Font Size: a A A

Research On Gas Load Forecasting Based On BP Combination Model

Posted on:2017-04-28Degree:MasterType:Thesis
Country:ChinaCandidate:T WangFull Text:PDF
GTID:2272330485466777Subject:Engineering
Abstract/Summary:PDF Full Text Request
With the continuous development of the gas market, gas load forecasting has become an important work of the gas system management department. Accurate prediction of gas load can make the planning scheme of gas pipe network better, and it can be more reasonable to dispatch gas. It has important significance to improve the economic benefit and social benefit of the Gas Company, to maintain the safe and stable operation of the gas system, and to ensure the orderly conduct of people’s daily life.Firstly, this paper introduces the research background, significance and research status at home and abroad, and describes the basic theory of gas load forecasting; Secondly, the characteristics of gas load and the influence factors of gas load are described, and the data pretreatment of gas load data is described, which provides theoretical basis and preparation work for load forecasting.Then this paper mainly describes the method for prediction of gas load, due to gas load forecasting is a very challenging work, many influencing factors, large data changes, the prediction is difficult to achieve satisfactory accuracy. To determine more suitable forecasting model for gas load, this paper explores the two aspects of multi model comparison and optimization. In this paper, the method is based on the traditional mature theory to establish the grey theory model, support vector machine model and BP neural network model, through the comparative analysis to get the most suitable model; on this basis, for the BP neural network convergence rate is slow, and easy to fall into the local minimum value of the shortcomings, further research is carried out on the optimization of the model. The specific use of the two main programs, namely, parameter optimization and data classification. For the optimization of the parameters of the forecasting model, this paper introduces genetic algorithm, particle swarm optimization algorithm and the cuckoo search algorithm which several belongs to the biology of parameter optimization algorithms to respectively of BP neural network parameter optimization. After three kinds of parameter optimization algorithms, the population individual with good fitness as the BP neural network is trained and predicted by the combination of weights and thresholds. Through the experimental analysis, cuckoo search algorithm of BP neural network parameter optimization effect is better, model prediction accuracy is higher, so the gas load forecasting model which is used in this paper is CS-BP neural network model.On the other hand, the paper puts forward the load data into two kinds: one kind is the daily load data, one kind is the load data of special day(mainly is holiday), the practice is to separate the special day to establish a forecast model, and the special daily load forecasting is introduced into the similar day method, and the candidate similar days are selected by the fuzzy filtering of the feature vectors, then through the calculation of the similarity degree and the gray correlation degree, when the similarity measure a?5.0 and the grey correlation degree rank in the top 10, the final similar days are selected. And due to the time span of the load forecasting, the time correction method is also introduced to the load data of similar days in this paper, and then determine the input vector of CS-BP neural network. Experimental results show that the classification on the one hand reduces the impact of special days on the overall prediction accuracy, on the other hand, the load forecasting accuracy is improved.
Keywords/Search Tags:gas load forecasting, BP neural network, cuckoo algorithm, similar days, ordinary day, holidays
PDF Full Text Request
Related items