Research On States Estimation And Optimal Control Of Micro-grid System Based On Bayesian Method

Microgrid is a key enabling solution for the reform and transformation of future smart grids by integrating distributed renewable generators,and the core technology for integrating distributed renewable to the main power system.It is also important guar-antee to realize the safe,clean,economic,efficient,flexible and reliable power supply of the smart grid.Compared with the traditional centralized main grid,it has the advan-tages of low transmission loss,power generation flexibility,high installation flexibility,high reliability and high security.However,due to the influence of environmental fac-tors on the power generation and load in the microgrid system,it shows high random-ness,volatility and intermittency.How to control and optimize the power flow in the microgrid has become one of the most important issues to ensure the economic operation of the microgrid on the basis of maintaining the uncertain balance relationship between the power supply and the load in the microgrid.One of the most important aspects to solve these issues is employing energy storage systems,which can not only effectively buffer the negative impacts of the distributed energy generation and load imbalance on the microgrid system,but also helpful for the optimal dispatch of the microgrid sys-tem,and can also improve the reliability of the power supply in the microgrid system.Lithium ion batteries have become one of the main choices of energy storage systems because of their advantages of high energy density,power density,low self-discharge rate and long cycle life.However,the Lithium-ion battery based energy storage system contains complex characteristics such as strong nonlinearity,state coupling,environ-mental sensitivity and aging-based performance attenuation,as well as the various dis-turbances due to inaccurate measurements in practical applications.How to accurately describe real-time dynamics and aging behaviors,accurately and robustly monitor state and parameters,evaluate the aging level of the energy storage battery,and optimize the energy dispatch in microgrid based on these factors,have practical significance for en-suring the security,economy,reliability and durability of the microgrid,They also play an important role in the the development and promotion of microgrid technology.In order to improve the accuracy and robustness of energy storage information ac-quisition and operation state perception in microgrid system,it is necessary to build the dynamic behavior description mechanism,and the theoretical foundation of on-line model parameters and state estimation for nonlinear time-varying systems with com-plex characteristics such as strong nonlinearity,state coupling,environmental sensitiv-ity and life decay.In order to optimize and control the power flow in microgrid based on uncertain information and uncertain optimization model,it is necessary to estab-lish an intelligent data driven energy optimization strategy based on evaluation of the load capacity of the energy storage system and the system constraints of the microgrid.The Bayesian methods have been introduced due to their advantages in the area of pa-rameter estimation and optimizations.At the meantime,the mathematical models for describing real-time dynamics and aging behaviors have been established based on neu-ral networks and stochastic process modeling.By combining the Bayesian estimation methods with these models,several approaches have been established for the on-line parameters estimation,health assessment and life prediction of the complex systems with nonlinearity,state coupling,environmental sensitivity,and life attenuation char-acteristics,which have also been applied into the energy storage systems.Based on the estimated state parameters of the energy storage systems and the predicted the power generation and load in microgrids,a data driven energy management strategy has beenproposed based on Bayesian optimization methods.The ground work and main contributions of this work can be summarized as:1)Aiming at the state estimation issues in nonlinear systems,the nonlinearity of the energy storage system have been addressed based on piecewise linearization and neural network tools.The model and parameter identification methods for the hysteresis char-acteristics have also been proposed.Aiming at on-line determining the model order,a data driven method has been proposed based on the combination of subspace identifi-cation method and particle filtering.To simultaneously estimate states and parameters for nonlinear and slow time-varying systems with parameters constraints,Constrained Bayesian dual filtering framework for simultaneous state and parameter estimation is designed.These works provide a theoretical basis for accurate monitoring of operation parameters and status of energy storage system under complex operation conditions.2)Aiming at determining instantaneous load capability of energy storage systems,a peak power capability determination method has been proposed based on real-time SOC estimation and terminal voltage prediction.The factors that influence peak power capability are analyzed according to statistical data,which provides an important basis for protecting battery and the optimal operation and control of microgrid system.3)Aiming at aging behavior description,an aging dynamic behavior model based on statistical modeling of Brown motion is proposed.Based on this model,the math-ematical expressions of state of health estimation and remaining useful life prediction are constructed.Besides,the parameter identification and drift parameter on-line esti-mation method have been proposed based on the combination of maximum likelihood estimation and particle filter algorithm,where uncertainty of the estimation and predic-tion is also quantified,Finally,the accuracy and robustness of the model and algorithm are verified by using different battery aging test set and aging threshold.4)Aiming at the state estimation issues of serially connected battery pack,a two time scaled dual filter is proposed to monitor the highest and lowest state-of-charge of the battery pack,which have also been employed to determine the peak power of the battery pack.In order to alleviate and eliminate the inconsistency of the in-pack cells,an equalization control method based on a two level parallel equalization topology and OCV(Open circuit voltage)estimation has been proposed.The simulation results show that the proposed OCV estimation method does not need to prior know the capacity and internal resistance of in-pack cells.The proposed equalization control method can also effectively alleviate the inconsistency of in-pack cells and improve the capacity of battery pack.5)In order to reformulate the economic dispatch problem,and solve it in a data-driven way,a data driven energy optimization strategy based on Bayesian-optimization-algorithm for a single grid-connected home microgrid has been proposed.The proposed solution formulates the optimization problem without a closed-form objective function expression.The objective cost function is on-line approximated by a Gaussian process.Then,Bayesian optimization can be employed to select the control decision that can maximize the profit of the microgrid.The experimental results show that this method can adapt to the change of objective function caused by parameter changes of microgrid,and is thus more flexible and reliable.