Agricultural Water Demand Prediction And Optimal Allocation Of Water Resources In The Shiyang River Basin

Shiyang River Basin, which is located in the northwest arid region, is one of the main basins where there exists water shortage. Water demand bases mainly on agricultural irrigation in the basin, which of each industry constantly varies as the climate changes and the population increases in recent years. Water shortage certainly leads to the bad phenomenon that the departments of water consumption fight for water. Therefore, it is necessary to conduct the study of agricultural water requirements and optimal allocation of water resources in order to realize the aim of sustainable development. In the paper, agricultural water use in the basin was studied by collected data, concluding the following results:(1)Meteorological data and crop data collected were used to establish the model of agricultural water demand prediction on the basis of 11 factors influencing agricultural water demand in Minqin, Tianzhu and the whole basin. The accuracy of the model was superior by testing.(2)There needed a large number of data to establish BP neural network model, which brought about tedious operation. The correlation analysis of 11 factors was conducted that influenced agricultural water requirements so as to forecast agricultural water demand in the cases of fewer data. Then these factors and the correlation of agricultural water demand were determined as well as the most important factors affecting the water requirements of the basin, that is, cultivated area and precipitation. 6 factors that obviously influencing water demand were also ascertained, namely cultivated area, precipitation, food crop area, accumulated temperature, sunshine and the average annual maximum temperature.(3)The hexatomic linear regression models of water requirements were established on the basis of 6 main factors multiple regression analysis by multiple regression analysis. The influencing factors were further optimized, thereby setting up binary linear regression model and BP neural network based on the two best important factors. And five-year data from 1999 to 2003 were used to check up the accuracy, which proved that the prediction effect of BP neural network model was better than that of binary linear regression model.(4)The grey model, exponential smoothing model and their combined model were used to predict water demand. The forecasting model was built in the Shiyang River Basin by analyzing agricultural water requirements over the years. The precision of the three models was tested. Then it was dictated that the absolute of the average relative error of grey model was 4.84percent, binary exponential smoothing model 6.14percent and the combined model, the minimum of the three, 4.04percent. The combined model ascertained was utilized to predict the agricultural water demand in the basin in the coming 10 years. The prediction value in 2004 was 17.677×10~8m~3, and it reached 19.178×10~8m~3 in 2013.(5)Economic benefit, social benefit and ecological benefit were taken into account and the maximum comprehensive benefit was its final objective. The theory of the multi-objective fuzzy optimal model of the crop planting structure was used to establish the multi-objective fuzzy optimal model of the crop planting structure. The object function was solved by the object function established in the two conditions—area and water yield. Thus, the main crop planting area was determined in the condition of the maximum comprehensive benefit.