Research On Water Conveyance Loss Prediction Method In Nansi Lake-Dongping Lake Section

China’s water diversion projects have made remarkable achievements,which have played a role in alleviating the contradiction between water supply and demand in waterscarce areas and supporting social and economic development.River channel water conveyance is the main water conveyance method of water diversion projects,and the loss of water conveyance caused by leakage and evaporation when using river canal water conveyance is difficult to avoid,resulting in water resources waste and other problems;At the same time,water conveyance loss is a key factor affecting the implementation effect of water dispatch plan and realizing refined water scheduling.Accurate prediction of water conveyance loss is conducive to formulating refined water allocation schemes,supporting precise scheduling decisions,and giving full play to the role of water transfer projects,so water conveyance loss have always been a hot issue in water transfer projects.In this paper,the canal of the Nansi Lake-Dongping Lake section(Lianghu section)was taken as the research area,and the hydrological monitoring data of the Lianghu section from 2013～2021 were used to study the prediction method of water conveyance loss.According to the different emphasis aspects,three methods are proposed,the water conveyance loss of the two lakes section is simulated and predicted,and the prediction results of different methods are analyzed based on the actual water conveyance loss data,and the prediction accuracy and characteristics of each method are compared.The main research contents and conclusions are as follows:(1)Water conveyance process simulation and water conveyance loss prediction based on MIKE11 modelAiming at the changes of water level and flow during water conveyance in the two lakes section,considering the influence of hydraulic buildings on water flow movement,MIKE 11 was used to simulate the evolution of water flow,clarify the water conveyance process of river channels,and predict water conveyance loss based on the model.Based on the river network,section,water conservancy project and boundary conditions of the two lakes section,MIKE11 was used to establish the simulation model of the water conveyance process of the two lakes section.Model parameter determination and model verification are carried out through measured data.The results show that the average error of water level in key sections is within 0.2m,and the average error of flow is within 10%,which realizes the uninterrupted simulation of the temporal and spatial changes of water level and flow during water transportation.On the basis of this model,the water conveyance loss prediction is carried out,and the average absolute error of the prediction results is 11,780 m3,the root mean square error is 12,360 m3,and the average absolute percentage error is 5.018%.The MIKE model has high simulation accuracy for water level and flow,and has good intuitiveness,which can understand the water loss process along the way.The accuracy of water conveyance loss prediction is within a reasonable range,which can be used as a conventional method for predicting water conveyance loss.(2)Predict water conveyance loss based on improved empirical formulas modelAiming at the fact that the leakage loss in the actual water conveyance process has the characteristics of dynamic change along the course,the integral idea is introduced to improve the common leakage empirical formula and characterize the dynamic change characteristics;Combined with the regional characteristics of the two lakes,the formula of relative humidity factor correction evaporation experience was introduced.Different geological segments are characterized in the form of segmented functions,and the genetic algorithm is introduced to determine the parameters according to the irrationality of the values taken by the empirical table.The combined and improved leakage and evaporation formulas constitute an empirical formula water conveyance loss prediction model,and the first combination is to improve the Davis-Wilson formula and improve the Dalton formula,and the second combination is to improve the Kostiakov formula and improve the Dalton formula.The results show that the empirical formula model has a good prediction effect on the water conveyance loss in the two lakes,with the average absolute errors of the first combination and the second combination being 11,130 m3 and 12,520 m3,and the root mean square error being 13,170 m3 and 15,030 m3,respectively,and the average absolute percentage errors being 4.663%and 5.162%,respectively.The prediction accuracy of the improved empirical formula on water conveyance loss is greatly improved compared with before the improvement,indicating that it is reasonable to consider the dynamic variation characteristics of leakage loss along the process and introduce relative humidity factors to modify the evaporation formula.(3)Predict water conveyance loss based on IPSO-ELM modelAiming at the characteristics of complex and diverse influencing factors of water conveyance loss,the influencing factors are optimized.Considering that the water conveyance loss mechanism is unclear and the expression form of influencing factors is difficult to determine,the neural network model is used to reflect the relationship between influencing factors and water conveyance loss,and the water conveyance loss prediction is made.Improved Particle Swarm Optimization(IPSO)is used to optimize the performance of the Extreme Learning Machine(ELM),establish the IPSO-ELM water conveyance loss prediction model,propose a water conveyance loss prediction method based on the IPSOELM model,and compare and analyze it with other neural network models.A total of three neural network models and four different influencing factor input combinations were set up for comparative analysis to verify the superiority of IPSO-ELM model and obtain the optimal influencing factor combination.The results show that the IPSO-ELM model has the highest prediction accuracy.Blindly increasing the number of influencing factors can not improve the prediction accuracy of the model,and the optimal combination of influencing factors can not only reduce the amount of model calculation,but also improve the accuracy of model prediction,and the optimal combination of influencing factors is water level,flow,wind speed,relative humidity and air temperature.The results show that the average absolute error of 6,830 m3,the root mean square error of 8,350 m3,and an average absolute percentage error of 3.157%.It shows that the IPSO-ELM model established with the optimal combination of influencing factors can predict water conveyance loss well,and the water conveyance loss prediction method based on neural network is feasible.