Research On Simulation And Management Of Groundwater Under Changing Conditions In Western Jilin

As a long time, the surface water(especially transit surface water) utilization rate of Western Jilin region(the study area) is low, groundwater has been the main source of water of the region’s industrial and agricultural production. For the past of many years, with the increasing exploitation of groundwater, groundwater levels in some areas continued to show a downward trend, excessive and irrational exploitation of groundwater and deterioration of ecological environment have become a widespread concern issues. In order to al eviate pressure on the exploitation of groundwater, protect the ecological environment in Western Jilin, in recent years, a series of large-scale water conservancy projects, including the "Nen river water to Baicheng engineering", "Hatta Mountain Hydro Project," "Da an Irrigation Project", "Large-scale irrigation renovation project" and "Water-saving actions increase grain", which are intended to pull more river water in the area, mainly for agricultural irrigation, construction and operation of these projects will bring profound impact to groundwater.Therefore, it is very important to research the influence of the new large-scale water conservancy projects to groundwater in Western Jilin, the establishment of groundwater flow numerical simulation model under varying conditions and forecast of groundwater level for the next period of time are necessary. Groundwater flow simulation model is a mathematical model who can simulation the groundwater actual system, it follow the physical, chemical and biological principles of the groundwater system itself, can portray input- output response relationship of the system, equivalent of the copies of actual groundwater system, for changes of natural conditions and human activities, you can use the simulation model to predict the water level change. But the simulation model can only solve the predict problem under the given input decisions, which cannot solve the optimization problems.The optimization model(also known as operations research model) is able to solve optimization problems, that is, under the given objectives and constraints, through the different input- control plan, achieve the optimal target. So, in order to avoid groundwater depth appears too low and occurrence of a series of ecological and environmental problems in some areas due to the excessive exploitation of groundwater, groundwater management model can be used, through the model solution, we can obtain the optimal groundwater exploitation plan who can meet certain objectives and constraints.It should be noted, optimization model must be based on simulation model, in order to ensure the optimization process must be complying with the principles and rules of the actual groundwater system(expressed by the simulation model). Therefore, it is necessary to embed the simulation model to the optimization model through some methods, making it an integral part of the optimization model, actualize connect of simulation model and optimization model. In the past, embedded method, response matrix method and the state transition equation method are used to solve this problem, but it should call the simulation model hundreds of times to solve the optimization model, and these methods always appear computational load disadvantage. In this research, surrogate model method is used to connect the simulation model and optimization model, kriging surrogate model was established to replace the simulation model, which can meet the accuracy requirements, then greatly reduce the computation load of the optimization process.Firstly, establish a hydrogeological conceptual model according to the natural geography, topography, geological and hydrogeological conditions in the study area, on this basis, a numerical simulation model of groundwater flow was established considering human activities, especially artisanal mining in recent years and large-scale water conservancy projects affect, and then use 72 water observation wel s measured data for model calibration and validation of the model; apply stepwise regression analysis and wavelet neural networks method to establish a precipitation prediction model for forecast period(2014~2023) precipitation predicting, and then set up two forecasting plan of whether to consider the impact of each new large-scale water conservancy projects to predict the groundwater level; select recent typical year(2010) as the equalization period, combined with simulation results, based on the amount of recharge and discharge of groundwater in the study area during the equalization period, carried out groundwater balance calculations, the balanced area were the whole Western Jilin and each groundwater parameters partition respectively, and then the workable reserves of groundwater was calculated in accordance with the mining coefficient method; the Changling region was selecting as the key study area, based on the simulation model, orthogonal array- based Latin hypercube sampling method was applying for sampling in controllable input variables(pumping) of simulation model in its feasible region, then call the simulation model to obtain input- output sample data set; use Kriging method to establish an surrogate model, using the input- output sample data set for training and testing; then the surrogate model would be coupling to the optimization model(containing the desired function and other constraints), which can establish a groundwater optimization management model; use genetic algorithm to solve the optimization model, the optimized groundwater exploitation plan and groundwater depth distribution will be found. Through the above study, the following results were obtained:(1) The groundwater flow numerical simulation model based on hydrogeological conceptual model under varying conditions in the study area, after model calibration and validation, the results show a high degree of fit between simulated water level and observation water level, which description that the simulation model established can reflect the impact of changing conditions to groundwater in the study area, so it can be used as representative of the actual system of groundwater for future groundwater level changes forecast.