Short Term Gas Load Forecasting Based On Modified Artificial Bee Colony Algorithm

High-quality,clean and environmentally friendly natural gas will greatly improve the environment,and its share in energy consumption will also increase.The demand for and consumption of natural gas continues to increase,and the reliability of supply related to it has also received increasing attention.The accurate prediction of gas load is of great significance to the reliability of supply.The main work of this paper is improving the model and strategy of prediction to obtain the best gas load forecasting accuracy.Firstly,based on the theoretical introduction of commonly used prediction models and algorithms,experimental simulations such as support vector machine,BP neural network and extreme learning machine are performed on gas load data.The experimental results show that the prediction performance of these models is greatly influenced by the parameters.Therefore,selecting the appropriate parameters is of great significance for improving the prediction performance.Then,in order to get the appropriate parameters,this paper studies the use of swarm intelligence algorithm for parameter optimization.The artificial bee colony algorithm has been proved that is superior to the genetic algorithm and the particle swarm algorithm,and has more superior optimization performance.Therefore,the artificial bee colony algorithm is used to select parameters for the prediction model.Because this algorithm has some intrinsic deficiencies,such as gradual stagnation,difficulty in jumping out of local optima and failing to achieve the theoretical optimum,etc.,in order to further improve the performance of the algorithm,this paper introduces improved search strategy and new search process,and the corresponding hybrid strategy.Mainly summarized as the following points: First,an adaptive search strategy based on stochastic new solution guidance is proposed.This strategy can effectively alleviate the problem that the decline of population diversity and falling into local optimal solution easily.Second,The repeated Gaussian search mechanism is introduced to maintain the balance between the quality and the population diversity.Third,the searching process based on the ideas of cat swarm optimization algorithm is introduced.This process will perform a full and fine local search on the better solution,and guide the poorer solution to the global optimum to reduce the number of invalid searches.The above points are the hybrid artificial bee colony algorithm in this paper,and then the algorithm is used in the parameter optimization of the prediction model.In addition,in order to get the best forecasting accuracy,this paper uses the fuzzy C-means clustering algorithm and the appropriate forecasting model strategy.Firstly,the training samples are clustered by the fuzzy C-means clustering algorithm.Then,under the premise of choosing the parameters for the model with the optimization algorithm,the prediction models are constructed for the every class based on the training data that belong to corresponding class.Finally,when the test data arrives,the test data is divided into different classes,and the trained predictive model of the corresponding class is selected to predict the load value for the test sample.Finally,this article uses the above method to establish a complete forecasting scheme for gas load forecasting,namely forecasting based on the fuzzy C-means clustering algorithm and the hybrid artificial bee colony algorithm.At the same time,the scheme is implemented by using least squares support vector machine and extreme learning machine as prediction models,and compared with the prediction method without application of this scheme.The result proves that the proposed prediction scheme has better prediction performance.