The Research On Gas Load Forecasting Based On Empirical Mode Decomposition And Optimal Neural Network

Natural gas has been widely applied because of the advantages of safe, reliable, clean and environmental. Gas scale expands unceasingly in our country, The requirements on construction of gas pipeline network, maintenance, and gas storage and other aspects of peaking of people are also rising. Load forecasting is an important reference for the above work, so load forecasting methods have been widely studied, improve forecast accuracy is also important.This paper first introduces the research background and current situation of gas load forecast and the basic knowledge about forecast. Because the system or human causes, will inevitably produce bad data, the real load change rule cannot be found. But the accuracy of historical load data is an important part of the prediction precision, so the bad data discovery and pretreatment are analyzed in detail, including fill the missing data, correction abnormal data and quantitative data of text type. Then the brief analysis of the key technology used, including the BP neural network and linear neural network and particle swarm optimization algorithm and the basic principle of empirical mode decomposition, for the gas load forecasting models of experiment and simulation laid a theoretical foundation.Then, it is the important part of the article, the establishment of the load forecasting model and the simulation process. Due to the global search ability of particle swarm algorithm. First of all,establish PSO- BP prediction model. In view of the faults of basic algorithm, this paper make some improvements:Using adaptive adjusting inertia weight w according to the fitness. When the fitness of the particle is smaller than the average adaptive value, corresponds to a small weight, enhance the ability of local of particles. When the particle’s fitness is greater than the average fitness, should increase w, make the particles search to a better direction, increase the global search ability; The particle is mapped to chaotic variables by logistic mapping. Use the ergodic of chaos to search, avoid PSO algorithm trapped in local minima; Local search. After PSO is initialized, get the current optimal value, select 1/5 before the optimal particle to chaotic search. Finally puts forward a new improved particle swarm algorithm ACLSPSO, and establish the ACLSPSO optimization neural network forecasting model to forecast simulation, proved the effectiveness of the improved algorithm.Finally, the gas load characteristics are analyzed, and puts forward the introduction of the EMD in load forecasting model. And analyzes the feasibility of the EMD used for gas load forecast. Establish empirical mode decomposition and ACLSPSO optimization neural network prediction model for the simulation. Gas Load mode is broken down into several components, establish an appropriate gas load forecasting model for each component of prediction and simulation. Then fit the total prediction. And gives the forecast results and the comparison test, analysis the performance of the model.