Neural network weight training method and application thereof

Abstract

The invention provides a neural network weight training method and application thereof. The training method for the weight of the neural network comprises the following steps: firstly, taking a predetermined number of short-term power load data as an initial training set, and taking an initial weight as a memory weight; obtaining a momentum factor, a resistance factor and an estimation factor of the weight training model by adopting a data linear analysis method; acquiring a weight updating speed according to the momentum factor, the resistance factor and the estimation factor, acquiring a weight variation according to the weight updating speed and the memory weight, taking an intermediate weight as a new memory weight, repeating the process for each piece of training data in an initial training set, and taking a final intermediate weight as an optimal weight.

Description

technical field [0001] The invention belongs to the field of neural network prediction, and in particular relates to a training method for the weight of a neural network and an application thereof. Background technique [0002] The main task of the power system is to provide users with stable, economical and power quality electrical energy in order to meet the user's power load demand, which requires that the power generation side of the power system should maintain a dynamic balance with the power load demand of the user side at any time. [0003] Since electric energy is difficult to store in large quantities, electric energy is easily subject to external interference during transmission, and the user's demand for electricity load changes momentarily, it is particularly important to accurately predict the user's short-term electricity load, which can enable the power system to generate electricity The power load demand on the side and the user side maintains a dynamic bala...

AI Technical Summary

Problems solved by technology

[0005] However, in the neural network used to predict the user's short-term power load in the past, when obtaining the optimal weight change, the hyperparameters of the weight change, such as the momentum factor, resistance factor, and estimation factor, are determined by the neural network on the user's The preset value of the short-term power load data before training does not change during the training process, and the time series of the user's short-term power load data is nonlinear, time-varying, easy to interfere, and has limited observation time. Therefore, using the previous neural network to train the user's short-term power load data will affect the convergence speed of the weight training model, so that the obtained optimal weight variation often falls into a local minimum, thus affecting the prediction of the user's short-term power load. precision

Images