Research On Price Forecast Based On PSO-Neural Network And Bidding Strategy Of Hydropower Plant

The market-oriented reform of the electric power industry is a develop trend of the world. Electrovalence is the economic lever of the power business under the power market environment which relate to the benefit of the market participator. Exact price forecast is the basic of bidding strategy and the key of the generate electricity corporation to participate in compete and debase the risk. To get the bidding strategy is the key for the generate electricity corporation's benefit and exist under the power market environment according to the addition of the power market of our country. This paper was research on the price forecast and bidding strategy of the hydropower plant.The paper analyzed the form of the price and factor that influence it, the characteristics of the periodicity, different change rule of the working day and weekend, price fluctuate under large load and rule of the "price nail" to appear etc. were researched. The method that chose the forecast model and arithmetic in reason according the variety characteristic and orderliness was put forward to advance the precision of the forecast.Neural Network has the trait of highly nonlinear fitting ability, can change the weight to get a quick reaction according to the change of the input, owing to the trait of the Particle Swarm Optimization algorithm that good at random global optimization, two algorithms were united in the paper, the configuration of three lays Neural Network was used to found a PSO-BP model, the Neural Network Toolbox of Matlab7.0 was used to realize it. Used Particle Swarm Optimization algorithms to optimize the connection weight of the net, Particle Swarm Optimization algorithm was used to confirm the initial value then achieve the definite precision by the study of the BP networking. It was proved that the model has a good forecast effect and resolved the problem that the weekend price should be modeled singleness in traditional method, it can nicely forecast and adaptability well etc.It was found that the price and load were not always connected closely by analyzed the pertinence of them. The formal forecast model was mended by inducted the pertinence coefficient to be the weight of whether to induct the load. When the price and the load connected closely and have large pertinence coefficient the load was inducted to be input of the model to add the sensitivity of the price fluctuant of the model. It was proved that the improved model have a higher forecast precision than the former, and a good effect for the biggish fulctu- ate weekend price and 'price nail', it resolved the problem of short time price forecast effectively. Compare to the chaos theory model and grey theory model the improved model has obvious advantage with few iterate time, higher forecast precision etc.The bidding strategy model was set up based on the price that forecasted by aforesaid method, and the model was solved by genetic algorithm. Most constraint conditions of hydropower plants was considered in the model, the target to get the biggest profit of the plants was achieved by assign the output of every period of time reasonable. And the model was proved effectively at improve the profit of the hydropower plants.