An Improved Whale Algorithm Based on Recurrent Neural Network Short-term Power Load Forecasting Method

Abstract

The invention discloses a cycle neural network short-term power load forecasting method based on an improved whale algorithm, and relates to the technical field of short-term power load forecasting. Using the cyclic neural network for short-term power load forecasting, using the similar daily load data of the day to be predicted as the input data of the cyclic neural network, the number of input neurons, the number of output neurons, the number of hidden layers, and the number of hidden layers of the cyclic neural network are determined. Learning rate and gradient descent algorithm. A predictive model of recurrent neural network is constructed. Using the differential evolution algorithm to improve the whale optimization algorithm, the high-dimensional global optimization ability of the ordinary whale algorithm is improved. Use the improved whale algorithm to pre-train the weights in the cyclic neural network. When the pre-training is over, put the trained weights into the cyclic neural network model, and then use the gradient descent algorithm to train the cyclic neural network. When the training is completed , get a neural network model with fixed weights, and then perform load forecasting.

Description

technical field [0001] The invention relates to the technical field of short-term power load forecasting. Background technique [0002] With the wide application of electric energy, the economic development of all countries in the world is more and more dependent on electric power, and the demand for electric energy and the requirements for the quality of electricity are also getting higher and higher. Due to the particularity of electrical energy, it cannot be stored in large quantities, and it needs to be used immediately. To establish a balance between the production of electrical energy, the transmission of electrical energy and the use of electrical energy, it is necessary to accurately estimate the load consumption in the power system to ensure the reliability of the power system. performance, optimize the dispatching of the power system, and improve economic benefits. [0003] Chen Gang, Zhou Jie, Zhang Xuejun, et al. Daily load forecasting based on BP and RBF cascad...

AI Technical Summary

Problems solved by technology

[0006] The main problems of the method proposed in the above literature are two points. One is to use a neural network without memory capacity for load forecasting, and the results of load forecasting may produce discontinuity; the other is to use a single gradient descent algorithm to predict Weight training, the neural network is easy to fall into a local optimal state and it is difficult to jump out, which affects the prediction accuracy

Images