Study On Multi-time Scale Nested Runoff Forecasting And Power Generation Scheduling Of Hydropower Stations

Hydropower energy,which is the high quality energy known as "never failure",has the feature of low cost,high efficiency,renewable and pollution-free.With the large-scale systematic development and utilization of hydropower energy,the function of large reservoirs dispatching is more and more obvious and the status of that is rising.This paper focused on the question of power generation dispatching of reservoirs in medium-long-term.The project's background is the Xiluodu hydropower station of Jinsha River downstream.According to the runoff forecast model in different time scale,the accurate runoff prediction results could be obtained,which could be used as the input of dispatching model.The model and algorithm of reservoir generation dispatching were deeply and systematically studied with the conventional dispatching model,optimal dispatching model and the multi-time scale nested model.The high-precision runoff forecasting is set as a scheduling plan to provide input.And the rolling correction and feedback affections of multiple time scale nested model on dispatching plan were explored.The main research results in this paper were summarized as following:(1)Research on medium-long-term runoff forecast based on the characteristics of abundance and drought.BP neural network was used to predict runoff.Considering that the accuracy of weather quantitative forecast in medium-long-term was low and the precision of qualitative forecast was high,whether the runoff forecast was plenty or lack was analyzed when forecasting the process in medium-long-term according to the meteorological forecast.And the meteorological factor was transformed into historical runoff rule to be considered.The historical data of the same period of abundance and drought were selected as model learning samples to train BP neural network and to determine the parameter rate in three time scales of ten,month and year.According to sample learning and prediction results,the prediction accuracy of the model could be tested.And the prediction rule of abundance and drought BP neural network model in various time scales was explored.What 'more,the runoff forecast was selected as the input of the reservoir dispatching system.(2)Research on power generation dispatching of reservoirs in medium-long-term based on runoff forecast.The typical runoff analysis method was used to draw the reservoir dispatching diagram which was used to guide the operation of reservoirs in medium-long-term.The optimal operation model of reservoir power generation was established and the dispatching model was optimized by using the stepwise optimization algorithm.Runoff forecasting of drought BP neural network and the annual actual runoff forecasting were used as the scheduling input.The process of planning operation and the actual operation of hydropower stations was solved by conventional methods and optimization methods.The rationality of drought BP neural network model in setting the hydropower dispatching planning was further verified.And change rules of power generation dispatching of reservoirs in medium-long-term were revealed.Model selection is given for multi-time scale nested modeling.(3)Research on real-time modification of hydropower station dispatching plan based on nesting of multi-time scale.The nested coupling model of reservoirs in multi-time scale was proposed based on the nested relation among models in different time scale.The short-time scale model and the long-time scale model are nested with each other based on the actual incoming water and runoff forecast in different time scales.Whether the actual operation met the water level constraints of dispatching plan in long-time scale was considered to realize the feedback control.What 'more,the effective model solving method was analyzed based on stepwise optimization algorithm to have the dispatching plan more satisfy with the actual operating conditions.