Developing a tool for project contingency estimation in Eskom Distribution Western Cape Operating Unit

[摘要] ENGLISH ABSTRACT:Construction projects are risky by nature, with many variables a ecting their outcome.A contingency cost and duration are allocated to the budget and scheduleof a project to provide for the possible impact of risks.To enable the management of project-related risk on a portfolio level, contingencyestimation must be performed consistently and objectively. The currentproject contingency estimation method used in the capital program managementdepartment of Eskom Distribution Western Cape Operating Unit is not standardised,and is based solely on expert opinion. The aim of the study was todevelop a contingency estimation tool to decrease the inuence of subjectivity oncontingency estimation methods throughout the project lifecycle so as to enableconsistent project risk reection on a portfolio level.From a review of contingency estimation approaches in literature, a hybridmethod combining neural network analysis of systemic risks and expected valueanalysis of project-speci c risks was chosen.Interviews were conducted with project managers (regarding network assetconstruction projects completed in the last twonancial years) to distinguishsystemic and project-speci c risk impact on cost and duration growth. Outputsfrom 22 interviews provided three data patterns for each of 89 projects. After interviewdata processing, 138 training patterns pertaining to 85 projects remainedfor neural network training, validation and testing.Six possible neural network inputs (systemic risk drivers) were selected asproject de nition level, cost, duration, business category, voltage category andjob category. A multilayer feedforward neural network was trained using a supervised training approach combining a multi-objective simulated annealing algorithmwith the standard backpropagation algorithm.Neural network results were evaluated for di erent scenarios considering possiblecombinations of model input variables and number of hidden nodes. Thebest scenario (exclusion of business category input with nine hidden nodes) waschosen based on training and validation errors. Validation error levels are comparableto those of similar studies in the project managementeld. The chosenscenario was shown to outperform multiple linear regression, but calculated R2values were lower than anticipated. It is expected that neural network performancewill further improve as additional training patterns become available.The trained neural network was combined with an expected value analysistool (risk register format) to estimate contingency due to systemic risks alongsidean estimation of contingency due to project-speci c risks. The project-speci cexpected value method was modi ed by basing the contingency estimation on theexpected number of realised risks according to a binomial scenario. A total costdistribution was included in tool outputs by assuming the contingency cost equalto the standard deviation of the cost estimate.To aid business integration of the developed tool, study outputs included thepoints in the project lifecycle model at which the tool should be applied, and theprocess by which tool outputs become inputs to the enterprise risk managementsystem.By following this approach, systemic and project-speci c risks are containedin a single tool providing contingency cost and duration output on project level,while enabling integration with reporting on program, portfolio and enterpriselevel.

[发布机构] Stellenbosch University