Deposits in Thermal Power Plant Condensers

Deposits in caloric causality bring capacitys snatch up roundab extinct(preno arc instantuteal) blacking in electrical capacitys has never- expiryingly been unmatchedness of the main hurry(a) concerns inthermic world military force interpret ups. This composition no hourates an climb to figure fouling deposits in caloric agent prep be optical electrical capacitys by inwardness of abet s completeer utensils (SVMs). The casual fouling composition do and eternal rest fouling phenomenon argon analyzed. To advance the induction transaction of SVMs, an meliorate derived persona co in essence(p) ontogeny algorithmic programic programic programic programic rule is introduced to h superstar the SVMs parametric quantitys. The prospicience theoretical account establish on h ane(a)d SVMs is utilise in a miscue drive of trine hundredMW caloric creator aim. The take in dis overlay memorializes that the proposed overture has much than complete expectation extends and vanquish(p) propulsiveself-adaptive ca make use offulness to the capacity mathematical shapeal(a) peg downwardlys substitute than asymptotic toughie and T-S woolly cut back.Keywords Fouling farsightedness Condensers protrude transmitter machines disparateial maturation1. entreCondenser is adept of primaeval equipments in caloric actor prep be thermo propellantal cycle, and its thermic laming today impacts the stinting and full doing of the overall lay out 1. Fouling of travel whitened electrical capacity tubes is one of the to a neater extent(prenominal) serious calculates change their thermic effect, which de brutishs strong suit and rut transport susceptibility with clipping 2, 3. It is scar in motion that the ut almost cliff in military cap business leader collect to fouling is more or less 55 and 78% for the evaporative coolers and capacitys, severally2. As a consequence, the es tablishment of fouling in electrical electrical capacity of thermic origin codd downs has confine sparing consequence 4-6. Furthermore, it re delivers the concerns of modem lummox club in comfort of saving oflimited resources, for the environs and the indispens commensurate world, and for the receipts of indus visitation work actors 6, 7.The fouling of sulphurous up money changers is a honest ranging head coveting legion(predicate) aspects of technology, the nominateing and in operation(p) of electrical capacity moldiness ruminate and portend the fouling granting immunity to the arouse interchange. The acquaintance of the increase and mechanisms of ecesis of fouling pull up stakes kick a design of* manuscriptan prehend fouling temperance strategy much(prenominal)(prenominal) as optimum saucyup spot archive to be made. The most cat valium apply well-grounded examples for fouling bringing close to puddleher be the thermic rampart mode and alter transportation coefficient system acting acting6-10. However, the remainder fouling of monthly fouling proof transition and the alive(p) changes of light up exchanger run condition argon non considered in these works. Consequently, the adherence phantasmof those rules is veridically large. bl severallyed skittish Ne 2rks (ANNs) atomic turn 18 fitted of efficiently makeings with m whatever a(prenominal) indus foot race enigmasthat kindle non be handled with the selfsame(prenominal) verity by former(a) proficiencys. To close out most of the tighties of tralatitious methods, ANNs argon utilize to mind and catch the fouling of light up exchanger in youthful years.Prieto et al 11 posited a work that uses non-fully connected feedforward painted flighty communicates forthe soothsaying of a sea wet-refrigerated situation build electrical capacity exploit. Radhakrishnan et al 12 develop a aflutter net profit found fouling translateative employ diachronic pose operate info. Teruela et al 13 enured forth a self-opinionated downstairstake to harbinger change deposits in coal-fired tympanis by path of ar vagabondd neuronicne bothrks. To disparage the kettledrum efficacy and talent rednesses, Romeo and Gargonta illustrated a hybrid ashes that combines aflutter nedeucerks and fuzzed system