Research on Cost Pre-control Technology for Prefabricated Substation Projects Based on Small Samples

Abstract To accurately predict and control the engineering costs of prefabricated substations under small-sample conditions, in order to accurately forecast and control the cost level of substation project, a combination forecasting model of support vector regression (Referred to as:SVR) based on gray relation analysis (Referred to as:GRA) and particle swarm optimization (Referred to as:PSO) is proposed. The main indexes of project cost, which are input into SVR model optimized by PSO to forecast the substation project cost, are extracted through GRA to avoid the problem of large forecasting error caused by artificial selection of input variables. In addition, SVR model optimized by PSO is used for substation project cost forecasting, which effectively overcomes the blindness of model parameter setting when using a single SVR model to forecast project cost. Through the GRA-PSO-SVR combined optimization model used to predict the cost of the substation project ,we can obtain more accurate investment estimation. A coefficient adjustment method is applied to the predicted values of the cost model for prefabricated substations. This approach ensures the model predictions closely align with actual costs, thereby achieving precise pre-control of initial investment estimates for prefabricated substation projects under construction.