Improved Artificial Bee Colony Algorithm Combined Model For Middle And Long Team Load Forecasting

The urban planning is based on power system construction, and the power loadforecasting is the basis of the power system planning. The medium and long-term loadforecasting play an important role on developing of power system, the higherprediction level is, the more benefit comes to the optimal management and morereasonable for the electric power constructing. It is also more to the benefit ofimproving the social benefit and economic benefit. So the study on medium andlong-term load forecasting has important theoretical significance and practical value.The middle and long term load forecasting affected by many factors an dpredicted over the span of nearly a decade. There are many load forecasting method inrecent years, however, each forecast model can only simulate the effects of one or afew factors, it is difficult to obtain the satisfactory predicted results. The comb inationforecasting method can synthesize the predicted results for multiple forecastingmethods, it is more comprehensive than a single forecast model. The key to get theaccurate predicted results is that whether we can determine the best weight value ofeach single model in the combination forecasting.In response to this problem, combination of ABC algorithm with the combinationforecasting model is applied to the medium and long term power load forecasting inthis paper, used for optimizing the weights of combination prediction in objectivefunction. ABC algorithm is a kind of swarm intelligence algorithms based on theprocess of bees finding food source. This algorithm is more flexible and has betterability of global optimization compared to Genetic Algorithm (GA) or Particle SwarmOptimization (PSO) and other evolutionary algorithms. However, it has the problemof slow convergence speed in the early and easy to fall into local optimum in the late.In order to solve this problem, the method proposed a disturbing term and worsthoney substitution that can overcome the problems of convergence rapid and localoptimum of the existing ABC algorithm. The results research shows that thedeficiencies are significant improved in the proposed method.