price probability under different sub-sites.In this paper,we use the electricity price and load data in Singapore as an example to forecast the short-term electricity price probability density.The results show that the support vector quantile regression method combined with wavelet analysis can solve the short-term price forecast probability density problem well.The paper also makes a deep research on the predictive model of short-term electricity price using the quantile regression of support vector quantile with different kernel functions,and analyzes the influence of kernel function on predictive performance of support vector quantile regression model.The results show that using radial basis.Support vector quantile regression algorithm of kernel function is superior to the linear kernel function on short term price probability density prediction.The research results have certain guiding significance for the short-term price forecasting.