Short-Term Electric Load Forecast Based CSO-BP Neural Network

Short term Electric Load Forecast is the fundamental process to electric system optimization, it s outcome is paramount for the system electrical economy development as well as power system security. Currently, there are dozens of methods to do Load forecasting.However, engineers are expecting a more accurate result. Therefore, it’s significant to develop a faster and preciser way to do Load forecast in the future.This paper introduces the research contents of short-term load forecasting of power system in detail, summarizes load forecasting methods at home and abroad, puts forward the advantages and disadvantages of each forecasting method, analyzes the shortcomings of traditional neural network prediction algorithm, and proposes a new algorithm. The main contents of this paper include the following aspects:Analyzing the load characteristics of a certain country in the United States, summarizing that the load provides characteristics of weekly periodicity, daily periodicity and holiday characteristics, and analyzing the relationship between all factors affecting the load and the load.The paper indicates the causes of bad data in historical load data and the influence on load forecast; adopts improved FCM algorithm to cluster the daily load curve and generate various characteristic curves; applies the lateral similarity of load curve to identify the bad data; and finally utilizes characteristic curve for bad data adjustment, so that bad data is corrected and load curve glitches are eliminated.The method establishes BP neural network load forecasting model considering the relevant factors of daily characteristics; normalizes temperature, load, and week type and takes as the input variables of the BP neural network prediction model. The paper introduces cat swarm optimization. And test function shows that the cat swarm optimization is faster than the genetic algorithm in convergence and is not easy to fall into the local optimal solution.In order to overcome the shortcomings of slow convergence of neural network,and falling into local minima due to improper selection of network initial value, and improve the short-term load forecasting accuracy of power system, the paper combines the cat swarm optimization with BP neural network, and proposes the short-term load forecasting model based on cat swarm-BP neural network.The weight and threshold of BP neural network are optimized by cat swarm optimization, so that the blindness of the initial weight selection of neural network is avoided, and the BP neural network prediction model is trained to obtain the optimal solution. Results show that the BP neural network optimized by the cat swarm effectively improves the shortcomings of the BP neural network and improves the efficiency and accuracy of the neural network for the short-term load forecasting accuracy of the power system. Compared with the genetic neural network, the method has higher accuracy, faster convergence and a certain practicality.