Study On System Identification And Control Law Under Complex Conditions Of Pumped Storage Unit

Clean energy such as wind and photovoltaic power can be connected to the power grid on a large scale and play their green advantages of high-quality energy only by the absorption and coordination of large-scale energy storage technology.Among the existing energy storage technologies,pumped storage unit has the largest reserves and relatively mature technology,which has outstanding advantages compared with other forms of energy storage technology.Pumped storage power generation has the advantages of fast load response,high operation flexibility,and fast condition conversion,which can almost track the load change of power grid in real time.However,there are still bottlenecks in the joint development of pumped storage and renewable energy.The access of renewable energy puts forward new requirements for the operation mode of pumped storage units.New operation modes such as longtime no-load standby,long-time load operation and more frequent load change bring new challenges to the operation of pumped storage power station.Especially in the aspect of stability and safety,due to the inherent anti "S" unstable operation characteristics of reversible units and the increasingly significant coupling effect in the process of hydraulic-mechanical-electrical energy conversion of the govening system,the traditional pumped storage technology has been unable to meet the requirements of future power system.Under this background,in view of the key scientific problems and technical difficulties that need to be solved urgently for the stable,safe and efficient operation of pumped storage units,this paper takes the system identification of pumped storage units and the study of control laws under complex working conditions as the starting point.On the basis of fully discussing the dynamic mechanism and nonlinear characteristics of each component of the pumped storage regulating system,the speed governing system with complex conduit system and hydraulic-mechanical-electrical coupling system are established respectively.Based on intelligent optimization algorithm,artificial neural network,multiobjective optimization theory,eigen analysis and model predictive control method,the research on parameter identification,model identification,improvement of dynamic characteristics of anti "S" area and coupling control law of speed and excitation regulation system are carried out.The modelingidentification-control of pumped storage units has been established systematically.The main innovation achievements and research work of this paper are as follows.(1)The different model expressions and applicable conditions of the speed governing system and excitation system of the pumped storage unit are studied systematically.In view of the difficulty in modeling the S-shaped region of pump-turbine,the logarithmic projection method and the improved Suter transformation are introduced to preprocess the whole characteristic curves of pump-turbine,which solves the multi-value problem of interpolation calculation in the S-shaped region.The governing system model with complex conduit system and the hydraulic-mechanical-electrical coupling model are built,which are suitable for different research conditions.The model foundation is established for the subsequent system identification and the research of control law under complex conditions.(2)In order to solve the problem of parameter identification caused by the coupling characteristics and complex conduit system,the method of parameter identification based on intelligent optimization algorithm is studied.An improved intelligent optimization algorithm is proposed by introducing artificial sheep algorithm and combining with levy walk,chaos mutation and elastic boundary treatment strategy.A parameter identification framework based on improved artificial sheep algorithm is established.The high-precision modeling of the complex conduit system is realized through the on-off process of the unit,and the high-precision integrated parameter identification of the hydraulicmechanical-electrical coupling model of the governing system is realized through the regulating process of the grid connected operation.(3)In view of the long-term dependence of data and the problem of gradient disappearing in the training of general neural network,the long short-term memory(LSTM)neural network is introduced to realize the high-precision off-line model identification of the speed regulating system of pumped storage unit with complex conduit system.Aiming at the problems of slow training convergence and difficulty in online adjustment faced by common back-propagation algorithm,an online sequence extreme learning machine with forgetting factor(WOS-ELM)is introduced,which has the advantages of strong nonlinear description ability of common BP neural network and simple calculation of recursive least square method,to realize high-precision online model identification of hydraulicmechanical-electrical coupling system of pumped storage unit.(4)Aiming at the problem that the speed oscillation is easily caused by the S-shaped characteristic in the start-up process under low water head,a multi-objective optimization(MOO)framework considering the rapidity and stability is proposed and the multi-objective artificial sheep algorithm(MOASA)is used to solve the problem,which can effectively suppress the speed oscillation of the unit in the start-up process under low water head.In order to improve the dynamic characteristics of the pumped storage unit in S-shaped area fundamentally,this paper discusses the feasibility of using the variable speed mechanism to avoid the unit operating into the S-shaped area for the first time.The results show that the unit can avoid the S-shaped area effectively under the low water head condition by reducing the rotating speed to move the operating area of the full characteristic curves to the left,so that the unit has better dynamic characteristics.It also provides a new reference for the transformation and development of conventional fixed-speed pumped storage units.(5)In order to realize the control of the hydraulic-mechanical-electrical coupling system,the small signal stability analysis of the regulating system is carried out by using the eigen analysis.On the basis,the multi-objective control strategy and decision-making method of the classic "PID+AVR+PSS" is proposed.An intelligent model predictive control(MPC)strategy based on online sequence extreme learning machine with forgetting factor,stair-like control incremental constraint and rolling optimization with artificial sheep algorithm is proposed.The effectiveness and superiority of the proposed intelligent model predictive control method for speed and excitation coupling control are verified by comparing the experiments under different working conditions with the classical control strategy.The stability of intelligent model predictive controller is analyzed by nonlinear dynamics theory.