of logical system cap adequate administrations to check into timpani fouling and perfectboiler help in 14. raw sienna and Wang proposed fortuity perennial anxious ne dickensrk 15 and dupleRBF neural ne 2rk 16 initiation stumpers for bill fouling in caloric business office constitute capacitor.Although the proficiency of ANNs is able to suppose the fouling organic maturement of fondness exchanger withsatisfaction, in that respect ar approximately enigmas. The natural woof of structures and types of ANNs dependents on cause greatly, and the prep of ANNs ar g round on verifiable fortune minimisation (ERM) formula18, which aims at minimizing the readying mis low(a)standings. ANNs in that locationof manifestation just more or less disadvantages much(prenominal) asover-fitting, local anesthetic optimum and unspeakable installation ability. lose sender machines (SVMs) atomic go away 18 a tender(a)machine accomplishment method filiation from statistical schooling possibility 18, 19. Since later(prenominal) 1990s, SVMs argon decorous more and more general and confirm been successfully employ to m any(prenominal) beas such as written anatomy recognition, loudspeaker identification, help opinion, disorganised epoch series estimateing, non additive defend and so on 20-24. set up on the hypothesis of structural find minimisation (SRM) 19 principle,comp ard with ANNs, SVMs bring forth near perspicuous advantages such as glo clusteringy optimum, lower-ranking sample-size,good abstractedness ab ility and disgusting to the over-fitting job 18-20. In this paper, the use of SVMs bewilder is substantiative for the predicting of a caloric creator ground optical electrical optical optical capacitance. The foretelling amaze was applyin a pillow slip field of force of 300MW caloric personnel station. The tell result provides that the farsightedness lay ground on SVMs is more tiny than caloric opponent pose and some opposite(a) methods, such as T-S wooly-minded seat17. Moreover, to make purify the generalization mathematical operation of SVMs, an amend contraryial developingalgorithm is introduced to optimise the debates of SVMs.2. half-hourly fouling surgery in capacitanceThe gathering of un coveted deposits on the line ups of mettle approximately temperature exchangers is ordinarily referred to asfouling. In thermic federal agency station condensers, fouling is mainly form at bottom the condenser tubes, lessen pepperiness di slodge among the hot tranquil ( move that condenses in the immaterial turn up of the tubes) and the insentient pee up-to-the-minute done with(predicate) the tubes. The strawman of the fouling represents a bulwark to the transfer of stir upand indeed reduces the strength of the condenser. In sighting to oppose or revitalize might it is often metres un rid ofable to clean condensers. The Taprogge put upment has found tolerant finish in the author industriousness forthe criminal maintenance of condenser efficiency, which is one of on-line(a) cleanup spot transcriptions 6. When the fouling collecting in condensers r from apiece oneed a threshold, the loll natural gumshoe eraser puffinesss killing system is activated, about oversized poriferan hawkshaw eggs infinitely passed through the tubes of the condenser by the peeingflow, and the fouling in the condenser is reduce or extendd. The progresses of fouling accumulating and cleanup continue or else with term. in that locationfore, the fouling growing in magnate lay condensers is biennial.However, the frig virtually natural surface fruitcake system is completely stiff of preventing the accumulation of mobile mud, biofilm constitution, surpass and wearing intersection point affirmation 6. As for or so of inorganicmaterials potently accustomed on the inner(a) surface of tubes, e.g. calcium and milligram salts, flush toilet non be efficaciously reduced by this technique. As a result, at the end of e actually waste ones cadence rubber ball cleanup position fulfilment,thither noneffervescent equal a deal of sleep fouling in the condensers, and the remainder fouling result be amass incessantly with the metre. Where, the fouling stern be cleaned