Modeling And Research On Direct Auxiliary Power Consumption Of Pumped Storage Power Station Based On The Support Vector Machine

As an important technicaland economic indicator measuringeconomic operation, energysaving and emission reduction of the Pumped Storage PowerStation, the plant powerconsumption ratio has a direct impact on the prosecution and operatingefficiency andenergyconservation and emission reduction effectiveness of the Pumped Storage PowerStation.Currently, in-depthresearch on the power consumption ratio of the Pumped StoragePowerStation both at home and abroadis relatively lacked,and mostonlyfocus ontheoverallstudyofcomprehensive plant power conservationratio. Therefore, what is of greatsignificance is to refine andanalysis the plant power consumption situation, and toscientificallyconstruct the plant power consumption calculationmodel. Research canprovideascientific basis forthe plant power consumption analysis and management assessment, and bebeneficial to promote the meticulous management of the Pumped Storage PowerStation, andto improve energy saving management level and prosecution and operationefficiency.The paper analyzed and studiedtheplant power consumptionstructure and management,assessment indicators and influence factors of the Pumped Storage PowerStation, andproposedreplacingthe comprehensive plant power consumption ratio with the directplantpower consumption ratio asassessment indicators of plant power consumption, and illustratedtherationality. On the basis of research on the fundamental theory and application of SVM,K-meansclustering sample optimization and particle swarmoptimization， the paper putforward the SVM regressionmodel based on the K-meansclustering optimization and particleswarm parameter optimization, toachieve the target that improving the plant powerconsumption modelaccuracy of the Pumped Storage Power Station and controlling thecalculation errorof the whole year within5%.The plant power consumption calculation model of the Pumped Storage Station studiedin this paper was divided into two parts. First of two, the overall calculation model, namely,regressionmodel of the direct plant power consumption and generation, pumping powerconsumption, temperature, hydrology and other factors constructed with SVM based on thehistorical data, had the overall direct plant power consumption fitting calculation andprediction capabilities; second of two, modular computing model, applying module divisionto plant power consumption according to the system category, and calculating powerconsumption of each module based on the qualitative analysis of the relationship of the directplant power consumption and generation, pumping power consumption, temperature,hydrology, operation rules output curve and other factors of each device and the quantitative calculation of the electrical equipment parameters, the generation sets start-stop conditionsand run-time and other data, meanwhile, constructing modular fitting model with theapplication of SVM, had the function of calculation, fitting and forecasting of the plant powerconsumption of each module. Two parts of the model can realize the function of checking,analysis and assessment of for the year and prediction of plant power consumption for thenext year of the Pumped Storage Power Station of CSG Peak andFrequencyRegulationGeneration Company through combination and verification mutually.Eventually, the paper established the direct plant power consumption calculation modelof Huizhou Pumped Storage PowerStationand developed the supportingapplicationsoftware,the final calculationresult verified theaccuracy andapplicability of themodel, and the precisionmet the energy savingand emission reductiontarget that the PeakandFrequencyRegulation Generation Company controls the plant power consumption ratioand standarderror of its jurisdiction within5%, which is clearly required in the CSG " TwelfthFive-Year " energy savingand emission reduction plan.