A Study On Pulverized-coal Concentration For Power Plant Boiler By Soft-sensing Theory

For pulverized coal fired power plant, the pulverized-coal concentration in pulverized coal blast pipes is very important to safe and economical operation of power plant boiler. So it is very necessary to measure the pulverized-coal concentration on-line. Because of the complex of gas-solid two-phase, it is short of a credibility parameters measurement for a long time. It is significant to do research on the measurement of pulverized-coal concentration.Recently, soft-sensor measurement solves so many problems hard to measure by the method of hardware and it provides a way to solve the problem mentioned above. The paper solves the problems about measurement for pulverized-coal concentration based on the clue of researches on soft-sensor theory .The main contents of this dissertation are as follows:1. Summarizing current status of data pretreatment methods. Data collected from locale takes errors because of influence by instrument or circumstance ineluctability .For chances errors, the paper provides a method based on signal filter and combining two or more ways to reduce random disturb. The paper puts forward a method to detect and proofread gross errors based on principal component analysis .2. Solving the lag problem of measurement for pulverized-coal concentration in boilers whose pulverized-coal is transport by hot air. The paper proposes a change coefficient compensation method based on temperature .It has applied in the project and received good effect.3. The paper focuses on a method for calculating the concentration of pulverized-coal transported by depleted exhaust gas based on gas-solid two-phase flow theory. Furthermore, an experimentation has been performed. Based on abundance data acquired from locale, combination mechanism and regression analysis, we obtain an algorithm model. The result is inspiring.4. As an algorithm of the computer study, the support vector machine resolves a large number of practical problems such as small sample, high dimensions, over learning, local minimum, and this algorithm has very well generalization ability. This thesis applies the support vector machine theory in the soft sensor model , introduces to the support vector machine in detail of related theories, and a breeze powder soft-sensor model based on SVM is given in the paper. By computer simulation, The paper contrasts the soft-sensor model based on SVM with RBF Network algorithm which is applied very abroad in the field of soft sensor .Simulation analysis indicates that the generalization ability of the SVM soft-sensor model is better than nerve network soft-sensor model.5. There is so much work to do if we want to gain practicality value from soft-sensor theory. The paper analyzes the choice of hardware and the software framework in detail for the measurement of pulverized-coal concentration on-line. It presents a software design ideal based on COM. The software features open and common-use by developing modules.