Study On Water Demand Forecast Of Rural Water Supply Project

Water is the basic condition for human survival,and water plays an important role in our life.Because of the uneven supply of water resources in urban and rural areas,the unscientific construction planning of rural water supply projects,the neglect of rural drinking water quality and the irrationality of rural water supply management,water resources supply problems have been brought to many areas of the country.In this context,the integration of urban and rural water supply was born.Driven by the integration of urban and rural water supply,urban and rural water resources will gradually be optimized allocation.Scientific and reasonable prediction of rural water demand is the basis and premise of optimal allocation of urban and rural water resources.This paper mainly studies the problem of forecasting the water demand of rural water supply projects under the condition of lack of data in the process of urban-rural water supply integration,aiming at solving the problem of determining the scale of urban-rural water supply system in the process of new construction or expansion.In this paper,we propose a Multi-Filling method based on user correlation to fill a small part of missing data in the original data,which makes up for the low accuracy of single mean filling in time series with high discreteness.After data pretreatment,the rural areas are classified by fuzzy C-means clustering.On the basis of literature analysis,official data analysis and data analysis,the new elements synthesized by the main cause analysis method are added into the impact factor group.Pearson correlation analysis is used to reduce the dimension of the impact factors for each type of rural areas to get the key factors.Finally,all kinds of water use data and key influencing factors are imported into Elman neural network to predict and the prediction results of Elman neural network are corrected by rough set theory.This model not only has a good fit,but also is relatively stable,so it can provide a new way of thinking for the research of rural water supply project water demand in the absence of data.Taking 16 villages around Baocheng Town in Baoting County of Hainan Province as an example,the water consumption of villages from 2010 to 2016 is used as training data to forecast the water consumption of villages in 201 7.Prediction results show that rough Set-fuzzy C-means clustering Elman neural network is used to predict unfilled data,single-mean filling data and multiple filling data respectively.Compared with the former two,the latter’s MAPE is reduced by 0.099 and 0.058,respectively,which proves the validity of data filling.Then,rough Set-fuzzy C-means clustering is applied to multiple filling data respectively.Elman-like and Elman-like neural networks predict that the former MAPE is reduced by 0.163,which verifies the superiority of the algorithm.