Access the full text.
Sign up today, get DeepDyve free for 14 days.
One of the most critical factors affecting boiler efficiency and hazardous-gas-emission reduction is the volume of excess air mixed with fuel. A knowledge-based approach is proposed to model the efficiency of a 320-MW natural-gas-fired steam power plant in Isfahan, Iran by applying fuzzy-modelling techniques to control the boiler efficiency. This model is based on fuel and air entering the boiler. First, the fuzzy-model structure is identified by applying the fuzzy rules obtained from an experienced human operator. The proposed method is then optimized using a genetic algorithm to increase the fuzzy-model accuracy. The results indicate that, by applying a genetic algorithm, the precision of the proposed fuzzy model increases. The error between the actual efficiency of the plant and the output efficiency of the proposed model is low. This model is developed by applying the fuzzy rules and modelling-related calculations. Finally, to optimize the efficiency of the boiler, a fuzzy proportional-integral controller is designed. The closed-loop control simulations are run by applying both the proposed controller and the manual controller to demonstrate the influence of the suggested method. The simulation outcomes indicate that the recommended controller adjusts the excess-air percentage correctly and increases the unit efficiency by 0.70%, significantly reducing fuel consumption. Received: 1 November 2020; Accepted: 5 March 2021 © The Author(s) 2021. Published by Oxford University Press on behalf of National Institute of Clean-and-Low-Carbon Energy This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (http:// creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com 229 Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 230 | Clean Energy, 2021, Vol. 5, No. 2 Graphical abstract Main Fuel Total combustion efficiency loop Main PI controller Main air Air Process New additional e Genetic control loop Excess air Algorithm Fuzzy Efficiency Model Manual Adj. Fuzzy PID controller Generating best Desired excess air value value Keywords: boiler modelling; excess air; fuzzy control; genetic algorithm; proportional-integral fuzzy controller and highly non-linear dynamic system. The modelling for a Introduction 1000-MW power-plant boiler is assessed in []. 9 Considering Nowadays, the optimization of fuel consumption in most the nonlinearity and time-varying quiddity of the combus- industries has become necessary due to the limitation of tion process, the effects of non-linear and time-varying vari- fossil fuels and the high annual growth in energy consump- ables are more significant for a large-scale boiler operating tion. Despite the increasing studies on alternative energy re- at higher energy levels. A load-dependent model for the sources, combustion in thermal power plants will maintain boiler is proposed in [10] by considering the main vapour its importance in the coming decades. Most of the energy pressure, the main steam temperature and the reheat consumed in the majority of industries is generated from steam temperature as important boiler-output variables. burning fossil fuel in furnaces and steam boilers. By properly Different classical methods like optimal control [11], pre- adjusting the excess air and increasing the efficiencies of dictive control, robust and adaptive control [12–14] are pro- furnaces and steam boilers, most of the outgoing energy that posed to control the combustion in boiler systems. Due to exits with hot gas can be reduced, causing a reduction in the the existence of data-monitoring and collection systems in consumed fuel cost with a considerable effect on red ucing most thermal power plants, the fuzzy method is efficient for the emission of pollutants into the environment [2 1]. , In modelling and controlling the boiler combustion system. As general, boiler inputs consist of air and water, and the major these data-collection and monitoring systems are present output of the boiler is the superheated steam. The primary in most thermal power plants, the use of fuzzy methods is boiler control signal (master) should affect the two crucial effective in controlling the boiler system. The conventional variables: air and fuel. A multi-objective model predictive data-mining techniques such as the artificial neural network controller is proposed in [] to optimize boiler combustion. 