System Level Sequential Subspace Design Optimization of Switched Reluctance Motor Drive Systems Based on Space Reduction Strategy

Abstract

In this paper, a novel system level sequential design optimization method is presented for switched reluctance motor (SRM) drive systems. First, the multiobjective optimization problem for the SRM drive systems is defined. Then, all parameters of the drive systems, including the motor level and control level are divided into two subspaces according to their influences on the objectives. Finally, the optimization of each subspace is performed sequentially until a convergence criterion is met. The sensitivity analysis, the approximate models, and the genetic algorithm are employed for the implementation. Besides, the space reduction method is introduced to reduce the computational cost. The step size of each control factor will be halved in the next iterative process. To verify the effectiveness of the proposed method, An SRM drive system with a 12/10 configuration and angle position method, is investigated.I. IntroductionDue to the absence of any windings and magnets in the rotor, the switched reluctance motors (SRMs) provides the best alternatives for the other machines under harsh environments of extremely high temperatures and pressures [1]. The trade-offs to the overall consideration and the implementation of multiobjective optimization are necessary to meet the requirements for different applications. Previous works are mostly on component level rather than the system level [2]. Theoretically, assembling individually optimized components into a system cannot guarantee the optimal system performance as each part is coupled to each other. Thus, the perfect cooperation of motor and control levels should be optimized simultaneously. The main problem in the system level optimization is the computation burden due to the abundant optimization variables. In this paper, the space reduction strategy will be integrated into the system level optimization flowchart.II. System-Level Sequential Subspace Design OptimizationAn SRM drive system with a 12/10 configuration and angle position method, will be investigated, as shown in Fig. 1.The optimization flowchart is presented in Fig. 2, and it can be divided into the following steps.Step 1: Define the system level optimization model, including objective function, constraints, and design parameters.The optimization problem can be defined asmin: f(xs)s.t. gi(xs)≤0, i=1,2,…,n (1)xsl≤xs≤xsuwhere xs, f, gi are the design parameter vector, objective and constraints of motor, respectively, xs consists of motor parameter vector and control parameter vector, xsl and xsu are the lower boundary and upper boundary, respectively.Step 2: Divide all the parameters in motor and control aspects into two subspaces. X1 and X2 represent the significant and non-significant subspaces, respectively.Step 3: Optimize X1. Generate samples and simulate their response by using the orthogonal array (OA) [1]. An OA with a higher level with 5 levels can be assigned to X1. The Kriging model [2] is developed based on the data. Then, conduct the optimization by using non-dominated sorting genetic algorithm (NSGA) II. The optimal solution is selected for the next subspace.Step 4: Optimize X2. In this step, an OA with 3 levels is assigned to X2. The optimal solution is selected by the same implementation in step 3.Step 5: Termination judgement. Compute the motor performance with the obtained optimal design and compare it with the last objective. If the relative error is less than a given value, terminate the optimization process and output the optimal design. Otherwise, go to the next step.Step 6: Reduce the design space of the design parameters in X1 and X2, respectively. Assume the initial design space of an optimization variable is [a, b], and the optimal value of this variable is x0.For X1 in which there are 5 levels for each design parameter with a step size 2d, the next design space and new levels are[a, a+d, a+2d, a+3d, a+4d ], x0–2d<a ;[b-4d, b-3d, b-2d, b-d, b ], x0+d>b ; (2)[x0-2d, x0-d, x0, x0+d, x0+2d ], othersFor X2 in which there are 3 levels for each design parameter with a step size 2d, the next design space and new levels are[a, a+d, a+2d ], x0-d<a ;[b-2d, b-d, b ], x0+d>b ; (3)[x0-d, x0, x0+d ], othersIt can be found that the step size is halved during the space reduction. **