Performance optimization of engineering systems with particular reference to dry-cooled power plants

[摘要] Computer simulation programs were developed for the analysis of dry-cooling systems for powerplant applications. Both forced draft direct condensing aircooled condensers and hyperbolic naturaldraft indirect dry-cooling towers are considered.The results of a considerable amount of theoretical and experimental work are taken into account tomodel all the physical phenomena ofthese systems, to formu1ate the problems in formal mathematicalterms and to design and apply suitable computational algorithms to solve these problems effectivelyand reliably.The dry-cooling systems are characterized by equation-based models. These equations aresimultaneously solved by a specially designed constrained nonlinear least squares algorithm todetermine the performance characteristics of the dry-cooling systems under fixed prescnoedoperating conditions, or under varying operating conditions when coupled to a turbo-generator set.The solution procedure is very fast and effective.A capital and operating cost estimation procedure, based on information obtained from dry-coolingsystem component manufacturers and the literature, is proposed. Analytical functions express theannual cost in terms ofthe various geometrical and operating parameters ofthe dry-cooling systems.The simu1ation and the cost estimation procedures were coupled to a constrained nonlinearprogramming code which enable the design of minimum cost dry-cooling systems at fixed prescribedoperating conditions, or dry-cooling systems which minimize the ratio of total annual cost to theannual net power output of the corresponding turbo-generator set. Since prevailing atmosphericconditions, especially the ambient temperature, influence the performance of dry-cooling systems,wide fluctuations in turbine back pressure occur. Therefore, in the latter case the optimal design isbased on the annual mean hourly frequency ofambient temperatures, rather than a fixed value.The equation-based models and the optimization problems are simultaneously solved along aninfeasible path (infeasible path integrated approach). The optimization model takes intoconsideration all the parameters that may affect the capital and operating cost of the dry-coolingsystems and does not prescribe any limits, other than those absolutely essential due to practicallimitations and to simulate the systems effectively. The influence that changes ofthe constraint limitsand some problem parameters have on the optinmm solution, are evaluated (sensitivity analysis). The Sequential Quadratic Programming (SQP) method is used as the basis in implementing nonlinearoptimization techniques to solve the cost minimirnti~n problems. A stable dual active set algorithmfor convex quadratic programming (QP) problems is implemented that makes use of the specialfeatures ofthe QP subproblems associated with the SQP methods. TIrls QP algorithm is also used aspart of the algorithm that solves the constrained nonlinear least squares problem This particularimplementation of the SQP method proved to be very reliable and efficient when applied to theoptimization problems based on the infeasible path integrated approach.However, as the nonlinear optimization problems become large, storage requirements for the Hessianmatrix and computational expense of solving large quadratic programming (QP) subproblemsbecome prohibitive. To overcome these difficulties, a reduced Hessian SQP decomposition strategywith coordinate bases was implemented. This method exploits the low dimensionality of thesubspace of independent decision variables. The performance of this SQP decomposition is furtherimproved by exploiting the mathematical structure of the engineering model, for example the blockdiagonal structure ofthe Jacobian matrix. Reductions ofbetween 50-90% in the total CPU time areobtained compared to conventional SQP optimization methods. However, more function andgradient evaluations are used by this decomposition strategy.The computer programs were extensively tested on various optimization problems and provide fastand effective means to determine practical trends in the manufacturing and construction of costoptimaldry-cooling systems, as well as their optimal performance and operating conditions in powerplant applications.The dissertation shows that, through the proper application of powerful optimization strategies andcareful tailoring of the well constructed optimization model, direct optimization of complex modelsdoes not need to be time consuming and difficult.Reconnnendations for further research are made.

[发布机构] Stellenbosch University