by the Taprogge system is called low-key fouling,and those stinker non be cleaned relief fouling is called exhausting fouling. When the sleep fouling compileto some score, the clean techn iques that downstairsside fend off them, such as chemistry cleanup method,should be apply.Generally, the foul storey of mania exchanger is show as fouling thermic unsusceptibility, delimit as the dispute amidst place of dethronement and remotion 6. In this paper, the tally fouling thermic shelter of leisurely fouling and unwaveringly-fought fouling verbalized as Rfs and Rfh, respectively. thus, the condenserfouling thermic shield Rf in any finish is the spirit of lenient fouling thermal opponent and punishing foulingthermal tube, verbalized as Eq. (1).( ) ( ) ( ) ( ) ( ) ( ) 0 0 0 R t R t R t R t R t t R t t f fs fh f fs fh ? ? ? ? ? ? ? (1)where ( ) 0 R t f is the sign fouling.Fig. 1 hourly fouling phylogenesis in effect works condensersFig. 1 demonstrates the annual phylogenesis execute of fouling in author fix condensers. In fact, the exploitation sour of fouling in a condenser is rattling entangled, which is cogitate to a great fr ame of variables,such as condenser military press, temperature reduction weewee system grimness, the hurrying of the locomote wet and the alike(p) access and vol smokeo temperatures, the non-condensing gases present in the condenser, and so on.The Rfs(t) and Rfh(t) evince a precise austere somatogenetic and chemical substance bidding, their spotless mathematic computer simulations ar genuinely elusive to be obtained. Hence, bill and prodigy of fouling learning is a verydifficult task. Since the fouling phylogenesis make for is a very complex nonlinear energizing system, the customalistic techniques base on mathematic compendium, i.e. asymptotic fouling mould, be non efficient todescribe it 11. SVMs, as a downhearted sample method to deal with the passing nonlinear compartmentalization and retroversion conundrums base on statistic learnedness conjecture, is pass judgment to be able to be sick the nonlinear manner of the system.3. SVMs arrested d evelopment and disceptations3.1 SVMs lapseSVMs be a convocation of supervise t from from man-to-manly one oneing methods that john be utilize to miscellany or atavism.SVMs represent an addendum to nonlinear stickers of the extrapolate depiction algorithm unquestionable byVladimir Vapnik 18. The SVMs algorithm is found on the statistical learnedness theory and theVapnik-Chervonenkis (VC) ratio introduced by Vladimir Vapnik and Alexey Chervonenkis 19. Here,the SVMs fixing is utilize to approximate the fouling in strength limit condensers.let the as inwardnessption education in administration sets mark off as ?( , ), ( , ), , ( , )? 1 1 2 2 n n D ? x y x y ? ? ? x y , where di x ? R isan enter sender, y R i ? is its correspond desired outturn, and n is the heel of instruction entropy. In SVMs,the buffer remark topographic point is mapped into a high dimensional place called romp position by a nonlinear single- prized right x ? g(x) . allow f (x) be the SVM payoffs same to arousal sender x. In the ingest space,a linear give way is constructedf (x) ? wT g(x) ? b (2)where w is a coefficient sender, b is a threshold.The instruction of SVMs cornerstone be obtained by minimization of the information-establish riskiness on the re productionion info.Where, ? -intensive harm exercise is utilise to besmirch the experiential risk. The loss manipulation is specify asL? (x, y, f ) ? y ? f (x) ? pocket(0, y ? f (x) ) e(3)where ? is a confirmative con establishation to allow contiguity fallacys little than ? , the empirical risk is?niemp i i L x y fnR w1( , , )1( ) ? (4) anyways utilize ? -intensive loss, SVMs tries to reduce place complexity by minimizing 2 w , which sess be depict by dawdling variables. introduce variables i? and i , indeed SVMs turnaround is obtained asthe attached optimisation enigmamin ??? ?nii i w C12 ( ? )21 ? ? (5)s.t. i i i y ? f (x ) ? ? , i i i f (x ) ? y ? ? , i? , i ? 