Data-Driven Modeling And Data-Driven Control In Smart Grids

Modeling the accurate physical models for these plants may be impossible or infeasible in smart grids.Even if their physical models are available,they may be too complex to be tractable for controller design,system monitoring,performance evaluation,etc.However,huge amounts of measurement data can easily be collected due to the well-developed information technology,both in the form of stored historical data from prior measurements and online data in real time during process runs.It would be very significant if we can make full use of those online or offline measurement data to directly design controller,predict and assess system states,evaluate performance,make decisions,perform real-time optimization,and conduct fault diagnosis.In this dissertation,in-depth studies have been conducted on the data-driven modeling and data-driven control in innovations and applications of smart grids,and the following research results have been achieved:Different from conventional load modeling methods,this section proposes the sparse learning methods to solve following two key problems,one is the coordination of the accuracy and conciseness of load model structure,the other is the parameters calibration of the load model,which achieve structural selection and parameter calibration of load models simultaneously.A novel data-driven method is proposed to estimate equivalent damping coefficients and topological parameters of the network-reduced model simultaneously so that we can identify the network-reduced model to characterize low frequency oscillation.More specifically,the proposed method utilizes the sparse representation to select the most dominant nonlinear terms from a set of dictionary functions,which finally balances the data fitness and achieves dynamics learning.There are low frequency oscillations in power systems.Locating low frequency oscillation sources exactly and taking appropriate control measures to enhance system damping and suppress system oscillations are an important guarantee for the secure and stable operation of power grids.It is necessary to find mechanism of the negative or weak damping of low frequency oscillations to interpret and propose effective criterion of location.We cut in this problem from perspective of nonlinear analysis and use the power angle relationship to equivalent the second-order rotor model.Furthermore,we utilize Hamilton analysis to reveal mechanism of oscillation.Accordingly,the problem of locating low frequency oscillation is transformed to identify equivalent negative damping coefficients.Propose a deep Koopman model predictive control strategy for improving transient stability of power grids in a fully data-driven manner.Due to the high-dimensionality and the nonlinearity of the transient process,we use the Koopman operator to map the original nonlinear dynamics into an infinite dimensional linear system.To facilitate the control design,we first utilize the deep neural network method to efficiently train observable functions to approximate Koopman operator so that the obtained dynamics in the higher dimensional space is a linear system.We then propose a model predictive control strategy for the obtained higher dimensional linear system.The proposed control scheme utilizes energy storage units which inject or absorb real power at the synchronous generator buses to enhance transient stability.Study a framework of system-wide stability assessment for future distribution systems that are comprised of multiple microgrids.Especially,we study the system-wide stability of coupled microgrids under the practical situation where the transmission line is lossy.Enabled by the advances in power electronics and synchrophasor technologies,the droop control method is introduced for autonomous real power sharing among coupling-operated microgrids.The steady state stability of coupled microgrids are analyzed through smallsignal stability approach considering both frequency and voltage dynamics.The condition of system-wide stability is then provided with the designed Lyapunov function,and the criterion solved by the linear matrix inequality is based on Lur’e system theory.At last,this dissertation summarizes the entire research,and proposes a prospect for further research on the data-driven modeling and data-driven control,and the future development direction.