Research On Improved GM(1,1) Load Forecasting Model Based On Numerical Analysis

Power system load forecasting is an essential foundation in power system planning and operation of work. Accurately forecast electric load is an important feature to ensure that provide reliable power supply for the various departments of the national economy and people’s lives, at the same time, it is the first condition to ensure the power industry to stable development. There are a lot of load forecasting methods, however, lots of them exist varying degrees of inadequacies, therefore, it is a critical work that make a reasonable choice of forecasting methods and as much as possible to improve the existing methods.Based on the gray theory GM (1,1) load forecasting model, and aim at the inadequacies in this model, this article proposed a new GM (1,1) load forecasting model improved by numerical analysis algorithm. The main contents are as follows:1、According to analysis of the modeling mechanism of GM (1,1) model, consider that the gray power system load forecasting model, essentially, is a index prediction model, therefore, the smoothness of original series and the variation of forecasting object affect the model precision directly; The series of load forecasting is discrete series, GM (1,1) gray model differential fitting a fitting equation is a discrete approximation of differential equations, the gray GM (1,1) fitting equation is difficult to consistent with the differential equations of actual system. Therefore, it will produce the inherent error when we modeling the model in this way.2、To solve the above problem, this paper proposes a method of adjustable logarithmic transform, in order to improve the series smoothness, use the improved smoothness data to modeling and calculate, it can effectively avoids the inherent error of the model; In addition, introduce a method of differential equation numerical algorithms when building the fitting equation of gray GM (1,1), the introduction of the differential equations of the numerical algorithm, so, the model can be more effective connection with differential equations and greatly enhance the degree of approximation between the model and the system differential equations.Compare with traditional GM (1,1) model, the accurate of improved model was improved greatly, at the same time, by adjustable logarithmic transformation of data to further improve the scope of model application. After the example calculation calculation results, verify the validity and correctness of improved method of this article GM (1,1) model.