| OCR Text |
Show o c: Q) 0.25 -r--------------------, loads > 250 MW with Q2 1Q5 5- 0.15 ~ I.L. Q) . ~ 0.1 ~ 110 Without GNOCIS I 0.05 iowlth GNOCIS , I O ~I ~~ __ ~~~~~~~--~~~~ o 2 4 6 8 10 12 14 16 18 20 Fly Ash Carbon, Percent 0.4 ~---W-l".t,hCo-u-t-:-----~W:I::tti;h ::----------, 0.35 88.1 ' 88.5 load> 250 MW ~ 0.3 c: ~ 0.25 a- ~ I.L. 0.2 ~ Q) ~ 0.15 ~ Qj a:: 0.1 0.05 o WIthout GNOCIS I IOWlth GNOCIS O ~ __ ~~~~--~~~~~~----- 87 87.5 88 88.5 89 89.5 90 Boiler El'ficlency, Percent Figure 4. Gaston / Maximize Efficiency / Closed-Loop Since these initial tests, GNOCIS has been converted to closed-loop operation. Although testing is yet to be completed, example results are shown in Figure 6. Data shown in this figure are from when the unit is under economic dispatch and between 2S0 and 270 MW. Also, excess O2 was excluded from the optimization determinations (i.e. no recommendations were made for O2) . As shown, LOI was reduced by approximately 2.S percent and boiler efficiency improved by nearly 0.4 percent. THE KINGSNORTH TRIAL Kingsnorth Power Station is located near London. Unit 1 is a SOO MW tangential-fired unit equipped with a NEt International Combustion boiler and is capable of meeting full load on either pulverized coal or residual fuel oil. The furnace has a central vertical dividing wall which forms two identical combustion chambers. The four burner boxes in each chamber have independent tilt control which can be moved nominally between +20 and -200 from the horizontal. Each mill fires a single level within the furnace. The furnace is fitted with a low NOx firing system with separated and closecoupled overfire air. The digital control system at Kingsnorth uses an in-house system known as CUTLASS which is based on DEC PDP 11 hardware and tnstem I/O equipment. The primary objective at Kingsnorth is to minimize CIA !n the fly ash while maintaining NOx below the current level of 390 ppm. With the current DCS configuration, only seven parameters are adjustable by the operators - burner tilts (ganged together as one setting), excess air, and five mill settings; therefore these parameters were selected for use in GNOCtS. Model Development The data acquisition system was receiving and storing data twenty-four hours a day throughout the Kingsnorth trials. However, not all data was suitable for use in the models; in its raw form the data covered periods when the instrumentation was faulty and when the plant was operating in a regime outside of the GNOCIS specification (zero and low load). Data was therefore preprocessed to remove invalid data and data not corresponding to GNOCIS operating regimes. Predictive models were constructed, as a first step, to give an indication for the overall accuracy of the modeling and to highlight potential difficulties. The performance of the model was evaluated by selecting up to four periods of operation, omitting these from the input data and then running the model on this data. Similar to the predictive model evaluation, four periods of operation were chosen against which the model performance would be assessed. This was done for the control models by taking expert advice on what information should have been given to an operator to prompt any desired control action. The experts were the Kingsnorth Efficiency Engineer and a PowerGen combustion expert. This advice was then compared with control advice suggested by the model. S |