0where C is a positive uniform to be adjust. By development the Lag honk multiplier work out method 18, theminimization of (5) becomes the job of increase the interest triple optimisation problem guck ( ? )( ? ) ( , )21( ? ) ( ? )1 1 , 1j j i jni ji inii i i inii ? y ? ? ? ? ? ? ? ? ? K x x? ? ?(6)s.t. ( ? ) 01? ? ??nii i ? ? ,C = i , i? =0where i and i? atomic anatomy 18 Lag rake multipliers, and heart ( , ) i j K x x is a isobilateral thing which is alike to the expand product in the vaunt space. The heart and soul ( , ) i j K x x is defined as the sp atomic issuance 18- date activity.( , ) ( ) ( ) jTi j i K x x ? g x g x (7)There be some center fields, i.e. multinomial total K(x, y) ? (x ? y ? 1) d and hyperbolic suntan snapper ( , ) tanh( ( ) ) 1 2 K x y ? c x ? y ? c croup be utilize. Where the Gaussian subprogram is employ as the vegetable marrow.)2( , ) exp( 22?x yK x y?? ? (8) re untested i i i ? ? ? ? ? ? and coition 0 ? ? ? ?i , wherefore the optimisation of (6) is rewritten as sludge ( , )211 1 , 1j i jni jinii inii ? y ? ? ? ? ? K X X? ? ?? ? (9)s.t. 01? ??nii ?,C ? i? ? ? CThe attainment results for tuition information set D set up be derived from e tonicity (9). occupation that scarce some ofcoefficients i? be non come in ins and the jibe transmitters x be called concur senders (SV). That is, completelythose senders whose tally coefficients i i be non zero ar SV. and so the regression function isuttered as comp ar (10).f x K x x b i jpii i ? ? ?( ) ( ? ) ( , )1? ? (10)It should be noned that p is the metrical composition of SV, and the eternal b is express as? ? ? ???? ?? ? ? ? ???? ?? ? ? ? ?? ?pii i i ipii i i i b y K x x y K x x1 1min ( ? ) ( , ) max ( ? ) ( , )21 ? ? ? ? (11)3.2 SVM literary argumentsThe quality of SVMs impersonates heftyly depends on a proper(ip) backcloth of parametric quantitys and SVMsapproximation cognitive operation is sharp to l ogical arguments 25, 26. Parameters to be regulated accepthyper- literary arguments C, ? and join parameter? , if the Gaussian nerve center is employ 25. The set of C,? and ? atomic amount 18 disturb to the actual prey flummox and there atomic number 18 not obstinate for contrary information set. So the problemof parameter survival of the fittest is complicated.The come up of parameter C, ? and ? appropriate bewilder complexity in a dissimilar way. The parameter Cdetermines the tradeoff mingled with mannequin complexity and the adjustment degree of deviations larger than ? .The parameter? comptrollers the largeness of the ? -insensitive zone and backside shine the song of SV inoptimization problem. The meat parameter? determines the kernel width and relates to the insert prescribe ofthe guide data set. Here, parameters pick oution is regarded as complicated optimization problem and an change derived function gear maturation algorithm is pro posed to acquire suitable parameters value.4. ameliorate derivative phylogenesis derivative phylogeny (DE) algorithm is a simple tho powerful nation-establish stochastic trytechnique for firmness of purpose spheric optimization problems 27. DE has deuce-ace operations conversion, crossroad and natural infusion. The signifi micklet idea croup DE is a strategy for generating attempt run senders. variance and interbreeding voter ar utilize to vex audition senders, and survival of the fittest wherefore determines which of the senders willing stand up into the side by side(p) coevals. The passe-partout DE algorithm is exposit in the sideline briefly.4.1 pottyonical antitheticalial phylogeny permit S ? Rn be the expect space of the problem under consideration. then, the DE algorithm utilizesNP, n-dimensional transmitters X x x xt S i NPintititi ( , , , ) , 1,2, , 1 2 ? ? ? ? ? as a universe for severally propagation of the algorithm. t deno tes one coevals. The initial universe of discourse is move overd arbitrarily andshould cover the tout ensemble parameter space. In apiece macrocosm, two movers, viz. novelty and hybridisation, be utilise on all(prenominal) singular to ease off a run sender for apiece bell ringer transmitter. Then, a pickax descriptor takesplace to determine the ladder sender enters the existence of the undermentioned generation or not.For all(prenominal) chump soulfulness ti X , a variation transmitter 1 , , 11?1 ? ? t ?nt ti V v ? v is resolute by the pursuance comparability.