The Implementation Of Power System State Estimation And Situation Awareness Based On Cellular Computation Networks

With the development of modern power system in the direction of high voltage,long distance and high capacity,the structure size of power systems is scaling up and becoming more complex.In the event of an accident,the impact will spread widely.Because of the penetration of renewable energy,distributed power supply,smart microgrid,demand-side management and so on,the prediction,control,and high-quality operation of power systems is more difficult.Dynamic state estimation(DSE)and realtime situational awareness of power systems is potential means for power companies to meet new challenges.The widespread use of Phasor Measurement Unit(PMU)makes it a new data source for dynamic state estimation and situational awareness of power systems.,This paper works on the dynamic state estimation,real-time situational awareness,and visualization of power systems based on Cellular Computation Networks(CCNs).The main contributions are as follows.(1)Considering the uncertainty of model parameters,measurement noise and distribution implementation during dynamic state estimation process,a CCNs based distributed dynamic state prediction and estimation method is proposed.According to the characteristics and advantages of the CCNs,an estimated CCNs model is built for distributed dynamic state prediction and estimation based on the physical connection of corresponding object components using PMU or generator measurements.In the process of CCNs implementation,a hierarchical scheme is designed to improve the prediction accuracy.The weight of each CCNs cell is optimized and trained using optimization method,and the proposed method is model independent and full data driven.The results show that the proposed method is not only suitable for distributed generator state prediction and estimation,but also works for distributed voltage and active power state prediction and estimation with high accuracy.(2)Considering continuous measure data loss,an adaptive innovation filter is designed based on N-step ahead state prediction.In the event of data loss,the missing measurement data is replaced with prediction state.The best state estimation is obtained using innovation when the prediction accuracy is lower than the reference error.With the hypothesis that the measurement matrix is a unit diagonal matrix,the innovation gain coefficient is determined by a priori error.The comparative results show that the proposed adaptive innovation filter based on N-step ahead prediction state has high robustness in the event of continuous measure data loss.(3)Considering the scalable and on-line realization problems to be solved during power systems dynamic state estimation,an online training and optimizing technique is constructed for CCNs based dynamic state estimation.With the idea of nonlinear CCNs state estimation,CCNs is implemented in state forecasting and filtering step respectively.An easy-to-train Echo State Network(ESN)is introduced in each CCNs cell to perform weight optimization training using a serial learning strategy.When the prediction error exceeds the error threshold,the output weight of ESN is updated by linear regression method,which is used to realize the online training of the ESN network weight.The power grid dynamic state estimation results show that the proposed online state estimation method has a high robustness in dealing with contingency,which has the potential for online realization.(4)A real-time situational awareness method considering transient stability and steady state security of power systems is proposed based on CCNs considering the realtime state tracking problem in current situational awareness methods.The proposed method introduces the energy function to evaluate the transient stability based on the power angle and speed estimation of the generator.A CCNs inspired hierarchical fuzzy system(CIHFS)is constructed for real-time grid safety monitoring.In the construction of proposed CIHFS classifier,fuzzy logic can transform the operator’s actual experience into machine-implementable rules through a linguistic design,to classify real-time security levels.And a CCNs inspired rule library can be adopted to reduce the number of rules for online security classification.During the contingency ranking process,taking into account the broken branch,an online contingency ranking indicator is proposed,meanwhile,a concentric relaxation method is introduced,and only the affected areas are analyzed to improve the ranking speed.(5)Based on the real-time transient stability and steady state security states,a visualization scheme including estimation and prediction results based on Web language is given.The proposed visualization design presents the results of voltage overslept,branch overload,contingency list system safety level,etc.to the operator in the control room with a vivid view through an efficient,practical,and easy-to-read graphical user interface.