A Study On Smart Grid Transient Failure Detection And Current Overload Preventive Control

With the rapid development of our country’s economy, the electric power demandis increased dramatically and the power system becomes larger and more complex.Large numbers of electric component integrated in the power system expand the safetyissue. Because the power grid is a nonlinear, large scale, tight coupling and dynamiccomplex system, the conventional power grid monitoring and control lack tightcooperation among measuring, computing, control and communication. However, theconcept of smart grid offers new opportunities to enhance the safety, flexibility andefficiency of power grid. Therefore, investigating new theory and approaches for thesmart grid have important practical significance.This dissertation investigates how to improve the safety of the power grid intransient, and proposed new schemes to deal with it. These schemes apply the faultdiagnosis method and a corrective control method. We construct a dynamic analyticalmodel of power grid integrated unified power flow controller, according to thetransmission and the dynamic characteristics of the power system. We discussed twofault detection approaches for power system transient process based on LRGF (LocallyRecurrent Global Forward) neural network and an adaptive principal componentextraction algorithm. For the security control in transient, this dissertation proposed anew corrective control strategy with unified power flow controller derived fromLyapunov method.①Proposed a pole assignment locally recurrent global forward(LRGF) neuralnetwork for modeling the power grid transient. The neural network divides the pole ofthe dynamic system into real pole and complex pole to avoid the complexity in stabilityanalysis. The proposed neurons in hidden layer are divided into real pole IIR andcomplex pole IIR for obtaining a better model. A switching function weights the two IIRfor getting the neuron output. The neural network is training by BP algorithm under thestability constraint.②Proposed a adaptive lifting scheme and threshold for residual signal analysis inpower system transient fault detection. The residual signal is figured out by subtractingthe output of the original power system with the output of the neural network. Then theresidual signal is extracted into a approximation signal and detailed by a lifting schemefor the fault information may hidden in the signal. At last, the approximation signal and detailed signal are analyzed by adaptive threshold scheme. The simulation results showthe effectiveness of the proposed fault detection scheme.③Proposed a power system transient fault detection scheme based on an on-lineadaptive principal component extraction algorithm. The on-line adaptive principalcomponent extraction algorithm can analyze high dimensional data and extract theprincipal component very fast, it has some advantage such as low computational costand high precision. The on-line data of the power system is analyzed by this algorithmto extract the principal component of the data adaptively, and then the T2statistics andQ statistics are figured out from the principal component transformation results to detectthe transient fault, the T2statistic imply the inner variation of the PCA model, in anotherhand, the Q statistics reveal the outer variation from the PCA model. The simulationresults of the fault detection scheme demonstrate the effectiveness of the proposedmethod.④For dealing with the instability in transient caused by contingency, thisdissertation proposed a new Lyapunov method based corrective control scheme withunified power flow controller. A control Lyapunov function is constructed by energyfunction and barrier functions for derive the control law. Compared with the simulationbased corrective control scheme, the time derivative of energy function is semi-negativedefinite, which bring much benefit in control design, and the barrier function restrict thecurrent flow on the transmission lines, when the state approaches the constraintboundary, the value of the barrier function will go to infinity. So that the system statewill converge to the stable working point without current overload on transmission lines.The simulation result of3bus system and162bus systems demonstrate theeffectiveness of the proposed corrective control scheme.