Studies On Watershed Long-Term Hydrological Forecast And Evaluation Method Of Water Resources Carrying Capacity

Water resources plays an important role in human survival and social development.Ensuring the sustainable development of water resources is the primary prerequisite to achieve sustainable economic and social development.Watershed hydrological forecasting and water resources carrying capacity analysis are two important issues in water resources system research.It is the foundation for water conservancy project planning and construction,optimal allocation of water resources and sustainable development of society.Since the twenty-first century,the sustainable use of water resources in China is facing multiple pressures such as population growth,water supply and demand imbalance and climate change.On the one hand,the global warming caused the melting of the polar ice cap,the sea level rise,the decrease of freshwater resources.Heavy rain,flood,drought and other extreme events increased significantly.On the other hand,the large-scale water conservancy projects produced a significant influence on the hydrological system,causing hydrological system deviates from the evolution law of the natural conditions.The significant changes in the atmosphere,surface,soil,river and groundwater processes of water circulation system lead to series problems of resources,ecology and environment.The variation of climate and water cycle processes puts forward higher requirements for hydrological forecasting and water resources carrying capacity analysis.In this paper,the key scientific problems of medium and long-term hydrological forecasting and water resources carrying capacity in the changing environment are analyzed.The Jinsha River and the southwest region are the main research objects.The temporal and spatial changes of runoff and rainfall characteristics in the Jinsha River Basin are analyzed.The fuzzy support vector machine and the artificial neural network monthly runoff-forecasting model are studied,and two important issues of forecasting factor selection and time series decomposition in model construction are compared.The meteorological factors,atmospheric circulation,the Kuroshio sea surface temperature,the summer monsoon index,which influence the rainfall in the middle and upper reaches of the Yangtze River,are selected as forecast factor by the sequence floating forward algorithm.Gaussian process regress is used for the Jinsha River Runoff forecasting.Finally,the evaluation model of water resources carrying capacity of southwest region with variable fuzzy set-set pair analysis is established.The main research contents and innovative achievements of this paper include:(1)The variation of runoff within a year in the Jinsha River is uneven.The interannual variation of the annual runoff is small,and the interannual variation in the lower reaches of the Jinsha River is smaller than that in the upper reaches.The precipitation in the Jinsha River Basin is concentrated in the period from May to October,and the rainfall is concentrated in the middle and lower reaches of the Jinsha River basin.There is a high correlation between precipitation and runoff in the Jinsha River.Co-occurrence probability of Jinsha River flood and Wujiang flood is very small,and when the Wujiang River has been a big flood,Jinsha River flood probability is great.(2)The formation of runoff is influenced by many factors such as hydrology,terrain and meteorology.Due to the change of climate and the watershed conditions,the contribution of the sample points to the forecast results at different periods is not the same.The fuzzy support vector machine introduces the concept of membership degree of fuzzy set theory into support vector machine.For medium and long-term runoff forecasting,the sample data have different error requirements according to the impocrtance.The results of the forecast in the Jinsha River Basin show that the FSVM model is better than the traditional SVM and GRNN model.The gray relational analysis is applied to the forecasting factor selection.The results show that the gray relational analysis method is superior to the traditional correlation coefficient method.FSVM based on gray relational analysis has greatly improved the prediction accuracy and generalization ability,especially has a better forecast results for the flood season runoff.(3)Sequential decomposition of the input factor has been shown to improve the prediction accuracy of the hydrological model.However,the existing time series decomposition studies are based on wavelet analysis,and the applicability of other decomposition methods in the field of hydrological forecasting is rarely studied.Using the discrete wavelet transform,empirical mode decomposition and STL decompose the prediction factor into sub-sequences,and then the decomposed sub-sequence is used as the model input.The results show that all of three kinds of decomposition techniques can improve the prediction accuracy of the model,and the result by discrete wavelet transform is improved greatly,the result by the empirical mode decomposition is improved minimal.(4)Runoff and the climate factor have a correlation of lagging behind.The sequence of floating forward selection algorithm is used to selected a set of forecast factor group from the meteorological factors rainfall,air temperature and air pressure,74 circulation factors released by the National Meteorological Center,the Kuroshio sea surface temperature closely related to the runoff in the Yangtze River Basin,and Pacific Ocean Oscillation,East Asian Summer Monsoon Index.The Gaussian process regression is introduced into the study of runoff forecasting and the sensitivity of predictor is analyzed.The climatic factor-based Gaussian process hydrological forecasting model has a good forecasting effect and belongs to the uncertainty forecast.The sensitivity analysis of the Gaussian process regressions hows that the factors influencing the runoff forecast in the flood season are the early runoff,the rainfall and the humidity,and the continental circulation in the eastern Atlantic and the North Atlantic oscillation teleconnection factor.(5)In this paper,we combined the objective weight method of information entropy theory and the subjective weight method of analytic hierarchy process;adopt the improved fuzzy variable model to evaluate the water resources carrying capacity.The maximum membership degree principle is a kind of analysis method,which has extensive application in decision-making and evaluation.However,the principle itself has many defects in information loss,which can lead to the deviation of evaluation.In order to avoid this problem,the research work introduces set pair analysis method,which can deal with the uncertainty problem.Through the identical different opposite analysis,use the degree of connection to describe the proximity between collections.The results show that the method is rigorous and highly reliable.