(2) Applying stepwise regression analysis and wavelet neural network method to establish a precipitation forecast model. The advantage of this model is that not only retains the strong pattern recognition and fault tolerance of wavelet neural network, but also avoid the model instability caused by the volatility of precipitation values. After training and testing of the model, the results show that use this method to forecast precipitation, predict a higher precision than single wavelet neural network. And then 6 weather stations in the study area of 2014 ~ 2023(forecast period) annual precipitation were predicted, which can prepare conditions for groundwater level forecast for the next step.(3) Combining planning and implementation of all new large-scale water conservancy projects, set up two groundwater mining plan forecast scenarios: plan 1, under the current groundwater mining plan(2013) to set the amount of groundwater exploitation, without considering the impact of major projects; plan 2, based on the current groundwater extraction, consider the impact of each project on groundwater. Forecast results show that to the end time, under plan 2, the groundwater level decline significantly reduced relative to plan 1, indicating that these large-scale water conservancy projects have a great impact on groundwater of Western Jilin, groundwater negative balance situation improved under changing conditions.(4) The recent representative year(2010) was selecting as the equalization period, respectively, the balanced area were the whole Western Jilin and each groundwater parameters partition respectively, according to the water balance method, combined with the results of the simulation model, groundwater balance calculations were carried out. In order to verify the results of the equalization calculation, compared it with groundwater storage variation amount, the results show that the difference was smal, the results of balance calculations was reliable. Then according to equilibrium calculations, using the mining coefficient method to estimate the workable reserves of groundwater amount. The calculation results can not only provide a scientific basis for the local control of groundwater exploitation, but also to provide an exploitation constraints for optimizing model.(5) Changling region was selected as the key study area, orthogonal array- based Latin hypercube sampling method was applying for sampling in controllable input variables(pumping) of simulation model in its feasible region, results showed that sampling obtained has better uniformity using this method with respect to the direct use of Latin Hypercube Sampling method; call simulation model optained input- output sample data sets, establish an surrogate model using Kriging method, using the input-output sample data set for model training and testing, the results show that under the same pumping input, the output water level in simulation model and surrogate model were nearly identical, indicating that in the process of establishment and solution of groundwater optimization management model, Kriging surrogate model can be used to replace the original simulation model, which can meet the accuracy requirements on the basis of greatly reducing calculation load.(6) Confirm the objective function and constraints of the optimization model, coupling the Kriging surrogate model and optimization model to establish the groundwater optimization management model; genetic algorithm was using to solve the model, on the basis to meet the amount of groundwater exploitation demand, exploitable constraints, each single well pumping limits and maximum limits for groundwater depth, the total amount of groundwater exploitation was reasonable al ocation in the pumping wel s, then obtain the optimal groundwater exploitation plan; the water table depth distribution will be more reasonable based on the optimal groundwater exploitation plan, which can effectively prevent a series of ecological and environmental problems due to the excessive exploitation of groundwater in some areas.Those results, in theory, the applying of stepwise regression analysis and wavelet neural network method effectively improve the accuracy of precipitation forecast; at the same time, the using of Kriging surrogate model for the establishment of groundwater optimization management model, explore a more reasonable and effective technology; In addition, in order to obtain the training and testing samples(input-output data set) of surrogate model, the application of the orthogonal array-based Latin hypercube sampling techniques provide a more reliable tool for this process. From the practical sense, according to groundwater flow simulation model forecasting results, the impact of the large-scale project to groundwater balance was determined, through the establishing and solving of groundwater optimization management model, more reasonable groundwater pumping plan can be obtained, which can play an active role in ecological and environmental protection.