3 approach are applied to model and optimize combustion ef- First, the utopia point is calculated in the model predictive ficiency [13, 15, 16]. system and then the optimization steps are determined by The main contribution of this study is to propose a finding a solution near the utopia point through the trial-and- knowledge-based approach to model efficiency based error method. The genetic algorithm (GA), particle swarm op- on the excess combustion airflow as an input, which in- timization and artificial intelligence algorithms are applied creases the degree of freedom to optimize the industrial in optimizing industrial boiler combustion [4–6]. The compli- boiler combustion. Moreover, controlling the percentage of cations and intricacy of the computation of these methods excess air for increasing the combustion efficiency and re- lead to weak performance in practice. A multivariable and ducing the emission of pollutant gases based on a fuzzy adaptive fuzzy-controller design is proposed in [ , 7 8] to im- proportional-integral controller in the 320-MW Islam-Abad prove the performance of the system in boiler combustion Power Plant in Isfahan, Iran is assessed. processes, where the combustion mechanism is a complex Knowledge-Based Rules Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 Kermani and Kargar | 231 This article is organized as follows: the effect of excess theoretical air needed for the stoichiometric combustion air on the combustion process and the description of the of the fuel is called the excess air [1]. The excess-air per - proposed approach are presented in Section 1; the fuzzy centage typically varies between 2% and 50% in different modelling is presented in Section 2; the controller is intro- systems and depends on different conditions, like fuel duced in Section 3; the simulations are made in Section 5; type and its components, and the boiler dimensions. If and the article is concluded in Section 5. the percentage of this excess air is not adjusted properly, the combustion process will not be complete, and it would allow a considerable volume of remaining unburned hy- 1 Combustion-process model drocarbons, the emission of carbon monoxide (CO) gas An overview of the boiler structure and the related compo- and the release of nitrogen oxide (NOx) pollutants into nents are illustrated in Fig. 1. the environment. NOx is composed of nitric oxide (NO > The combustion inputs are fuel and air regulated by the 90%) and more poisonous nitrogen dioxide (10% > NO) [1]. control system, and a portion of the inlet air as excess air is In addition to the formation of acid rain, NOx emissions tuned manually by the operator. Because we can never get allow ozone to be formed in the lower layers of the atmos- perfect mixing in an industrial-scale combustion system, phere, which is the primary cause of urban air pollution. there will always be some locations in which there is in- Breathing odourless and colourless CO poisonous gas can sufficient air for complete combustion and more air than be very dangerous and may even lead to death. For com- the theoretical (stoichiometric) air is needed to complete plete combustion and burning of all fuel, the air-to-fuel combustion. Suppose the theoretical air-to-fuel ratio is ratio must be chosen such that sufficient oxygen is avail- selected. Since 78% of the air is nitrogen, there will be ap- able in the combustion chamber. The relation between the proximately four nitrogen atoms per active oxygen atom, excess-air percentages and efficiency is shown in Fig. 2. which act as insulators, preventing better mixing of fuel As observed in Fig. 2, the maximum boiler efficiency and air. The air volume that enters the system beyond the is obtained in a restricted area. If the excess-air volume Electrical power to switchyard Superheater High- Low- Flue gas pressure pressure Steam boundary turbine turbine Drum Exhaust Generator Water Walls Reheater Stem condenser Burner Feedwater Flue pump gas Fuel stack Fresh cooling water Induced Header Air draft fan Air Ash handling SO Precipitator/ heater scrubber fabric filter Air Air Forced Fuel pump Fuel draft fan Fig. 1: Boiler structure Insulated tube downcomers Feedwater Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 CO HC’s Unburned fuel 232 | Clean Energy, 2021, Vol. 5, No. 2 is higher than the required amount, heat is absorbed by that would be released into the environment with other this extra air and exits the chamber instead of being useful combustion products. to the system, leading to an increase in the losses of the In the boilers with sulphur-containing fuels, as the ex- system and a decrease in boiler efficiency. When the excess cess combustion air increases, the sulphur in the fuel is air is less than required, the combustion is incomplete. more likely to react to form SO + SO (SOx), a significant 2 3 A portion of the carbon in the hydrocarbon composition contaminant, which in turn would increase the acid rain. of the fuel will be converted into CO instead of carbon di- By considering the destructive effects of pollutants on oxide (CO ), which is a form of energy loss. The volume of human health and the necessity for obtaining the optimum this loss will be proportional to that of the produced CO efficiency of power plants, determining the exact fuel-to- air ratio is of high essence. For this purpose, a method is developed in this article that, in addition to oper ating the system with maximum efficiency, would reduce SOx con- tamination and production in a simultaneous manner. As combustion in the boiler system is a highly non-linear, multi-input multi-output and time-varying process, it is challenging to obtain the optimal volume of air through the classical methods required for the system model. In this context, considering