Study On Small-Sample Estimation Model Used In Power Engineering Cost

The rapid development of national economy makes the need to power for enterprise increased constantly, and country enlarges the dynamics of investment to power and engineering development. In"11th Five-Year Plan"period, the state grid will invest more than 1 trillion Yuan, so large investment and heavy development mission bring forward higher demands to the power engineering cost. On the one hand, the traditional cost methods of engineer based on budgetary estimate and budget system can not meet the need of engineer gradually, especially in the investment management and approbation of investigate. On another hand, the audit department and investing party accumulated lots of history engineer, the engineering involved many factors, the analogized projects are limited, which make the history data are ineffective so that it is impossible to guide the cost management of new project effectively. Therefore, it is very important for cost estimation of new engineer to establishment practical engineering estimating model using cost data that already get with mathematic algorithm.The paper proposes an easy-operated and effective power engineering cost method of small-sample engineering estimating model based on the present condition and history project of power engineer cost in some region. Firstly, aiming at the characteristic of power cost history data sample engineers are limit and influencing factors are many in the region, the Support Vector Machine which put a good performance in the aspect of small sample learning is presented and theoretical derived. In order to simplify mathematical algorithm and meet engineering estimating demand , an improved algorithm-least squares support vector machine is given and rigorous theoretical derived. Secondly, the new comprehensive and independent indexes are got by pretreated data with principal component analysis to keep the information of original data furthest. Finally, combining the last two methods, the engineering estimating model based on mixed algorithm is proposed. And the best principal component number and result are got by contrasting and analyzing estimation effect of different principal component number with the constraint condition of descending the data error. At the same, because the selection of parameters in the aforesaid model depends on experiential data, so the parameter optimal model based on Genetic Algorithm is suggested in the paper which set up reasonable fitness function, find the best parameter and further increase the stability and estimation precision of the model. Two engineering estimating models are constructed, The configuration of the first model is so simple and operational that it is suitable to quick estimation and censor of engineering cost. And the second improved model can remedy the defects of the first one efficiently and raise estimation precision by optimizing parameters. The ideal effect can be got by simulating the two models and the results express that the method can give effective guide and advices to the power engineering cost in the initial stage of engineering development.