Research On Optimization Method Of Petrochemical Multistage Centrifugal Pump Based On Proxy Model

The hydrogenation feed pump is a key equipment in the hydrocracking unit of the refining process,which pressurizes and transports flammable and explosive media such as diesel,gasoline,and residual oil to the heating furnace.It is one of the most central multistage centrifugal pumps in the petrochemical field.Multistage centrifugal pumps have higher pressure and the complex flow structure inside the rotor leads to more energy loss,greatly limiting the efficiency of the pump;At the same time,more complex vortex structures can generate hydraulic excitation,causing unstable operation of multi-stage pumps.In order to improve the hydraulic characteristics of the hydrotreating feed pump,this paper aims at the head and efficiency,and takes the first stage double suction impeller and the second stage single suction impeller as the specific optimization objects.From two aspects of pump hydraulic performance prediction and optimization,the method based on PSO-LSSVR surrogate model is used to study the hydrotreating feed pump.The main work is as follows:(1)On the basis of establishing a three-dimensional model of the hydrogenation feed pump,the convergence analysis of the grid was carried out using the GCI criterion.The SST k-ωturbulence model was applied to analyze the multistage centrifugal pump at six operating points and was verified experimentally.The numerical simulation errors of head and efficiency were 4.86%and 4.91%,respectively,under the design conditions.(2)The optimization of the multi-stage pump proxy model based on the LSSVR algorithm and the determination of the kernel function were studied.The particle swarm optimization(PSO)algorithm was used to optimize the penalty parameter C and kernel function parameterσ~2 in the LSSVR proxy model,and the convergence problem of PSO was studied.The results showed that the Gaussian radial basis function kernel was suitable for the optimization of the multi-stage pump in this paper,and the linearly decreasing inertial weight showed obvious advantages in the convergence of PSO.(3)A performance prediction method for the hydrogenation feed pump based on Latin hypercube sampling method was proposed.Three performance indicators,maximum absolute relative error(MARE),root mean square error(RMSE),and relative root mean square error(RRMSE),were used to evaluate the accuracy of the surrogate model,and R~2was used to evaluate the fitting performance.The surrogate model predicted the sample values and analyzed the priority of different variables on optimization.The results showed that the three performance indicators met the accuracy requirements,and the R~2 analysis showed that the fitting performance of the surrogate model was good.The priority of the influence of different variables on optimization was analyzed,and it was found that the blade outlet width was the most significant variable affecting the head of the multi-stage pump,while the blade outlet angle was the most significant variable affecting the efficiency of the multi-stage pump.(4)The energy loss of multistage pump is analyzed based on entropy generation theory by comparing the changes of internal flow field before and after optimization with the total pressure loss coefficient,Euler head,speed and turbulent kinetic energy.The results show that the head and efficiency are increased by 3.99%and 2.91%respectively after optimization;after optimization,the impeller has better pressurization capacity,the low speed area and high speed area of internal flow are reduced,and the internal flow field is significantly improved.The analysis of energy loss shows that the energy loss of the multi-stage centrifugal pump with improved impeller structure is significantly reduced,mainly because the horseshoe vortex at the leading edge of the blade and the trailing edge vortex are restrained.After optimization,the amplitude of pressure pulsation decreases and the pressure distribution becomes more uniform.