the possibility of obtaining data from thermal power plants, data-mining techniques will be beneficial. In this article, controlling the percentage of excess air to increase combustion efficiency and reduce the pollutant gas emission from the 320-MW Islam-Abad Power Plant in 012 34 5 Isfahan, Iran is of concern. The excess-air control centre on % oxygen in Flue Gas the control desk of one of the 320-MW units of the Islam- Unburned fuel Excess air, lost Abad power station is shown in Fig. 3. heat up stack The newly designed boiler-system control structure of Fig. 2: The relation between excess-air percentage and efficiency [17] the existing plant is displayed in Fig. 4. First, a fuzzy model Fig. 3: The excess-air control desk and general view of control room Efficiency CO Stoichiometriccombustion Highest efficiency Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 Kermani and Kargar | 233 that applies fuel and air information to the control the experts periodically. The original PI controller would not system is identified; next, the GA optimizes this efficiency; remain accurate forever and the efficiency of the boiler the output of this model causes a reduction in error per - output would decrease after some time. By applying this centage. The error is the difference between the output ef- method, instead of having an operator controlling per - ficiency of the fuzzy model and the actual efficiency of the formance (manual mode), a fuzzy PI controller structure is plant. The obtained model is then applied to design this designed to accomplish the proper boiler efficiency auto- fuzzy controller. matically. If the boiler-output efficiency varies, the fuzzy As illustrated in Fig. 4, the main PI (proportional- PI controller generates a new excess-air signal according integral) controller output signal determines the fuelto the c hanges. Due to the new signal of the air and the and air volumes inside the boiler. One of the essential re- combustion, this new volume of fuel is chosen by the main quirements in the combustion operation is to control the PI controller, where the boiler-output efficiency remains at logical fuel-to-air mixture ratio inside the boiler, where optimum, i.e. saving in fuel consumption. there must always be enough air inside the boiler under The boiler gross capacity is 790 MW and 320 MW of this certain conditions for a specific volume of fuel. The fuel- value (i.e. net capacity) leads to electricity generation. demand control loop is designed to assure that this ratio Because online calculations of boiler-output efficiency are is at steady state. The fuel does not increase until the air difficult, the total efficiency of the steam unit is calculated by: increases. The air does not decrease during the reduc- η = η ∗ η total Boiler Turbo cycle (1) tion of the boiler load until the fuel decreases, indicating where η and η are the boiler and turbo-cycle ef- Boiler Turbo cycle the air has priority in the combustion process. The main ficiency, respectively. The boiler efficiency directly affects PI controller is directly affected by changes in the boiler the total efficiency of the unit. The typical boiler and tur - output (i.e. superheat and reheat steam temperature, bine cycle losses are indicated in Tables 1 and 2. pressure and flow rate, etc.) Fig. 1 and the output has slow The boiler and cycle efficiency are defined as follows: dynamic behaviour. To have better combustion, more air Steam flow rate × (Steam enthalpy − Feed water enthalpy) than required by theoretical calculations is needed. For Boiler Efficiency = × 100 Fuel f iring rate × Grosscaloric value this purpose, a manual mode is provided for the oper - (2) ator to adjust and control 20% of the inlet air in the form of excess air in its experimental sense. The design of an Cycle Efficiency = Q .H − Q .H + Q .H − Q .H − Q .H − Q .H ms ms fw fw rh rh rc rc ss ss rs rs intelligent controller regulating the excess-air volume is (3) presented in this article. The boiler parameters, such as the fuel-heating value, where, P, Q , Q , Q , Q , Q , Q , H , H , H , H , H and H ms fw rh rc ss rs ms fw rh rc ss rs weather conditions and an increase in deposits due to are generator power, main steam flow, feed water flow, hot variation in combustion, indicate that the control coef- reheat flow, cold reheat flow, super heat spray flow, reheat ficients of the main PI controller have to be changed by spray flow, main steam enthalpy, feed water enthalpy, hot Fuel Main efficiency Main Pl Process controller Main Air Air Genetic Excess Algorithm Air + Fuzzy Model Efficiency Manual Adj. Fuzzy PID controller Desired value Fig. 4: Boiler-system control structure Knowledge-Based Rules Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 234 | Clean Energy, 2021, Vol. 5, No. 2 Table 1: Boiler losses [18] Reducing Boiler data Dry gas losses 5.0% data Unburned combustibles radiation and unaccounted 2.0% Losses due to moisture formed by hydrogen 3.7% Data combustion fuzzification Losses due to moisture in fuel 0.2% Losses due to moisture in air 0.1% Total boiler loss 11% Fuzzy rules