Study On Real-time Operation Of Cascade Hydropower Stations Considering Runoff Forecast

With the global climate problem becoming more and more serious,the use of clean and renewable energy is an important measure to realize the adjustment of national energy structure.Among them,hydropower accounts for a large proportion of clean energy,and its development technology is relatively mature.In the past 20 years,China’s hydropower industry has developed rapidly,and the installed capacity of hydropower has made continuous breakthroughs.With the continuous improvement of hydropower installed capacity,the optimal operation of hydropower system is facing a huge challenge.In short-term and real-time optimal scheduling,there are two main difficulties: one is the uncertainty of the incoming water.There are some defects in the current runoff prediction method,which makes it difficult to achieve the ideal effect.There are differences between the predicted inflow and the actual inflow,which may lead to serious risk of water abandonment or water level exceeding limit during the operation and operation of the reservoir.The second is the uncertainty of load.Due to the diverse topography and special climate conditions in Southwest China,coupled with the impact of holidays,the load of power grid is generally difficult to predict.There is a big deviation between the actual production load and the planned load in the scheduling scheme,which seriously affects the safe and stable operation of the power grid.Based on this background,this paper focuses on the runoff forecast method considering flood and dry season stages and the real-time optimal dispatching model responding to grid load adjustment.The main research results are as follows:(1)In view of the problem that BP artificial neural network is easy to fall into local optimum,a staged prediction method of reservoir inflow in the middle period of hydropower station considering flood dry transition based on artificial neural network is proposed.According to the historical runoff data of dry season(flood season),the runoff forecast model of dry season(flood season)is established and the parameters are calibrated.In the dry(flood)period of the basin,the runoff forecast model of the dry(flood)period is used to get the forecast results of the dry(flood)period.In the transition period of flood and dry season,two kinds of prediction results are obtained by using two kinds of models respectively.Then,according to the membership degree of flood and dry season in the current time,the two results are weighted and averaged to get the prediction results of the transition period,so as to improve the prediction accuracy of the inflow runoff of the hydropower station reservoir.(2)Aiming at how to respond to the adjustment of real-time generation plan quickly and safely,taking Beipanjiang cascade hydropower stations as the research background,combined with the real-time scheduling related problems encountered by Beipanjiang centralized control center in the actual production and operation,this paper puts forward the objective function of minimizing the residual demand of power grid load during the scheduling period,comprehensively considering various complex constraints,and uses a test method The real-time optimal scheduling model is solved by considering the daily power deviation.