Study On Short - Term Gas Load Forecasting Model Based On Improved Hybrid Frog Leaping Algorithm

Natural gas is the world’s third largest energy, but also a green, efficient,economic, safe and environmentally friendly energy. With the increasingly serious pollution of the urban environment, the air quality is getting worse and worse, and improve the proportion of green natural gas in energy consumption is also increasingly important. Improving the proportion of natural gas in cities is not only conducive to the promotion of energy-saving emission reduction, but also to promote the sustainable development of economy and society, and reflects the people’s living standard is an important indicator. Therefore, the scientific and accurate prediction of natural gas is not only related to the interests of natural gas companies, but also related to the construction and development of the natural gas industry, and the vast number of people’s lives are also closely related to the development of the natural gas industry.The short-term load forecasting method is mainly dependent on the historical load data, its essence is based on the similarity principle of load forecasting.Traditional gas load forecasting is the forecasting optimization algorithm to improve the prediction accuracy, this paper based on fuzzy C-means(FCM) clustering analyses proposed a new method to select training set of least squares support vector machine(LS-SVM). Least squares support vector machine is a novel machine learning method to use high dimensional map, the switching nonlinear problem to a linear problem. In pattern recognition and regression estimation has good performance, however, when the mass input data is involved in the training model, the convergence time is longer.So, we can use the fuzzy C means clustering to classify the historical gas load on the basis of analyzing the characteristics of the gas load, and calculate the distance between the sample and the cluster center, then select the data in the class of the nearest distance as the similar days to be predicted, and as the training sample of LS-SVM. Traditional FCM clustering is easy to fall into local optimum, In this paper,we use the improved shuffled frog leaping algorithm of global optimization and depth of local search characteristic of the shortcomings of the traditional FCM clustering optimization. and To get the optimal classification results, so that the classification result conforms to the internal structure of the real data of gas load. In this way, the input output data regularity of the load sample which is involved in the LS-SVM training model is strengthened, and the training samples are guaranteed to meet thehigher degree of the same input output function, realize the effective combination of optimization of the FCM clustering analysis and LS-SVM algorithm. This method improve the prediction accuracy of LS-SVM model, and simulation results prove the effectiveness of the hybrid model mentioned, which reduction and optimization of training samples.Finally, this paper aimed at the Shanghai area of a year of gas load data on the Rstudio platform,We use the optimized FCM clustering to select the training samples to train the LS-SVM prediction model, And the prediction results are compared with the results predicted by the BP neural network algorithm and the traditional LS-SVM algorithm. The results show that the prediction algorithm based on Optimization of FCM clustering and LS-SVM algorithm can effectively improve the accuracy of gas load forecasting.