Table 2: Cycle losses [19] Defuzzification Heat rejected with perfect cycle and theoretical 32.8% working fluid Heat rejected due to imperfections in working fluid 7.7% Losses due to ΔP and ΔT in feed water cycle 2.2% Genetic Evaluation with Losses due to ΔP and ΔT in condensing system 1.6% algorithim actual data Loss due to ΔP in reheater 0.4% Total cycle losses 44.7% Fig. 5: Boiler-modelling structure reheat enthalpy, cold reheat enthalpy, super heat spray en- involve unreliable information that can corrupt the results. thalpy and reheat spray enthalpy, respectively. By adopting the regression method to fit a polynomial to As mentioned, the main PI controller of the unit is re- the acquired data in the prior phase, an ultimate data set sponsible for controlling various combustion variables. should be obtained for the output efficiency due to the During the test described in this work (doing the man- changes in excess air. At this point, obtaining an appro- oeuvre on the excess air), operators tried to keep constant priate database is expected, which would be followed by the conditions of the input energy to the turbo cycle (~η designing the fuzzy rules for system modelling and data Turbo ) from the boiler side (i.e. reheat and superheat tem- classification according to the efficiency of the boiler. At cycle perature, pressure and flow rate). The important and dis- this stage, the data are segmented into predefined classes tinctive point of this research is that the proposed method based on the classification criterion, i.e. the boiler effi- does not change the main control loop. Instead, an add- ciency. The fuzzy data-mining technique is applied to the itional control loop is added to the system to determine pre-processed data and appropriate results are obtained the best excess-air value. thereof. The fuzzy rules are designed to obtain the fuzzy- Thus, variations in excess air lead to changes in the model outcome, i.e. the efficiency. The GA is then applied total efficiency of the steam unit, which is available and to minimize the error rate between the actual data and displayed online on the monitoring system. Changes in the fuzzy data. The GA is a particular algorithm in which the efficiency of the whole unit are considered in all our evolutionary biology techniques like heredity, Darwin’s calculations. chosen principles and biological mutation are involved. This is a programming method in which genetic evolution as a resolving model has been generally applied to obtain 2 Fuzzy modelling the best solutions for optimization and search problems The proposed algorithm is presented in Fig. 5. due to biological operations such as crossover, selection For modelling a boiler, the required data must be col- and mutation. The problem that needs to be solved in- lected first. The boiler air and fuel flow are considered the cludes the inputs transformed into the solutions due to two inputs of a fuzzy model, with the efficiency as the the process modelled on genetic evolution. The responses output. The boiler control system keeps the fuel inlet flow are then considered as volunteers by the fitness function almost constant. Upon changes to the excess-air volume and then the algorithm ends if the output condition of the provided to the operator as a part of the boiler input, the problem is met. In general, this is an iterative algorithm in output efficiency is calculated and recorded by the plant which most of its parts are selected as random processes. monitoring system. Consequently, the excess-air per - The obtained input and output data from the monitoring centage increases from 0.17% to 1.3% at 320-MW output system are saved. The monitoring system of the subject power, which changes the efficiency rate recorded per plant is shown in Fig. 6. second. Because the measured data of efficiency were af- For different volumes of excess air and fuel flows, the fected by noise, the polynomial regression is applied to output total efficiency is recorded over ~1700 seconds and find the best fit for a series of data and reduce the noise shown in Fig. 7, where the efficiency curve is shown with effect. We disregard redundant data that are far from the drastic changes, but in fact the system dynamics are very standard deviation during the calculation. These data may slow. These changes occur due to the measurement noise. Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 Kermani and Kargar | 235 Fig. 6: The monitoring system in the 320-MW Islam-Abad Power Plant, Isfahan, Iran 38.6 Data 5th degree 38.4 38.2 37.8 37.6 37.4 37.2 0 200 400 600 800 1000 1200 1400 1600 Time (sec.) Fig. 7: Estimation of the boiler efficiency: data compared to fifth-degree polynomial fit In the case of noise, curve fitting is the method of forming a curve or mathematical function that has the best fit to a series of data. Polynomial regression is a class of curve fitting MODEL Fuel in which the relationship between the independent variable (Mamdani) x and the dependent variable is modelled as an y n th-degree polynomial in the presence of the sensor noise. Because Efficiency the measured variables used to calculate efficiency are af- Excess air fected by noise, the polynomial regression is applied to find the best fit for a series of data and reduce the noise effect. Fig. 8: Fuzzy structure of the proposed model We disregard redundant data that are far from the standard deviation during the calculation. These data may involve The excess air will be changed from the minimum to the unreliable information that can corrupt the results. The maximum of its authorized range per particular fuel flow. fifth-degree polynomial regression is selected and applied in The monitoring system records all the component data of the basic fitting function to obtain an agreeable waveform. the system during these changes. According to the moni- In practice, experts at the power plant do the same. toring data, the fuzzy method is applied to model the The change in excess air is done manually to evaluate boiler efficiency in terms of fuel flow and excess air. The efficiency. For this purpose, at different intervals of fuel fuzzy structure is shown in Fig. 8. The fuel flow and excess flow, the excess air is changed within the lowest to the air are considered as the fuzzy inputs and efficiency as the highest boundary limit of the unit to maintain stable com- output. The input and output variables can be expressed in bustion, which in turn changes the output efficiency of the the linguistic phrases, tabulated in Table 3. system. According to the operators’ experience, the fuel The membership functions of the boiler fuel flow are intervals include very high, high, medium and low states. defined in correspondence with the experimental data Efficiency (%) Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 236 | Clean Energy, 2021, Vol. 5, No. 2 Table 3: Linguistic phrases of input/output variables B MD G VG Linguistic variables Linguistic terms Excess air Gradient (Gd) Fuel Very High (VH), High (H), Medium (M), Low (L) 0.5 Efficiency (output) Bad (B), Moderate (MD), Good (G), Very Good (VG) 37 37.2 37.4 37.6 37.8 38 38.2 38.4 38.6 38.8 39 L M H VH Efficiency (%) Fig. 11: Boiler-efficiency output membership functions 0.5 Table 4: Fuzzy rules Rule # Diagnostic fuzzy rules R1 If (fuel is L) and (ex-air is Gd) then (eff is G) R2 If (fuel is M) and (ex-air is Gd) then (eff is VG) 82.5 83 83.5 84 84.5 85 85.5 86 86.5 87 R3 If (fuel is H) and (ex-air is Gd) then (eff is MD) R4 If (fuel is VH) and (ex-air is Gd) then (eff is B) Fuel gas flow (kNm /h) Fig. 9: Boiler fuel input data membership functions For evaluating the accuracy of the model, first, a set of different data is given as the test input to the boiler and, next, the efficiency obtained from the model is compared with the consequences acquired from the real operational system. The outcome shown in Fig. 13 indicates that this proposed model provides an accurate response. In this study, the inlet water temperature and environmental 0.5 conditions are considered constant due to their slow dy- namic behaviour with respect to the combustion process. According to the simulated data, the maximum root mean square (RMS) error between the actual (E) and estimated (Ê) 0 0.2 0.4 0.6 0.8 1 1.2 1.4 outputs of the identified fuzzy model is 16%: O2 (%) Ä ä E − E i i (4) Fig. 10: Excess-air boiler input data membership function i=1 RMS = Error In the next step, the fuzzy-modelling accuracy is increased shown in Fig. 9. The fuel is natural gas but the suggested by applying the GA. The cost function is expressed as: method will be the same for other fuels. The membership function is associated with the boiler J = |e|∗ t dt excess air illustrated in Fig. 10 subject to the actual model conditions. Note that %O = (Air demand) × K, 0.9 ≤ K ≤ 1.1, where e is the error between the actual boiler and the where, K is tuned by the operator and it means equivalent fuzzy-model efficiency. The GA minimizes the cost func- to 20% of inlet air. tion to obtain a better precise fuzzy model and it will also Also, the membership functions associated with the optimally select the parameters in the fuzzy PI controller output efficiency in terms of practical data are shown in [20]. The process of using a GA to generate optimal data Fig. 11. is that, first, the interval [0.7 3] is considered as the initial The definition and tuning of the membership functions population of the GA for each of the scale variables. Then, are based on experience and in consultation with power- the fitness of each person in this population is examined. plant specialists. After these functions are specified, the The selected N people are paired to create new parents. In fuzzy rules are obtained according to the data file extracted the GA, the impact factors of the mutation and crossover from the boiler, subject to the membership functions. The are considered to be 0.3 and 0.7, respectively. Then, the pre- fuzzy rules are defined in Table 4. vious generation is replaced by the new generation. The A 3D diagram of the rules is depicted in Fig. 12, where values of the fuzzy-controller scaling factors are obtained the output value of the efficiency is