Curve Fitting Principle And Its Application Research

Curve fitting refers to choose the appropriate type of curve to fit the life,society,and other activities,the data which from life,society,and other activities,and to fine the inherent laws of the data.People could use the curve fitting methods' results which from the power system load forecasting and nonlinear correction fields to let the plan of life and production to be more effective,and achieve greater economic and social benefits.Therefore,it has very important practical value.Firstly,this paper constructs the polynomial fitting model.Three kinds of curve fitting polynomial model were established according to the different kinds of polynomial.Three kinds of polynomials are orthogonal polynomials and widely used in contemporary analytical techniques,compared to a polynomial is more suitable for curve fitting.Secondly,this paper introduces three kinds of the currently the most widely used algorithms,including conjugate gradient method,gradient descent method and recursive least square method.The paper uses the optimization algorithms to calculate the polynomial fitting model parameters,the structure parameters as the weights of neural network,and the real data as the training sample.Using the curve fitting algorithms for training the weights of neural network,the method obtained polynomial model parameters.Thirdly,the power system load forecasting method based on combination of Hermite polynomial curve fitting model and recursive least square method was proposed in order to improve the prediction accuracy about the power system load forecasting.Beijing city electricity data was chosen as the sample data for training.Using recursive least square to calculate the parameters of polynomial fitting model,and then to realize power system load data fitting by using trained parameters.The method not only can effectively predict the load's change trend,but also has good accuracy.The compound correction model based on Laguerre polynomials of nonlinear correction and temperature compensation was proposed for the problem of nonlinear correction.The measured values of humidity sensor and the measured values of liquid level sensor in each type of temperature were selected respectively as the data sample.The method could use recursive least square to calculate the actual data with high accuracy.The simulation results showed that the proposed method has better prediction accuracy and fitting effect than the traditional prediction model in power system load forecasting,and has the ability to predict medium and long-term load.The proposed method has good nonlinear correction and temperature compensation effect.Therefore,the methods of this paper have important application value in the field of power system load forecasting and nonlinear correction.