Short-term Power Load Forecasting Based On Fractal Theory

Power load forecasting is one of the most important work in the power supply unit, Because reasonably predict the future load of hours, days or a month is the premise for scheduling department to make electricity generation plan. We can only make accurate, reasonable load forecast of a region ahead of schedule to ensure that the region's production and the normal life.Because the load curve has similarity in different time of the same space, so this paper can forecast the future load with the load data of similar days. And this paper uses the theory of the gray correlation analysis method to select similar days. After select similar days, this paper uses the fractal theory of fractal interpolation method to establish short-term load forecasting model. But there are two very important problems need to solve with the algorithm in the fractal interpolation method, one is a fractal interpolation algorithm of parameters, and the other one is a fractal iteration algorithm for interpolation generated curve. This paper use genetic algorithm and optimized fractal certainty iterative algorithm to solve the two problems. In addition, this paper also improves the two kinds of solving algorithm. The first is improvement of the basic genetic algorithm, and there are two places worthy to be improved. The first place is improvement of the objective function, which use a global fitting errors instead of local fitting error to get the optimal fractal interpolation parameters. And the other is the improvement of its selection operator and crossover operator. This paper use ideological choice of random league with the optimum selection method instead of proportion of choice, and make comparision to fitness of the crossover individuals, then keep the highest fitness individuals. The purpose is to protect the best individual not to be damaged; The second is improvement of the certainty iterative algorithm, and the improve thought is the initial iteration set improvement for iterative initial point, because using the initial set to make iteration can produce a lot of repeated points. But using the initial point, there won't be a lot of repeated points to be produced, also it can enhance iterative efficiency.After analysising algorithm, it will have to establish forecasting model. Based on improving algorithm, it will set up whole fractal basic prediction model and overall fractal improved prediction model in this paper. Finally considering the curve characteristics of power load and various meteorological factors on the effect of load, this paper also puts forward of the fractal basic prediction model in different period and fractal improved prediction model in different period. And comparing the pros and cons of four prediction models by using an example, to make feasibility analysis of the improved prediction model.