specified. with 100 iterations over an interval [0.7 3] to retrieve the Membership grades Membership grades Membership grades Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 Kermani and Kargar | 237 38.6 38.4 38.2 37.8 37.6 37.4 37.2 0.5 0 (%) 86 0 Fuel gas flow (kNm3/h) Fig. 12: The boiler model fuzzy rules 0 200 400 600 800 1000 1200 1400 1600 0.5 0 200 400 600 800 1000 1200 1400 1600 E actual model E fuzzy model 0 200 400 600 800 1000 1200 1400 1600 Time (s) Fig. 13: Boiler model output response with fuzzy modelling best response. The simulation results (Fig. 14) indicate that design a PI controller, a fuzzy structure is defined for each the maximum RMS error between actual output (E) and es- of the PI controller coefficients. In the fuzzy logic con- timated output (Ê) of the optimized fuzzy model is 7.5%, troller (FLC), the two inputs, named system error (e) (i.e. which is a more accurate response than that of the esti- the difference between expected and fuzzy efficiency) and mated output of the fuzzy model in Fig. 13. change of error (ce), are applied. The optimal efficiency for This modelling is made subject to maximum changes this intended steam unit is considered to be 38.8%. We (i.e. continuous change in the excess airflow to allow dif- have considered different operating modes for the boiler ferent frequencies to become excited and obtain the ac- inputs and, based on the step response in Fig. 15, the ap- curate model). propriate rules for the fuzzy PI controller are designed. The control input (u) is defined as: de u = k e + k e dt + k (6) 3 Proposed controller p i d dt The closed-loop control system of the boiler adjusts the (7) k = k + k p p p main air and fuel flow to control the boiler efficiency. To Efficiency (%) Excess air (O %) Fuel (KNm /h) Efficiency (%) Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 238 | Clean Energy, 2021, Vol. 5, No. 2 0 200 400 600 800 1000 1200 1400 1600 0.5 0 200 400 600 800 1000 1200 1400 1600 E actual model E fuzzy model 0 200 400 600 800 1000 1200 1400 1600 Time (s) Fig. 14: Boiler model output response obtained through the optimized fuzzy modelling Step response 1.8 1.6 1.4 1.2 1 4 0.8 0.6 0.4 0.2 0246 8101214 Time (s) Fig. 15: Different modes for error and error changes in fuzzy PI controller design (8) k = k + k i i i controller structure, it is possible to reduce the system where the constant parameters k and are the pro- p i steady-state error and fluctuations. To illustrate how the portional and integral coefficients of the PI controller, IF-THEN rules are defined, assume that: (i) the system error is high at the beginning of the operation, (ii) the respectively, and the parameters k and k are deter - output has not yet reached the proper volume and (iii) mined by fuzzy logic control for each step time, in their the slope of the error is high, point 1 Fig 15. In this case, adaptive sense. the error sign (E-Eref) is negative and the error-slope sign The membership functions and rules are adjusted to is positive. Therefore, to obtain the proper response, improve the FLC performance. According to the system the value must be increased to a moderate or large state error at any given moment in Fig. 15, the coefficient p positive volume (because the error slope has decreased k can be increased or decreased, which is defined by orientation). the fuzzy rules. By modifying the coefficient k in the PI Efficiency (%) Excess air (O %) Fuel (KNm /h) 2 Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 Kermani and Kargar | 239 Table 5: Fuzzy rule base table for k NB NM NS ZO PS PM PB CE/E NB NM NS ZO PS PM PB NB PB PB PM PM PS ZO ZO NM PB PB PM PS PS ZO NS 0.5 NS PM PM PM PS ZO NS NS ZE PM PM PS ZO NS NM NM PS PS PS ZO NS NS NM NM PM PS ZO NS NM NM NM NB –2 –1.5 –1 –0.5 0 0.5 1 1.5 2 PB ZO ZO NM NM NM NB NB Fig. 16: Structure of fuzzy-controller membership functions for error input e(t) Table 6: Fuzzy rule base table for k ce/e NB NM NS ZO PS PM PB NB NM NS ZO PS PM PB NB NB NB NB NB NS NS ZO NS NB NB NB NS NS ZO PS NS NB NB NS NS ZO PS PS 0.5 ZE NB NS NS ZO PS PS PB PS NS NS ZO PS PB PB PB PS NS ZO PS PS PB PB PB PB ZO PS PS PB PB PB PB –2 –1.5 –1 –0.5 0 0.5 1 1.5 2 Fig. 17: Structure of fuzzy-controller membership functions for input analysis is run in the three modes: manual excess-air con- error changes ce(t) trol, closed-loop with the PI controller and closed-loop with the fuzzy PI controller. The coefficients of the main PI For each one of the fuzzy inputs, seven linguistic vari- controller applied in this simulation are extracted from the ables, namely positive big (PB), positive medium (PM), Ziegler–Nichols method [21]. In the manual control mode, positive small (PS), zero (Z), negative small (NS), negative the excess-air volume is set at 0.5%. The boiler-system medium (NM) and negative big (NB), are of concern. The simulation with a fuzzy PI controller is analysed by con- membership functions are defined in range [–2 + 2] con- sidering the noise in both the excess air and fuel flow as sidering the error of the boiler system, as shown in Figs the two main boiler inputs. 