LSSVM Gas Load Forecasting Based On Improved Artificial Bee Colony Algorithm

With the development of economy,the increasing of population, and the resulting generally promotition of people’s living quality, people’s demand for natural gas is becoming more and more enormous.Energy crisis is affecting people’s lives, and natural gas is abundant, renewable as a green energy,therefore it is widely used.gas that it is Low-cost,non-polluting,and natural which has the very good development prospect. In order to use and transport natural gas more effectively and to improve the utilization efficiency of the weather, we must be able to forecast the natural gas loading mor scientific and effective for a period of time in the future according to the relevant impact factors of region’s natural gas load,such as the weather, temperature,holiday.whether can we forecast the local gas load accurately or not will directly affect the safety of electricity for residents,and also the interests of gas producers.While in the field of power loader forecasting, and the predicting of solar energy has been quite mature, we can’t directly copy the method of other areas.The general method is difficult to guarantee the performance of the prediction of gas consume,because the number of the sample is few. Support vector machine(SVM)is a kind of artificial intelligence method, and according to the analysis of the related literature, it is suitable in the prediction of gas consume.As a modified version,LSSVM encompasses similar advantages as SVM, but LSSVM offers a linear system of equation rather than Quadratic Programming, which is achieved by applying equality constraints instead of inequality constraints. Hence,such an approach simplifies the training process of a standard SVM to a great extent and finally promote the development and application of SVM, so this article uses the LSSVM algorithm as the theoretical basis of gas load forecasting.Selection of the hyper-parameters is critical to the performance of Least Squares Support Vector Machines(LSSVM), directly impacting the generalization and regression efficacy of the LSSVM. Improved artificial colony algorithm is used to optimize the LSSVM parameters,and finally analysis and forecast the gas consume of the local region combined with LSSVM.In order to solve the problem above, In this paper,the Artificial Bee Colony Algorithm, ABC, is improved, introducing a new cross processing method, studying a ABC based on double population policy, and putting forward a run-time parameteradjustment method. this paper propose a gas load model based on the improved artificial colony algorithm of LSSVM, and at the same time it analysis a contrast experiment against the improved artificial colony algorithm to prove the effectiveness of the algorithm.Finally, this artical analyze and forecastthe gas comsume with the input of history sample,It prove that the method we proposed has a higher than prediction accuracy than the traditional artificial colony algorithm.and meanwhile prove that the proposed artificial colony algorithm has good practicability in the LSSVM.