Investigation Of Real-Time Simulation Numerical Optimization Algorithms For Power Electronic Digital Design

The global energy crisis and environmental pollution are boosting the pressure on the production and consumption of energy.The power electronic devices are considered to be the key to high-efficiency energy conversion and environmental protection,which leads to a surge in the device demand.The scaled power electronic devices are sensitive to the cost,time and efficiency of research and development,which in turn puts forward new requirements for the research and development pattern.With the advantages of good flexibility,high efficiency and low cost,digital design is expected to be an effective way to deal with the scaled power electronic devices and follow the trend of power electronic design automation(PEDA).As the foundation and underlying core technology of PEDA,the power electronic simulation therefore is receiving increasing attention from both the academic and industrial community.Unfortunately,the key technologies of power electronic simulation are mastered by foreign countries.Meanwhile,due to the uncertainty of international situation and the periodicity of domestic policies,the localization of power electronic simulation technology is imminent.Moreover,the trends of high switching-frequency power devices application,multi-domain load objects and power electronic design automation have led to the great challenges in power electronic simulation.According to the clues from digital platform,to mathematical model and finally to numerical algorithm,the comprehensive and in-depth research is conducted from the numerical stability analysis,computation speed optimization and numerical stiff solution in electrical drive system,which includes the major parts of technical problems in power electronic simulation.As a result,a series of numerical optimization algorithms are proposed in this dissertation.Firstly,the numerical stability criterion and optimization method are proposed.Due to the typical features of topology switching in power electronic system,the discontinuities in numerical conversion are easy to accumulate numerical errors,which is sensitive for the numerical stability.Moreover,with the application of high switching-frequency power devices,the introduced discontinuous events and discontinuous functions bring the great challenges for the numerical stability evaluation.For this issue,the factors affecting the numerical stability are systematically analyzed in this dissertation.Then,the phase-trajectory region criterion is proposed to obtain the numerical stability risks based on the dominant influence factors,and therefore the limitations of discontinuous event handling and low efficiency in traditional criteria can be basically covered.Besides,an optimization method based on first-order numerical integration is proposed to quantitatively modify the stability region and improve the numerical stability without reducing the numerical precision,which can provide better guidance for the numerical integration method design.Secondly,the parallel acceleration with numerical model decoupling based on computation front is proposed.The simulation step size of power electronic simulation is highly correlated with the switching frequency of power devices.With the application of high switching-frequency devices,the short time limit and real-time constraints lead to the great challenge to computation speed.With the trend of parallel architecture in digital platform,parallel computing is considered as an effective way to optimize the computation speed with short-time constraint.However,the state-of-the-art numerical models and numerical algorithms are based on serial computation,which are lack of redundant parallelism.For this problem,the optimized method based on computation front is proposed to construct the redundant parallelism for numerical integration.Then,the influences on numerical precision and numerical stability is analyzed,and the extensibility for high-order and multi-step integration methods are discussed respectively.Moreover,the parallel numerical model is designed and applied to the electrical drive system based on redundant parallelism constructed by the proposed method.As a result,the model-level numerical acceleration can be achieved.Thirdly,the variable-order quantized state stiff algorithm based on half-cycle phase shift pre-positioning method is proposed.The numerical stiff problem is the contradiction between numerical stability and computation speed,which is the great challenge in power electronic simulation with the application of high switching-frequency power devices.The solution efficiency of stiff system can be improved by variable step size control based on local truncation error(LTE)in traditional numerical stiff algorithm.However,the discontinuous switching behaviors in power electronic system increase the computation cost of variable step size process,which is limited in the application of real-time simulation.For this issue,the quantized state algorithm is introduced in this dissertation to solve the numerical stiffness in power electronic simulation.The real-time synchronization and switching events are pre-located by the half-cycle phase shift process,and therefore the extra switching cycle delay introduced by the quantized state algorithm in hardware in the loop(HIL)real-time simulation can be avoided.As a result,the application of quantized state algorithm can be expanded in real-time simulation.On this basis,the variable-order quantization state stiff algorithm is proposed to overcome the limitation of stiff solution caused by the discontinuities in power electronic switches.As a result,the computation amount of numerical simulation in electric drive system is reduced,which improves the efficiency of numerical stiff algorithm in real-time simulation.Finally,the aforementioned methods are integrated into the composite numerical optimization algorithms for power electronic real-time simulation.Meanwhile,the development and implementation process from digital platform,mathematical model and numerical algorithm in real-time environment is discussed in depth.As a result,the coordinated optimization scheme is embedded into the real-time simulation prototype of electrical drive system with typical power electronic behaviors,which is built to verify and demonstrate the validity of proposed algorithms.