Research On Data Preprocess Method For Thermal Parameters

Energy supply in China is dominated by thermal power at present. Thermal powerplant is complex industrial system as the main energy production for our country, and itshoulders great social responsibility, such as safe operation, energy saving, emissionreduction, and so on. In order to monitor the performance of various power equipmentplant many different aspects of analysis data is indispensable. Therefore it is an importantpremise for monitoring all kinds of power equipment system to measure the operatingparameters of power equipment accurately.In order to obtain more accurate measurement results from the sensor measurementthat is interfered by noise, this paper will give a comprehensive analysis of the results byusing a variety of treatments systematically. The thesis mainly talks about it from thefollowing aspects:(1)In order to solve the gross error existed in the thermal data, by improvingtraditional3σ criterion the method which is mentioned in this paper is used to test andmodify the gross error. Based on3σ criterion in data error theory, the improved methodwhich measures the rate of change of the observed quantity is introduced to preprocessgross error of the collected thermal data. Then this method should be used to test thegross error data and singular point of data, and modifying the random error. Theeffectiveness of proposed method is validated with simulation on coal grinding current;(2) To remove random errors existed in thermal measurement data, two methodswhich are the moving average method in the five point three moving average method andwavelet analysis filtering and the evaluation criteria are brought about. With comparisonof two denoising method to reduce the noise in the measurement data and the evaluationcriteria to analyze its results, such as mean square error, signal to noise ratio, smoothnessand correlation coefficient, there is a comprehensive conclusion that the waveletanalyzing method has more advantages on data denoising through the example ofcalculating the coal grinding current and furnace pressure;(3) In order to utilize measurements of multiple sensors, the multi sensor fusionmethod is used to improve the accuracy and reliability of measurement data. Based onthe theory of normalization and recursive estimation variance, multi sensor datareconstruction algorithm is constructed to deal with data in the thermal process. Though comparing multi-sensor fusion and the single sensor and computing the results, it isconfirmed that the accuracy of measurement data can be obviously improved by themethod of multi-sensor data fusion. The effectiveness of proposed method is validatedwith simulation on measured drum water level and boiler water flow.