16 and 17. The structure of each membership function is The excess-air adjustment by applying the fuzzy PI considered Gaussian. The proposed fuzzy rules for k and controller compared to the manual adjustment and the PI k are tabulated in Tables 5 and 6, respectively. The first controller is shown in Fig. 18. According to the obtained re- row and column of the matrices indicate the fuzzy sets sults, with changes in fuel levels during system operation, of the error variable (E) and the change in error variable the fuzzy PI controller, compared to the PI controller and (CE), respectively. The fuzzy output variable is presented manual mode, increases the output efficiency by ~0.7%. in the table. The average volume of each efficiency graph is obtained through Equation (9): 3.1 Stability analysis Eff (9) i=1 According to the manufacturer’s available documenta- Eff = tion and the control-system design, the system stability The output obtained efficiency is is guaranteed subject to certain conditions, one being the Eff = 38.05%, Eff = 38.39%, and Eff = 38.73% in excess-air changes of ≤20% of the inlet air, which the oper - Manual PI FuzzyPI manual mode and by using the PI and fuzzy PI controller, ator takes care of manually. The control system is designed respectively (Fig. 18). The same result is shown in Fig. 19, to preserve system stability, despite excess-air changes in where, with a change in the fuel graph, the fuzzy PI con- the boiler of ≤20%. Therefore, this proposed method would troller adjusts the excess air and increases the system ef- contribute to adjusting the operator’s excess-air volume ficiency by ~0.64%. and guarantee system stability. The output efficiency is obtained by Eff = 37.34%, Eff = 37.36%, and Eff = 37.75% in Manual PI FuzzyPI 4 Simulation manual mode and by using the PI and fuzzy PI controller, respectively. This simulation is run according to the proposed excess- In the following, the simulation of the boiler system is air controller and the fact that the main PI controller auto- addressed by considering the noise in both of the main matically controls the fuel. The simulation results with boiler inputs (i.e. the excess air and fuel flow). The white different fuel volumes are evaluated. The boiler-system Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 240 | Clean Energy, 2021, Vol. 5, No. 2 Manual 39 PI PI fuzzy 0 200 400 600 800 1000 1200 1400 1600 84.5 83.5 82.5 0 200 400 600 800 1000 1200 1400 1600 Manual PI 0.5 PI fuzzy 0 200 400 600 800 1000 1200 1400 1600 Time (sec.) Fig. 18: Excess-air adjustment by applying fuzzy PI controller compared to the manual tuning and the PI controller of the power plant Manual PI PI fuzzy 0 200 400 600 800 1000 1200 1400 1600 85.4 85.2 0 200 400 600 800 1000 1200 1400 1600 1.5 Manual PI PI fuzzy 0.5 0 200 400 600 800 1000 1200 1400 1600 Time (sec.) Fig. 19: Excess-air adjustment with PI and PI fuzzy controllers with another graph of fuel noise with a power of 0.1 is added to the boiler fuel input fuel changes. The output efficiency is obtained to be 37.27%, and the simulation is performed by considering the PI and 37.63% and 37.86% in manual mode and by applying the PI fuzzy PI controller. The boiler-output efficiency results are and fuzzy PI controller, respectively. In the next step, the shown in Fig. 20. white noise of 0.1 power is added to the boiler inlet air and As observed, the fluctuations in the boiler fuel actuator the boiler-efficiency results are simulated by applying both affect efficiency. When the PI controller is applied, the ef- the PI and fuzzy PI controller (Fig. 21). ficiency is reduced due to the noise. When the fuzzy PI is As seen in Fig. 21, the fuzzy PI controller accurately applied, the precise adjustment of the excess air increases adjusts the amount of excess air for combustion com- the output efficiency range despite fluctuation with the pared to the manual mode and PI controller, thus causing 3 3 Excess air (O %) Fuel (KNm /h) Excess air (O %) Fuel (KNm /h) 2 2 Efficiency (%) Efficiency (%) Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 Kermani and Kargar | 241 Manual PI PI fuzzy 0 200 400 600 800 1000 1200 1400 1600 85.5 84.5 0 200 400 600 800 1000 1200 1400 1600 1.5 Manual PI PI fuzzy 0.5 0 200 400 600 800 1000 1200 1400 1600 Time (sec.) Fig. 20: Efficiency output response by considering the noise at the fuel inlet Manual PI 38.5 PI fuzzy 37.5 0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600 Manual PI PI fuzzy 0 200 400 600 800 1000 1200 1400 1600 Time (sec.) Fig. 21: Efficiency output response by considering the noise at the fuel and air inlets an increase in the boiler efficiency. The results indicate 5 Conclusion that the efficiency is highly sensitive to the excess-air In this article, the fuzzy data-mining method is adopted volume and its oscillations. The fuzzy PI controller min- to model boiler efficiency based on fuel flow and excess imizes the effects of fluctuations and increases the ef- air. The rule-based fuzzy model is presented by using ficiency by ≤0.6% compared with the PI controller and the practical database of the