Long-term Power System Load Forecasting Research And Application

Mid-and-long term load forecasting of power system, served as the base of formulating electric power development plan, is one of the most important tasks for power department. Improve the mid-and-long term forecasting level, is beneficial to the optimal management of planed use of electricity, to the proper plan of power construction, and to the economic and social effect, hence its significance to power system.The power system capacity is influenced by many direct and indirect factors, so the change of it is complex, whose pattern is difficult to describe by a single forecast model, and the precision is changeable over time and even will be far apart from the actual value due to fierce fluctuation. Combination forecast model integrates multiple forecast methods which complements each other, and can improve the precision while reduce the risk. However, traditional combination methods always allocates equal weight to each single one, and lack flexibility when precision changes, so the forecast result is not the optimal one in overall situation.On account of theses problems, this paper provides a combination method based on Inducing Ordered Weighted Geometric Average Operator (IOWGA) and Stack Markov Chain (SMC). In this method, weight is allocated according to the forecast precision of each single forecast method in each year. The weighting is determined by precision instead of each single method, since larger weight is allocated to more precise forecast method. So the previous problems are solved by the correspondence of weight coefficient and fit precision in combination forecast model.The basic assumption of this forecast model is to establish a combination model via IOWGA. IOWGA is a special kind of operator which can allocate weight to each single method in sequence, according to their precision at each time point. SMC divides the object into multiple statuses, forecasts the development tendency on the basis of transition probability from the original state and possible state, and regards forecast precision as a random process. The forecast precision of single forecast method within forecasting year can be inferred quantitatively through SMC, and the weight allocation of single forecast method is then inferred based on the precision level of single forecast year.This paper expounds on the theoretical background, characteristics, modeling thoughts and calculating procedures of this forecast method. A forecast of electricity consumption of one province is conducted by applying this method, and the forecasting results are analyzed by comparison with that of ordinary combination methods. It proves that the present method has particular advantages as well as practical values.