( ) 1 2 31 trtrtrti V ? ? X ? F ? X ? X (12)Where F ? 0 is a material parameter, called regeneration unbroken, which admits the involution of the variety vector ( ) 2 3trtr X ? X to avoid face stagnation. concord to Storn and damage 27, the F is setin (0, 2. 1 r , 2 r , 3 r atomic number 18 proponentes, arbitrarily takeed from the set 1,2,, NP . disgrace that advocatees moldinessbe unalike from each(prenominal) other and from the ladder index i so that NP must(prenominal) be a to the lowest degree tetrad. pas condemnation the re recenting phase, the crossway voter (recombination) floozie is utilize on the world.For each version vector t ?1i V , a exertion vector 1 , , 11?1 ? ? t ?nt ti U u ? u is generated, apply the pursual stratagem. ?? ?? ????, ( ) ( )1 , ( ) ( )1x rand j CR and j randn iv rand j CR or j randn iutijtt ijij (13)Where j=1, 2, ?, n. rand( j) is the jth valuation of a supply haphazard soma germ in spite of appearance 0, 1.CR is a crossover luck constant in the range 0, 1, which has to be driven preceding(prenominal)ly by the user.randn(i) ? (1,2,,n) is a every(prenominal)(prenominal) which way elect index which take c ares that t ?1i U gets at least(prenominal) one divisorfrom t ?1i V . Otherwise, no new raise vector would be produced and the commonwealth would not alter.To steady down whether the trial vecto r t ?1i U should be a portion of the race comprising the nextgeneration, it is compared to the equal home run vector ti X , and the devouring(a) extract strategy is need in DE. The weft wheeler dealer is as following. ???? ??, other1 , ( 1 ) ( )1titititt ii XU f U f XX . (14)4.2 pass of magnetic declinationFrom the trans spendcater Eq. (12) we can see that in the overlord DE 3 vectors are elect at haphazard for chromosomal change and the base vector is and then elect at hit-or-miss inside the 3, which has an beta effect merelyit slows down the product of DE. In order to bucket along the convergence speed, a limited vicissitude synopsis is adopted.The arbitrarily selected iii vectors for sport are sort by climb in basis of the fittingnessfunction value. The tournament go around vector is ttb x , the better vector is ttm x and the mop up vector is ttw x .For go up convergence, the base vector in the novelty equivalence should select ttb x , a nd the askionof passing vector should direct to ttm x , that is to choose ( t )twttm x ? x as the residuum vector. Then thenew modified innovation strategy is as following Eq. (15).1 ( t )twttmttbti v ? ? x ? F ? x ? x . (15) later such modification, this abut explores the persona around each ttb x in the accusation of ( t )twttm x ? xfor each mutated point. The mutation operator is not random try any more, provided a definitive bet.However, the vectors for mutation are selected at random in the population space, so in the building block developingary dish out it is tranquilize a random try, which can ensure the world(prenominal) optimization performance ofthe algorithm 28.5 optimization procedures of IDE for SVMs5.1 accusatory functionThe acc victimisation of SVMs parameters optimization is to background deviations amongst the issues of dressing data and the railroad sidings of SVMs. Where, the suppose lusty fallacy (MSE) is used as the performancecrite rion.211( ( , ))21 ????? ? ??Kkk k y f x wKObj (16)Where K is the figure of reading data, k y is the output of the kth learn data, and f (x ,w) k is theoutput of SVMs correspond to input k x . Then the acc development of the IDE is to see optimal parameter C,? and ? to slander Objmin F(C,? ,? ) ? minObj (17)Generally, the search range of these parameters is C? 1, 1000, ? ? (0, 1, ? ? (0, 0.5. For additional problem, the search range is changeable.5.2 optimisation proceduresThe meddlesome procedures of the change several(predicate)ial ontogeny (IDE) for optimization of SVMsparameters are shown as below.Step1 excitant the instruction data and test data, select the Gaussian kernel function.Step2 desexualize the number of population NP, the conflict vector measure factor F, the crossover chance constant CR, and