Study On The Mixed Model Of Construction Cost Estimation Of Power Transmission Project

Rapid development of electricity demand and power grid construction with economy developing fast as well as increasingly diversified electrical equipment and construction technologies promote higher requirements on cost management of power grid projects. The long power transmission construction cycle and high investment cost with the various factors of natural environment, the price levels and the maintenance requirements influencing the total investment cause the traditional estimation index valuation method to be too general to meet the increasingly precise cost management requirements. Therefore, it is necessary to build fast and accurate cost estimation model based on existing engineering data recently with the reasonable mathematical modeling methods, together with the present situations of electric power engineering construction, cost and management.This paper puts forward the new simple, rapid and effective cost estimation model for power transmission projects with the rigorous scientific theory support to effectively solve the cost management problems in the decision-making stage, from the reality conditions, combining with the characteristics of transmission engineering sample and cost, based on historical data of engineering cost. First of all, analyze the basic theories of project cost and the cost estimation characteristics of the transmission project, explore the cost rules and the data characteristics, make comparison of machine learning algorithm commonly used on the basis of the limited sample quantity and many factors affecting the cost, select gaussian process regression algorithm as the foundation of modeling to effectively deal with the small samples and high dimension data model parameter as the foundation of modeling. Secondly, put forward the data pretreatment process based on principal component analysis through combining with the cost characteristics of power transmission projects so as to provide the more clean, accurate and neat data for the model above on the premise of reducing original data information loss as far as possible. Then, fruit flies optimization algorithm is proposed for improving parameter initial value in view of super parameter initial value greatly influencing parameter automatic optimization results in the gaussian process regression model. Build a new hybrid model by integrating with the researches above, adopt the contrast method of simulation results for optimize the covariance function and sample input dimension in gaussian process to reduce the errors of model, and carry out simulation verification by the 110 kV project as the sample of electric power supply company. The simulation results show that the estimation of new model is higher accurate and the effect is more stable to meet the requirements of electricity cost management, comparing to the traditional models, and has good practical feasibility and extension through combining closely with engineering management practice.