Risk Analysis Of Electricity Market Based On GARCH Model

In power market, the policy of separation between network operations and generation, and making electric generation market competitive are implemented. Generation companies have to bid before market participation so as to obtain profit as much as possible. Generation companies expect to maximize their profits through specific bidding strategies. However, the uncertain factors in power market make it difficult to measure their risk quantitatively. Therefore, Generation companies all expect to either lower their risk under fixed interest, or maximize their profit under fixed risk. This paper studies on the forecast method of market clearing price and the measurement method of risk in power market caused by price volatility.Price is an economic gear to supply and demand which holds the balance between them. Price forecast is one of the basic requirements for strategies optimization. However, forecasting the electric price is a difficult task as the electric price is affected by many factors. The accurate forecast of uniform clearing price is rather benefit for the generation companies to make suitable bidding strategies. In this paper, a novel forecast method which combines the Markov chain and back-propagation neural networks is presented. With the feature of the transferring probability, the weighted Markov chain could effectively show the affection of random factors, and the back-propagation neural networks has a strong adaptability, the combine method is used to forecast the uniform clearing price. Results of calculation examples show that the proposed method is practicable.Risk for the generation companies mainly comes from the volatility of market clearing price. The risk and also risk measuring methods are introduced firstly. With the advantage of GARCH(Generalized Autoregressive Conditional Heteroskedasticity) in analysis time series, electric return series is concluded by handling the logarithm in the market price series in PJM. The analyses of these return series show that the heteroskedasticity feature is exist in them. Akaike Information Criteria and Schwarz Criterion are used to choose suitable autoregressive conditional heteroskedasticity model which captures the volatility of the return series. And then, the standard deviations of return series in different distribution functions are obtained. Under the given confidence level, the effectiveness of different models are comprised within different distribution function and different distribution function of return series. Even more, the Value at Risk(VaR) and Conditional Value at Risk(CVaR) of the return series are obtained At last, the probable bidding strategies for power suppliers will be then analyzed according to the risk obtained by VaR and CVaR which could measure the power market risk quantitatively.