Forecasting Of Transmission Congestion Using Artificial Neural Networks In Electricity Market

With the gradual development of the power market and power system, the efficiency of electricity production is improved, and the whole society gain more from the reform of economic and social benefits. Generation competition, transmission opening, and user selection, has became the three pillars in a competitive power market. As the transmission channel between generaion plants and users, the change of transmission system not only affect the security and reliability of power system, but also affect the power competition and market efficiency. The sources of energy in our country are very uneven: the eastern region is electricity-intensive area. The main power sources used in the eastern region are transmitted form the western area. So in order to make rational and efficient use of the energy, we establish a lot of projects, such as the West-East Electricity Transmission Project, the Long- distance High-voltage Direct Current project, and so on. Meanwhile we should more attention to the transmission congestion because of the more and more long-distance transmission.The traditional research of the transmission congestion focuses on the management for alleviating the congestion. This paper proposes a transmission congestion forecasting model using artificial networks.This paper introduces the main reasons to cause the transmission congestion through the congestion definition and the congestion management. There are several uncertain factors in the power system of market operation, such as: load demand, generation cost, generation plans, and so on, which will lead to uncertainty of the transmission congestion.To clear the degree of correlation between each factor and electricity prices, using the method of one-place linear regression. Electricity prices is the direct manifestation and the important index of the transmission congestion. It is also the important basic source to research the transmission congestion. Through the data from California electricity market in USA and other relevant knowledge, the relation between the electricity prices and the reasons can be learned.According to the knowledge of congestion and ANN, a model of the congestion forecasting is established. This model includes the input layer, hidden layer and output layer. Sampling time point of each factor (such as: the total system load, the system actual output, etc.) and the product of its weight as input layer neurons, and the corresponding time point of the price for the output layer neurons. The data from California electricity market in USA is used to test and verify the validity and practicability of the proposed model.In order to learn deeper about the transmission congestion, a congestion forecastiong model is set up based on time series. Using the historical data within a day as the unit of sample input, and the corresponding time point of the price as the output. In order to learn better about the transmission congestion, a new index as the degree of the transmission congestion is defined to describe it.Similarly, the data from California electricity market in USA is used to test and verify the validity and practicability of the proposed model. The differences between the two models and the advantages and disadvantages were discussed.