Research On Water Temperature Regulation Technique Of Large Reservoirs Based On Artificial Intelligence Algorithm

Relying on strong infrastructure capacity and the advantageous terrain of southwestern China,a batch of large-scale hydropower project have been built to provide sustainable drive for social and economic development.While utilizing water energy to achieve generating electricity benefit and promote energy conservation and emission reduction,these hydropower project have also caused a series of eco-environmental problems.The disadvantageous ecological effects of outflow with low temperature on the fish have been widely concerned.The stratified water intake facilities are important engineering facilities to mitigate the negative impact of low temperature water on ecological environment,and the operation effect of stratified water intake facilities is a focues issue.The operation effect of stratified water intake facilities is closely related to the inflow conditions,meteorological condition,operation mode.Under actual conditions,the inflow is always in a state of flux.According to the inflow conditions,timely and rapid prediction of water temperature and implementation of stratified water intake is a technical issue concerned by the engineering management organization and also a scientific issue concerned by the academic community.Mathematical model based on physical meaning is widely used in environmental impact assessment,reservoir water temperature structure analysis and water temperature prediction under typical operating situations.However,due to the high specialty,the complexity of parameter calibration and the time consuming of calculation,it is difficult for this model to meet the demand of quick decision of reservoir operation under complex inflow conditions.In addition,most of the stratified water intake facilities of large hydropower stations are in the stage of trial operation or debugging in China,and the amount of observed data is limited,and the evaluation method system of operation effect of stratified water intake facilities has not yet been established.In recent years,the rapid development of information and data processing technology and its application in the field of engineering provide methods for solving the problems of multi-factor influence.In order to effectively solve the above problems,this study takes the opportunity of the rapid development of information and big data technology in recent years,combining the advantages of of mathematical model based on physical meaning and Artificial Intelligence(AI)algorithm,to explore the reservoir outflow temperature rapid prediction technology,and to study the evaluation and optimization of operation effect of stratified water intake facilities.Main research contents and achievements are as follows:(1)A program for AI fast prediction model of outflow temperature is developed based on Python3.5 language,the core module of the program includes Support Vector Regressioin(SVR),Back Propagation(BP)Neural network,Recurrent Neural Network(RNN),Long Short Term Memory(LSTM),and Gated Recurrent Unit(GRU)5 sub-modules to complete parameter setting,model training and predicting.Meanwhile,the program also has the function of data pre-processing and post-processing.(2)Based on synthesis consideration of flow,meteorology,inflow water temperature and operating scheme of stoplog gate,this paper designs 108 possible operating conditions for the actual operation of Jinping-I hydropower project,to meet the requirements of model training for data volume and operating condition.Using the EFDC model verified by Jinping-I observed data,the water temperature distribution and outflow temperature under various operating conditions were simulated.By organizing the boundary conditions of the design conditions and the simulation results of EFDC,a dataset containing nearly 20,000 sets was producted,which the data of flow,meteorology,inflow temperature,stoplog gate operation,reservoir water temperature distribution and outflow temperature in the dataset were one-to-one corresponding.(3)The outflow temperature is affected by many factors such as reservoir characteristics,operation mode,inflow condition,meteorological condition,etc.In this paper,a machine learning model suitable for outflow temperature prediction of Jinping-I is established by optimizing input factors and structural parameters.The result shows that the AI model can accurately predict outflow temperature by taking main and tributary inflow,main and tributary inflow temperature,outflow,stoplog gate operation mode,air temperature,solar irradiance,relative humidity and wind speed as the input factors and the outflow temperature of the reservoir as the output factor.(4)Comparing the performance of traditional machine learning algorithm and new deep learning algorithm,the result shows that the prediction accuracy and uncertainty of RNN,LSTM and GRU 3 new deep learning algorithms are significantly better than that of S VR and BP neural network.Meanwhile,the AI model can establish the response relationship between the historical data of key influence factors and the outflow temperature of the current period,and achieve the accurate prediction of the outflow temperature within 15 days under the condition of not using the stoplog gate,and 7 days under the condition of using the stoplog gate,so as to reserve time for the formulation of the water intake scheme and the operation of the stoplog gate.(5)Compared to traditional EFDC model,trained AI models are easy to operate.The outflow temperature will be predicted as soon as the input factor data of the current moment are input.The prediction speed of AI models can reach to the second response,which is far better than the traditionally mathematical model,and can meet the demand of the rapid decision of reservoir operation.(6)Based on the Analytic Hierarchy Process,the evaluation system of the operation effect of stratified water intake facilities are constructed.Based on the evaluation system and prediction model of AI outflow temperature prediction model,an optimization design process of operation scheme of the stratified water intake facilities was proposed,which was "scheme design-outflow temperature predictione-effect evaluation-scheme comparison".Taking high,normal and low flow years as examples,the optimal design of operation scheme of the stratified water intake facilities is carried out according to this process,and suggestions for the optimal operation of layered water intake facilities are put forward based on the water level conditions of different hydrological years.Overall,the ecological benefits of operation with multi-layer stoplog gate is generally higher,which can be used as preferred alternatives,but while the ecological benefits of various schemes are similar,the stoplog gate operation scheme with fewer layers can be chosen to reduce the loss of power generation.