Research On Forecasting Technology For Bus Load And Impacts Of Its Characteristics On Power Grids

An accurate short-term bus load forecasting is the foundation of realizing energysaving, consumption reducing and meticulous management of dispatching. Theresearch for bus load forecasting technology on power system has important practicalvalue and theoretical significance. Combining with current researches both domesticand abroad, on the basis of analyzing the contents and characteristics of theshort-term bus load forecasting, the existing problems in this aspect was elaborated.Specifically, electrified railway load is one kind of single-phase and nonlinear busload with strong impact. The impacts or hazards to the power system, leading by thepenetration of220kV high-speed railway traction load and the power quality problemsof110kV electric locomotive load, are worthy of attention and study. Focusing onanalyzing and studying certain issues of the short-term bus load forecastingincludingbad data pre-processing, similar days selection, forecasting methods, simulations ofthe high-speed railway traction load, negative sequence and harmonic characteristicsof the electric locomotive load, etc, the following specific work had been done:As an important link, the original data analysis would improve the accuracy ofshort-term bus load forecasting a lot. Thus, a bad data processing strategy based onstratified analysis of characteristic matrix was presented. Firstly, the AFS clusteringalgorithm for dividing sample set optimal clustering structure was studied. Referringto the characteristic curves, the horizontal and vertical eigenvectors reflectingproperties of the load points were calculated, and the characteristic matrix was formed.By applying the discriminant criterion, the stratified analysis for the characteristicmatrix of daily load curve was carried out, thereafter the corresponding bad dataprocessing strategies focusing on bus loads which had different variation ofcharacteristics were established.Appropriate selection of the similar days can effectively improve the accuracy ofbus load forecasting, and an integrated forecasting method based on the selection ofoptimal similar days was presented. In order to obtain a set of curves whose load levelis similar to the target date, the effects of daily characteristics factors on the loadlevel was calculated, and fuzzy object function is weighted over each factorimportance degree. Establishing discriminant function of all sets with similar shape,and target date was classified. Taking the intersection of load level set and curve shape set as the result of similar day selection. Sample with minimum date interval totarget date was selected as virtual predict object, and then the weight of eachalgorithm in integrated forecasting was calculated.A forecasting method based on decoupling mechanism was presented.Forecasting process was divided into two parts, i.e., the load level prediction andper-unit curve prediction, and prediction strategies were developed to adapt to theircharacteristics respectively. The results were selected according to the similar loadlevel, and the similar sets were trained by utilizing the least square-support vectormachine method for predictions. In addition, the per-unit curves of the target datewere classified and the weighted average processing was carried out upon thesehistorical per-unit curves according to the similarity.High-speed railway load, as a typical high power bus load with impact, waschosen as the study object with its influence upon power grid in operationanalyzed.The characteristics of the high-speed railway traction load were primarilyanalyzed, and then PSCAD was utilized to build suitable traction power supplysimulation system for high-speed railway. The piecewise linear U-I curve of UMECmodel was equivalent to saturated characteristics of the iron core, and then madeno-load closing simulation. On this basis, simulations of single-phase short circuitfaults and switch-in with load when three-phase short circuit faults in line side wereperformed.Electric locomotive load was selected as the instance of single-phase nonlinearbus loads, with its negative sequence and harmonic characteristics studyed.In order todeeply analyze the impacts of the negative sequence component on power grid fromelectric locomotive load, establishing the negative sequence source model suitable fornegative sequence characteristics analysis and power flow calculation was vitallyimportant. A modeling method based on the classification and synthesis of negativesequence characteristics was proposed. The clustering tendency analysis wasconducted on the sample set, with the measured response space of fundamentalnegative sequence current and the actual running state of locomotive as eigenvectors.On the premise of clustering possibility, the clustering validity function was used,aiming at obtaining the optimal classification result of the sample set. By means ofmechanism analysis for the negative sequence characteristics of traction load, themodel structure was determined, while the optimal model expression was achievedapplying the method of stepwise multiple regression. Besides, the newly addedsample was classified into the class within the minimum Euclidean distance from the clustering center, and the related model was tested.According to lots of measured recorded data of electrified railway and by use ofthe modeling idea based on overall measurement-identification, harmonic sourcemodels under different traction operation conditions were built, in which three-phasefundamental positive-sequence active and reactive power exchanged betweenlocomotive and traction power supply system were regarded as the energizing and thereal and imaginary parts as the responses, and then recommended parameters for eachkind of harmonic source model were determined.On the basis of the negative sequence source model and harmonic source model,a system for computing power flow was achieved. The program included thefundamental negative sequence power flow calculation and the harmonic power flowcalculation. By simply inputting the data of generators, transformers, transmissionlines, general loads and traction loads, the users could obtain the harmonic or negativesequence voltages for all network nodes, and thereafter can solve harmonic ornegative sequence currents of each branch.In order to realize distributed management of load conveniently for the ElectricPower Dispatching Center, a bus load forecasting system for Hunan power grid hadbeen developed. The system, which was based on the Microsoft Visual Studio.NETplatform and Microsoft SQL Server database, employed the advanced multi-layersystem Browser/Server (B/S) as its structure and was consisted of six modules,including bus load forecasting, bus data query, bus load analysis, bus load statistics,reporting and evaluation and system management center.