boiler and the GA is used manual control. The output efficiency is obtained by to minimize the modelling error. The simulation results Eff = 37.47%, Eff = 37.63%, and Eff = 37.85% in Manual PI FuzzyPI indicate that the accuracy of the fuzzy model increased manual mode and by using the PI and fuzzy PI controller, from 84% to 92.5% by applying a GA. To control the ex- respectively. cess air, a controller is designed based on the PI control 3 3 Efficiency (%) Fuel (KNm /h) Excess air (O %) Fuel (KNm /h) Excess air (O %) Efficiency (%) Downloaded from https://academic.oup.com/ce/article/5/2/229/6275218 by DeepDyve user on 18 May 2021 242 | Clean Energy, 2021, Vol. 5, No. 2 [8] Xakimovich SI, Maxamadjonovna UD, Askarxodjaevna BH. structure combined with the fuzzy rules. The fuzzy con- Adaptive system of fuzzy-logical regulation by temperature troller adjusts the air volume according to the input fuel mode of a drum boiler. IIUM Engineering Journal, 2020, 21:182–192. changes to achieve maximum efficiency. The simulation [9] Liu XJ, Kong XB, Hou GL, et al. Modeling of a 1000 MW power results indicate that the efficiency increases compared plant ultra super-critical boiler system using fuzzy-neural to the manual mode and PI controllers by applying this network methods. Energy Conversion and Management, 2013, proposed controller. In addition to the increased ef- 35:518–527. ficiency, another result obtained here is the optimal [10] Sarda P, Hedrick E, Reynolds K, et al. Development of a dy- namic model and control system for load-following studies combustion control in the boiler due to the reduction of supercritical pulverized coal power plants. Processes, 2018, in fuel consumption and pollutant emission into the 6:226. environment. [11] Liu X, Kong X. Nonlinear fuzzy model predictive iterative learning control for drum-type boiler-turbine system. Journal of Process Control, 2013, 23:1023–1040. Conflict of Interest [12] Zhang F, Wu X, Shen J. Extended state observer based fuzzy None declared. model predictive control for ultra-supercritical boiler-turbine unit. Applied Thermal Engineering, 2017, 118:90–100. [13] Kusiak A, Song Z. Combustion efficiency optimization and References virtual testing: a data-mining approach. IEEE Transactions on [1] Kozioł M. The effect of the ratio of excess air and the process Industrial Informatics, 2006, 2:176–184. time on CO and NOx emissions from the combustion of al- [14] Wu J, et al. Robust H(infinity) tracking control of boiler-turbine ternative fuel in a grate furnace. E3S Web of Conferences, 2020, systems. ISA Transactions, 2010;49:369–375. 154:03003. [15] Weiping L, Jinshuai Z. Research and analysis of 300MW [2] Zhao W, Zhao G, Lv M, et al. Fuzzy optimization control for boiler combustion system based on fuzzy control. In: NOx emissions from power plant boilers based on nonlinear Proceedings of the 2018 Chinese Control And Decision Conference optimization. Journal of Intelligent and Fuzzy Systems, 2015, (CCDC), Shenyang, China, 9–11 June 2018, Vol. 1, 5460–5463. 29:2475–2481. [16] Teorey T, Lightstone S, Nadeau T,et al. Database Modeling and [3] Lin B, Shen J, Li Y, et al. Boiler combustion optimization using Design. 5th edn. Burlington, MA: Morgan Kaufman Publishers, multi-objective model predictive control. In: Proceedings—2015 2011. Chinese Automation Congress, CAC 2015, Wuhan, China, 27–29 [17] Industrial Heating. Achieving Efficient Combustion (part 1). November 2015, 2140–2145. https://www.industrialheating.com/ articles/91625-achieving- [4] Shi Y, Zhong W, Chen X, et al. Combustion optimization of efficient-combustion-part-1 (10 February 2021, date last ultra supercritical boiler based on artificial intelligence. accessed). Energy, 2019, 170:804–817. [18] American Society of Mechanical Engineers. Performance [5] Pan H, Zhong W, Wang Z, et al. Optimization of industrial Monitoring Guidelines for Power Plants. Performance Test Codes. boiler combustion control system based on genetic algo- New York, NY: ASME, 2010, 9. rithm. Computers & Electrical Engineering, 2018, 70:987–997. [19] American Society of Mechanical Engineers. Performance [6] Liang H, Zou J, Zuo K, et al. An improved genetic algorithm op- Monitoring Guidelines for Power Plants. Performance Test Codes. timization fuzzy controller applied to the wellhead back pres- New York, NY: ASME, 2010, 10. sure control system. Mechanical Systems and Signal Processing, [20] Chao CT, Sutarna N, Chiou JS, et al. An optimal fuzzy PID 2020, 142:106708. controller design based on conventional PID control and [7] Zhang YF, Xu LW, Chen P, et al. Multivariable fuzzy control nonlinear factors. Applied Sciences, 2019, 9:1224. doi:10.3390/ of CFB boiler combustion system. In: Proceedings of the World app9061224 Congress on Engineering and Computer Science 2013 Vol II, WCECS [21] Bequette BW. Process Control: Modeling, Design, and Simulation. 2013, San Francisco, USA, 23–25 October 2013, 1–6. Upper Saddle River, NJ: Prentice Hall PTR, 2003, 198–202.
Clean Energy – Oxford University Press
Published: Jun 1, 2021
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.