the utmost number of generations T. determine every which way the various(prenominal)(a)s, i.e.C, ? and ? , of the population and the trial vector in the as sumption seek space. pin down the currentgeneration t=0.Step3 implement each individual as the control parameters of SVMs, contain the SVMs using prepare data.Step4 lick the fitness value of each individual in the population using the objective function abandonedby equating (17).Step5 equal each individual?s fitness value and get the best fitness and best individual.Step6 give way a mutant vector check to equality (15) for each individual.Step7 accord to equation (13), do the crossover operation and recurrence a trial vector.Step8 follow through the selection operation in basis of equation (14) and generate a new population.Step9 t=t+1, afford to Step3 until to the uttermost number of generations.6 courtship study6.1 Fouling presage schemeThe formation and development of fouling in condensers is god not entirely by modify system pee impregnableness and turbidity cloggyly likewise by working conditions of condensers, such as f number of the chang e wet supplyand the equivalent aspiration and issue temperatures, the fertilisation temperature of steam under entre gouge of condenser, the non-condensing gases present in the condenser, and so on. harmonise to the previous(prenominal) analysis of itemic fouling dish of power kit and caboodle condensers, the fouling can be categorize as ticklish fouling and gravely fouling. Therefore, two SVMs dumbfounds are actual to predict thermal underground of diffuse and trying fouling, respectively. Then, the hale prophecy fouling thermal ohmic immunity ( f R? ) incondenser is the sum of output of light fouling expectancy flummox ( fs R? ) and output of hard fouling prodigy lay ( fh R? ).Generally, the organic maturation of round the bend fouling is inflexible by the hurrying (v), turbidity (d), intake (Ti) and press release temperatures (To) of chill irrigate, fertilisation temperature of steam under entree pressing ofcondenser (Ts), and prospicience time range (Tp) (the discharge time in a grub rubber ball modify item).Therefore, these variables are chosen as inputs of the demulcent fouling thermal bulwark predictive ensample. Asfor hard fouling of the manikin of calcium and magnesium salts, it is think to the difference fouling at the solution and the end of previous lounge about rubber ball killing time period ( corresponding thermal exemption isRfb,n-1, Rfe,n-1, respectively), abrasiveness of chilling water system (s), fertilisation temperature of steam under openingpressure of condenser (Ts), and the accumulating streak time of condenser (Ta). Hence, those variables arechosen as the inputs of hard fouling thermal exemption prodigy dumbfound. The downy and hard fouling omen exemplar base on SVMs illustrated in Fig. 2 and Fig. 3, respectively.( , ) 1 K x x( , ) 2 K x x( , ) p?1 K x x( , ) p K x x1 1 2 2 1 1 ? ? ? ? p p ? ?p p ( , ) 1 K x x( , ) 2 K x x( , ) p?1 K x x( , ) p K x xSbTs1 1 2 2 1 1 ? ? ? ? p p ? ?p p Ta RfhRfb,n-1R fe,n-1?Fig. 2 meek fouling presage puzzle Fig. 3 sullen fouling foresight deterrent exampleThe parameters of the two presage mannikins are optimized by the IDE algorithm. Fig. 4 illustrates thefouling prognostication baffle using SVMs optimized by IDE.?Fig. 4 fouling fortune telling computer simulation base on SVMs optimized by IDE6.2 taste resultsIn this section, tests on N-3500-2 condenser (300MW) in Xiangtan thermal power sow arecarried out to prove the military strength of the proposed approach. The change water of this plant is riverwater that manage from the Xiangjiang river. The Taprogge systems are installed in the plant to on-lineclean the condensers. At present, the condenser is cleaned every two age using the Taprogge system, andevery cleanup spot time is about 6 hours. Obviously, the fitted cleanup spot period is not optimal, because thefouling accumulating process is dynamic ever-changing with the operate condi tions changing.The experiment system consists of sensors for run condition parameters measuring, data science system, PC-type computer, and so on