Dam Deformation Prediction Model Based On LSSVM Optimized By Improved Artificial Bee Colony Algorithm

With the development of surveying technologies,the dam deformation monitoring data is more and more complicated.So how to deal with the dam deformation data and establish an accurate deformation prediction model is very important for the safe operation of dams,Which can provide scientific decision-making basis for the use of dams.Aiming at the disadvantages that the traditional deformation prediction models have a high requirements for the original monitoring data,and it is difficult to describe the quantitative relationship,This paper proposed a least squares support vector machine(LSSVM)model that optimized by improved artificial bee colony algorithm(IABC),and applied it to the dam deformation prediction.First of all,proposed three improvement measures for the shortcomings of the ABC algorithm.On the one hand,ABC algorithm generates the initial population in an uncertain way,so this paper proposed a dual-population initialization strategy based on forward-backward learning,and gave the formula of exchanging the current optimal solution,which enriched the diversity of the population and made the initial population distribute in the solution space more evenly and reasonably;On the other hand,the ABC algorithm searches the solution with random direction and step size,and it is easily falls into local extremum.this paper,Therefore,added the current optimal solution to the search formula,and designed an adaptive weight function to guide the direction and speed for the search solution.What is more,the ABC algorithm eliminates all the bad solutions according to the greedy rule,which limits the exploration ability.Therefore,Metroplis criterion was introduced to replace the greedy rule and an adaptive cooling function was designed to control the probability for accepting the bad solution,which balanced the exploration and development ability of the algorithm.Combined the improvement measures with the original ABC algorithm to get the IABC algorithm,and the optimization performance of IABC algorithm was tested by four benchmark functions.The experimental results showed that compared with the ABC algorithm,the optimal value and the average value soved by IABC algorithm,which were closer to the actual value,and the solution standard deviation has several times lower.It is proved that the IABC algorithm is practical and advanced.Next,established a LSSVM dam deformation prediction model based on IABC algorithm.The minimum value of LSSVM prediction error is the final target value of IABCalgorithm.Therefore,constructed a LSSVM prediction error formula with kernel parameters and detailed parameters as independent variables,and which was regarded as the objective function formula of IABC algorithm.Then solved the optimal parameter combination by IABC algorithm,and substituted it into LSSVM to construct the IABC-LSSVM dam deformation prediction model.The last but not the least,the actual application.Took the measured cumulative settlement of the dam of Guandi Hydropower Station as an example,the IABC-LSSVM model program was written based on Matlab,and the experimental results show that the convergence accuracy of IABC algorithm is significantly higher than that of ABC algorithm in the parameter optimization process of LSSVM,which verified the effectiveness of the improvement algorithm again.In addition,compared with the results of the other four prediction models in the estimation of the average absolute error,the average absolute percentage error,and the root mean square error,the ABC-LSSVM models were 0.396 mm,10.56%,and 0.487mm;The GS-LSSVM models were 0.313 mm,7.75%,and 0.433mm;The GA-LSSVM models are 0.463 mm,13.46%,and 0.614mm;The PSO-LSSVM models are0.410 mm,12.47%,and 0.473mm;while the IABC-LSSVM models are 0.189 mm,4.82%,0.256 mm,all the error indicators are the smallest,which showed high precision,strong stability characteristics,and provided reference value for the safety monitoring of dam.