Study Of Power System Parameters Identification

The param eters of generator system a nd load m odel are the basis of power system analysis, control, monitoring and planning. Its Un-accuracy directly affects the security, stability and economy of power system operation, dispatch, and planning. On one hand, it m ight cause huge waste to inve st too m uch capital to take unnecessary measure to strengthen system architecture for the pessimistic analysis results. On the other hand, the system might take large risk to operate at dangerous situation, even at great possibility to collapse for the aggressively optimism analysis results. Therefore, how to obtain accurate parameters of generator system and load model has been a hot topic for years.To focus on the def icits and dif ficulties of parameters of generator system and load model, a thoroughly exploration and st udy have been done in this thesis and many satisfactory achievem ents are obtaine d. The m ain achievements and research contents are listed as follows.Firstly, a new on-line m ethod of unit para meters identification based on Digital Fault Recorder for generator system is pr esented. All the mathematical model for all sub-systems, including generator, exciter and governor-turbine, are first derived out in the method; then the function relationship of input, output and the variables whose parameters needing to be tested is acquire d; the optim al values for the param eters concerned are achieved by iteratively adjusting and optimizing their values around the initial values provided by m anufactories through PSO(Particle Swarm Optimization) algorithm, in order to let the m odel outputs of each sub-sy stem optimally match the real outputs recorded by DF R; after the iteration is conver gence, the optim al parameters are gained. The advantage of the method is that it is only based on the data recorded by DFR to realize all the parameters identification without doing experiment on the field. The results of the example s hows that the method has high identification accuracy, fast calculation speed, and is convenient for the practical application.Secondly, a new m ethod for load m odel and new method for param eter identification are further proposed. By i mproving the linearized-GNLD model, a new load model with higher accuracy of curve fitting is achieved. Then, a new m ethod of parameter identif ication f or load mode l based on random fuzziness clustering algorithm is presented, the idea of which is to cluster the sample load curves first, and then to identify one set of param eters for each cluster of sample load curves. By th e method, node voltage, activ e power and reactiv e power are set as feature vecto r to classify all the collected sam ple curves in to multiple g roups curves , each grou p corresponding to a load model and a set of parameters identified for it, which could be also called to class ify the load character istic by the load curves, and realize the synthesis of each g roup curves in a certain extent. Thus, th e model obtained by the method proposed has higher accu racy and strong promotion capability than those b y other methods.The methods of modeling and parameter identification for generator system and load model proposed in the thesis has high practicability; so the software developed based on it has a strong potential to be applied for practical power system...