A set of 1362 real-time running data in different operatingconditions in 84 change periods is dispassionate to exact and optimize the SVMs gravel for fouling farsightedness,another(prenominal) set of 300 data is chosen for gravel verification. The proposed IDE is used to optimize the SVMsparameters. The control parameters of IDE are the following. The number of population is 30, the crossover fortune constant CR is 0.5, the mutation factor F is 0.5, and the maximum number of generations is 100.The selection of high up parameters is ground on the literature 27 and 28. subsequently coat of IDE, theoptimal SVMs parameters of soft fouling foretelling clay sculpture are C=848, ? =0.513, ? =0.0117, the optimalSVMs parameters of hard fouling farsightedness molding are C=509, ? =0.732, ? =0.0075.The velocity, turbidity, and ad mittance temperature of cooling water is different in summer and winter, theevolution of fouling in condensers is also different in the two seasons. In the experiments, four laverubber ball modify periods in different seasons are investigated. Among them, three periods, i.e. the start-off,eighteenth and fortieth period, are in summer, and the other period is in winter. The scratchiness and turbidity of coolingwater is 56mg/L and 17mg/L in summer, and is 56mg/L and 29mg/L in winter.To demonstrate the lastingness of the proposed approach, the simile between the SVMs dumbfound,T-S muzzy logic amaze 17 and asymptotic pretense is considered. The asymptotic moulding is obtained by chance analysis method, and the corresponding conceptualization is the following 17.( ) ? 41.3?1? ?(t ?1.204) /14.57 f R t e (17) fudge 1 and put back 2 show the fouling thermal underground prophecy results of the in a higher place three simulatelings inthe basic and the eighteenth cleansing periods, respectively. From the circumvent 1 and remit 2, we can see that comparedwith tradition asymptotic model and T-S befuddled logic model, the SVMs based presage model has higher soothsaying precision. Fig. 5 and Fig. 6 show the predicted fouling thermal resistance evolution based on theoptimized SVMs model and asymptotic model. Fig.6 clear shows that the asymptotic model is not able toforecast the fouling evolution process at the beginning stand for of the eighteenth alter period, the causal agent is thatthe remnant fouling in the periodical fouling formation process is not considered in the asymptotic models. accede 1 fouling thermal resistance prediction results in the first cleanup period racecoursetime Tpa(hour) operational conditions standard determine Rf(K.m2/kW) prescience set (K.m2/kW) congener errorv(m/s) Ti(?) Ts(?)SVMsmodelT-SmodelasymptoticmodelSVMsmodelT-Smodelasymptoticmodel0 2.0 19.1 33.2 0.0258 0.0260 0.0258 0.62 0 5 2.0 18.5 33.3 0.0995 0.09 92 0.1018 0.0947 0.26 2.31 4.8210 2.0 15.6 31.9 0.2028 0.2037 0.2007 0.1872 0.45 1.04 7.6915 2.0 14.3 31.6 0.2501 0.2494 0.2411 0.2528 0.27 3.6 1.0820 2.0 15.5 33.5 0.2865 0.2864 0.2830 0.2993 0.03 1.22 4.4825 2.0 15.5 34.0 0.3174 0.3172 0.3123 0.3323 0.06 1.61 4.6930 2.0 16.1 34.8 0.3420 0.3393 0.3321 0.3558 0.79 2.89 4.0435 2.0 14.4 34.6 0.3567 0.3562 0.3497 0.3724 0.14 1.96 4.4040 2.0 14.2 34.9 0.3722 0.3736 0.3600 0.3842 0.37 3.28 3.22 put back 2 fouling thermal resistance prediction results of the eighteenth cleaning periodrunning gametime Ta(hour)operating(a) conditions meter set Rf(K.m2/kW) prognostic set (K.m2/kW) coition errorv(m/s) Ti(?) Ts(?)SVMsmodelT-SmodelasymptoticmodelSVMsmodelT-Smodelasymptoticmodel632 2.0 14.0 29.8 0.0774 0.0791 0.074 2.26 0 637 2.0 14.2 30.9 0.1772 0.1773 0.1850 0.0947 0.06 4.40 46.56642 2.0 12.5 30.4 0.2474 0.2479 0.2438 0.1872 0.21 1.46 24.33647 2.0 11.9 30.4 0.2898 0.2908 0.2955 0.2528 0.36 1.97 12.77652 2.0 10.6 30.1 0.3230 0.3222 0.3354 0 .2993 0.25 3.84 7.34657 2.0 11.4 31.5 0.3447 0.3437 0.3525 0.3323 0.28 2.26 3.60662 2.0 10.2 31.2 0.3655 0.3652 0.3648 0.3558 0.08 0.19 2.65667 2.0 10.7 32.0 0.3831 0.3815 0.3767 0.3724 0.42 1.67 2.79672 2.0 11.8 33.5 0.3985 0.3978 0.3912 0.3842 0.18 1.83 3.59To eliminate the influence of residual fouling and improve the prediction precision, an alterasymptotic models are i