Study On Characteristics Analysis And Implicit Stochastic Optimization Operation Method Of Hydropower Stations And Reservoirs

The scientific implementation of the refined operation of hydropower station reservoirs and deep tapping of the power generation potential of hydropower stations that have been put into operation are of great significance to realize the efficient development and utilization of hydropower energy,reduce the consumption of fossil energy in the power system,and promote the improvement of the energy structure of the power system.The refined operation of hydropower station reservoirs puts forward new and higher requirements for the accuracy of reservoir operation model,especially the accuracy of power output calculation of hydropower stations.How to accurately describe the volume characteristics of reservoirs,the dynamic characteristics of hydropower stations,and the tail water level characteristics of hydropower stations,so as to realize the accurate calculation of power generation output,is a key problem.In addition,affected by the uncertainty of reservoir inflow,the refined operation of hydropower station reservoirs is essentially a sequential decision-making problem with uncertainty.With the continuous improvement of hydrologic forecasting technology,it is possible to use runoff forecasting information for dispatching decision-making.Therefore,how to make full use of runoff prediction information for reservoir operation and overcome the impact of runoff prediction uncertainty on operation decision-making is another key scientific problem.Focusing on these two problems,the characteristics analysis and modeling of hydropower station reservoirs,the stochastic simulation of single-site and multi-site runoff,the medium-and long-term runoff forecast,and the implicit stochastic optimal operation of hydropower reservoirs have been studied in depth.Some valuable results have been achieved,and can provide model and method supports for the refined power generation operation of hydropower station reservoirs under the condition of uncertain inflow.The main research contents and innovative achievements of this paper are as follows:(1)In view of the lack of physical significance in the functional formal representation of the reservoir volume characteristics,the irregular cone model of reservoirs were constructed,and the power function curve representing the reservoir volume characteristics was derived according to the cone model.At the same time,the reasonable value ranges of parameters of the power function curve were given according to the mathematical properties of the reservoir volume characteristics.Finally,through the rationality analysis,goodness of fit evaluation and reservoir operation application verification,it was proved that the proposed power function curve was a kind of reservoir volumetric characteristic representation curve with wide applicability,high reliability and clear physical significance,which can provide basic curve support for the refined power generation operation of reservoirs.(2)Focusing on the analysis and modeling of the dynamic characteristics of hydropower stations,firstly,the dynamic characteristics of hydropower stations were analyzed in depth based on the historical operation data of hydropower stations.It was found that the dynamic characteristics of hydropower stations had multiple values for a single independent variable(water head or power generation flow).At the same time,it was also found that the dynamic characteristics of hydropower stations had a certain degree of time scale effect.On this basis,considering the dual effects of water head and power generation flow as well as the influences of time scale,the curved surface models of the efficiency characteristics,water consumption rate characteristics and power characteristics of hydropower stations were constructed by using polynomial and neural network,and the accurate description of the dynamic characteristics of hydropower stations was preliminarily realized.The case studies showed that the efficiency characteristics,water consumption rate characteristics and power characteristics surface models had higher power generation output calculation accuracy than the efficiency characteristics and water consumption rate characteristics curve models with head as input,and can provide model supports for the accurate calculation of power output of hydropower stations.(3)Focusing on the analysis and modeling of the tail water level characteristics of hydropower stations,firstly,based on the historical operation data,the variation process of the tail water level of hydropower stations was deeply analyzed,and it was found that the variation process of tail water level of hydropower stations has obvious aftereffect characteristics compared with the discharge flow process of reservoirs.On this basis,by using Pearson correlation analysis method,the key influencing factors of the tail water level in the current time period were identified,and the polynomial fitting models and support vector regression model for the tail water level characteristics of hydropower stations were constructed.The performance of various models was compared and analyzed based on the goodness of fit indexes.The results indicated that the support vector regression model with the reservoir discharge flow and the downstream hydropower station water level(or the downstream tributary inflow)in the current and previous periods as the inputs had the better practicability,reliability and accuracy,and can provide model support for the accurate calculation of tail water level,water head and power output of hydropower stations.Finally,the time scale effect of the tail water level characteristics of hydropower stations was studied.It was found that the time scale effect was not significant,indicating that it was not necessary to consider the influence of time scale when constructing the mathematical model of tail water level characteristics,unless there was higher requirement for the prediction accuracy of the tail water level.(4)In view of the shortcomings of traditional methods that need to assume the probability distribution of hydrological variables and the dependence structure of runoff series,the Gaussian mixture model(GMM)which can approximate any continuous distribution with arbitrary accuracy was introduced,and the new stochastic simulation methods of single-site runoff series based on GMM and seasonal GMM were proposed,and the new stochastic simulation methods of multi-site runoff series based on GMM and seasonal GMM were also proposed.The applicability and effectiveness of the proposed methods were tested by the stochastic simulation experiments of the single-site and multisite ten-day runoff and monthly runoff.The results showed that the proposed single-site and multi-site methods can effectively maintain the main statistical characteristics of the measured runoff series,and can be used to generate a large number of simulated runoff series that implicitly reflect the random characteristics of runoff,and then provide data support for the derivation of reservoir operation function.(5)Focusing on the power generation dispatching problem of hydropower station reservoirs under the uncertainty of inflow,a reservoir operation function derivation method based on the least-squares boosting decision tree was proposed,and was used to derive the optimal operation function considering multi-step runoff forecast information.Then,in order to provide runoff prediction and its uncertainty information to the reservoir operation function,a mid-and long-term runoff forecasting model driven by remote correlation factors and regional meteorological and hydrological factors based on the least-squares boosting decision tree was established,and a Gaussian mixture model of the multi-step runoff forecasting error sequence was constructed to quantify the uncertainty of the mid-and longterm runoff forecasting.The application results of the Three Gorges and Ertan hydropower stations showed the proposed optimal operation function considering multi-step runoff forecast information had strong robustness and was effective and practical.It can effectively reduce the influence of multi-step runoff forecast uncertainty while making full use of the runoff forecast and its uncertainty information,and provide reliable decision reference values and decision reference intervals for dispatching decision makers.