Research On Accurate Load Modeling Considering Load Time-varying Characteristics

Power system digital simulation is the foundation of system planning,design,scheduling,operation,and analysis.The difficulty of load modeling lies in dealing with the randomness and time-varying characteristic of the load.Due to the relatively simple structure and low parameter timeliness of traditional load models,it is difficult to accurately describe the comprehensive time-varying characteristics of various load equipment in the new power system.The adaptive interactive multi-model algorithm is an effective means in the field of maneuvering tracking to handle variable structure system and time-varying parameter problems.It transforms the process of identifying load model parameters into maneuvering tracking of load characteristics,providing strong support for modeling work considering load time-varying.Taking large capacity variable speed pumped storage units as an example,constructing a generalized load model for those new components has become an urgent need to enrich load modeling work under the current situation.This article focuses on the two key issues mentioned above and is committed to promoting the construction of generalized load modeling towards variable speed pumped storage units and the research on adaptive interactive multi-model load modeling methods.The specific content includes:(1)In order to facilitate a physical understanding of the transient behavior of variable speed pumped storage units,a dynamic model based on the energy coupling relationship of internal components of large-capacity variable speed pumped storage unit is proposed in this paper.This model takes doubly-fed variable speed machine and grid-side converters as interfaces with the grid,including pump turbines and electromechanical coordinated control,and considers the low voltage ride through mechanism of actual variable speed pumped storage units under large disturbances.The model of doubly-fed variable speed machine is divided into two categories that can be switched to each other according to the working state of the AC excitation system.A unified electrodynamic force model motivated by power of doubly-fed variable speed machine and grid-side converters in both generating and pumping states is constructed based on the energy coupling relationship of the main energy storage components within variable speed pumped storage unit.It exhibits the characteristics of fast response of power decoupling control,frequency stability support,and short-term asynchronous operation of doubly fed variable speed machine caused by excitation system locking under large disturbances.The simulation verifies the effectiveness of the proposed model and uses the identification method to identify the parameters of the expanded comprehensive load model in the constructed scenario.(2)To improve the tracking performance of interactive multi model load modeling methods for time-varying loads,this paper proposes a method to adjust the transfer probability in real-time based on system model information.The transfer probability matrix is an important factor that affects the input interaction.The likelihood function of each load model contains information about the matching degree with the current actual load characteristics,which can be used to distinguish whether the load model needs to be switched.All likelihood functions are obtained through parallel filtering,the weights of each model are updated,and the estimated values of the output state variables and their covariance are fused.When the change of the likelihood function meets the model switching threshold conditions,the transition probability matrix is adaptively adjusted using the gradient of the weight update,which helps to overcome the response lag problem and the phenomenon of instantaneous error peak in the process of model switching.(3)Aiming at the problem of vicious competition among models in interactive multi model load modeling,which leads to a decrease in tracking accuracy,a method for power load time-sharing classification model library and model selection mechanism is proposed.The domain knowledge of load modeling is stored structurally by using the related technology of knowledge graph,and the time-sharing classification load model library is initially constructed.Extracting information on various relevant factors that affect the real-world time-varying load,and calculating through fuzzy rule reasoning and selection algorithms,the real-time optimal model set is obtained to participate in input interaction.This paper provides theoretical and methodological support for accurate online modeling of time-varying loads by constructing a generalized load model for variable speed pumped storage power plants and proposing an interactive multi model transfer probability adaptive adjustment method and model selection mechanism based on the power load time-sharing classification model library.