Hardware transient faults are proven to have a significant impact on deep neural networks (DNNs), whose safety-critical misclassification (SCM) in autonomous vehicles, healthcare, and space applications is increased u...Hardware transient faults are proven to have a significant impact on deep neural networks (DNNs), whose safety-critical misclassification (SCM) in autonomous vehicles, healthcare, and space applications is increased up to four times. However, the inaccuracy evaluation using accurate fault injection is time-consuming and requires several hours and even a couple of days on a complete simulation platform. To accelerate the evaluation of hardware transient faults on DNNs, we design a unified and end-to-end automatic methodology, A-Mean, using the silent data corruption (SDC) rate of basic operations (such as convolution, addition, multiply, ReLU, and max-pooling) and a static two-level mean calculation mechanism to rapidly compute the overall SDC rate, for estimating the general classification metric accuracy and application-specific metric SCM. More importantly, a max-policy is used to determine the SDC boundary of non-sequential structures in DNNs. Then, the worst-case scheme is used to further calculate the enlarged SCM and halved accuracy under transient faults, via merging the static results of SDC with the original data from one-time dynamic fault-free execution. Furthermore, all of the steps mentioned above have been implemented automatically, so that this easy-to-use automatic tool can be employed for prompt evaluation of transient faults on diverse DNNs. Meanwhile, a novel metric “fault sensitivity” is defined to characterize the variation of transient fault-induced higher SCM and lower accuracy. The comparative results with a state-of-the-art fault injection method TensorFI+ on five DNN models and four datasets show that our proposed estimation method A-Mean achieves up to 922.80 times speedup, with just 4.20% SCM loss and 0.77% accuracy loss on average. The artifact of A-Mean is publicly available at https://github.com/breatrice321/A-Mean.展开更多
To enhance the fault transient performance of aerospace multiphase permanent magnet synchronous motor(PMSM)system,an adaptive robust speed control is proposed regardless of the phase open-circuit(OC)and short-circuit(...To enhance the fault transient performance of aerospace multiphase permanent magnet synchronous motor(PMSM)system,an adaptive robust speed control is proposed regardless of the phase open-circuit(OC)and short-circuit(SC)fault in this paper,which can be applied for both the redundant motor system and fault tolerant motor system.For aerospace multiphase PMSM system,besides external load disturbance and system parameter perturbation,there inevitably exists the electromagnetic torque ripple in fault transient process,which can degrade the system performance and even cause the system instability.To cope with this issue,the electromagnet torque ripple of the multiphase PMSM system in fault transient process is first analyzed.Then,by considering the electromagnet torque fluctuation caused by fault transient as a system uncertainty,a novel adaptive robust speed control scheme is proposed,while the adaptive law is constructed to emulate the total system uncertainty bound,which include the load disturbance,the parameter variation,and the electromagnetic torque fluctuation due to fault transient.The resulting control can ensure the speed control performance even in fault transient process regardless of the uncertainty,in which no prior estimation of the uncertainty bound is required.In addition,the proposed adaptive robust speed control is demonstrated by a six-phase PMSM experimental platform.The novelty of this research is to explore a novel adaptive robust speed control to strengthen the fault tolerance performance of multiphase PMSM system even in fault transient process,which requires no prior estimation of the uncertainty bound.展开更多
Slack-Decode Simultaneously and Redundantly Threaded （SD-SRT） is proposed for detecting transient faults in processors. SD-SRT boosts the previously proposed SRT performance via definitely eliminating redundant inst...Slack-Decode Simultaneously and Redundantly Threaded （SD-SRT） is proposed for detecting transient faults in processors. SD-SRT boosts the previously proposed SRT performance via definitely eliminating redundant instructiou fetches. First, the fetch stage is moved out of the Spheres of Replication （SoR）, and a unified instruction-fetch-queue （IFQ） is exploited by both the leading and trailing threads. Second, a scheme called slack-decode cooperates with the unified IFQ to harmonize proceeding of the two threads. The simulations show that SD-SRT outperforms original SRT in terms of IPC by 15%, and decreases I-cache access by 42%. Meanwhile, SD-SRT leads to a lessened size and complexity for hardware structures such as load-value-queue and store-buffer.展开更多
The amount of electrical energy produced by wind mills is constantly increasing.Nowadays detailed analyzes considering the impact of wind energy integration on the transmission system are required.The goal of this stu...The amount of electrical energy produced by wind mills is constantly increasing.Nowadays detailed analyzes considering the impact of wind energy integration on the transmission system are required.The goal of this study is to investigate the dynamic response of a wind turbine with doubly fed induction generator connected to the power system during grid disturbance.The current and future wind power situation is modeled as two cases and a transient fault is simulated.In order to analyze the impact of wind energy integration in electrical power grid,a power system model has been developed,integrated with wind turbine using doubly fed induction generator and transient analysis are performed.Here,an attempt has been made to compare the impact,in terms of voltages,currents,total harmonic distortion,etc.,of adding wind turbines into electrical power grid.展开更多
Since the fault dynamic of droop-controlled inverter is different from synchronous generators (SGs), protection devices may become invalid, and the fault overcurrent may damage power electronic devices and threaten th...Since the fault dynamic of droop-controlled inverter is different from synchronous generators (SGs), protection devices may become invalid, and the fault overcurrent may damage power electronic devices and threaten the safety of the microgrid. Therefore, it is imperative to conduct a comprehensive fault analysis of the inverter to guide the design of protection schemes. However, due to the complexity of droop control strategy, existing literatures have simplified asymmetric fault analysis of droop-controlled inverters to varying degrees. Therefore, accurate fault analysis of a droop-controlled inverter is needed. In this paper, by analyzing the control system, an accurate fault model is established. Based on this, a calculation method for instantaneous asymmetrical fault current is proposed. In addition, the current components and current characteristics are analyzed. It was determined that fault currents are affected by control loops, fault types, fault distance and nonlinear limiters. In particular, the influences of limiters on the fault model, fault current calculation and fault current characteristics were analyzed. Through detailed analysis, it was found that dynamics of the control loop cannot be ignored, the fault type and fault distance determine fault current level, and part of the limiters will totally change the fault current trend. Finally, calculation and experimental results verify the correctness of the proposed method.展开更多
This series of papers report on relay protection strategies that satisfy the demands of a strong smart grid.These strategies include ultra-high-speed transient-based fault discrimination,new co-ordination principles o...This series of papers report on relay protection strategies that satisfy the demands of a strong smart grid.These strategies include ultra-high-speed transient-based fault discrimination,new co-ordination principles of main and back-up protection to suit the diversification of the power network,optimal co-ordination between relay protection and auto-reclosure to enhance robustness of the power network.There are also new development in protection early warning and tripping functions of protection based on wide area information.In this paper the principles,algorithms and techniques of single-ended,transient-based and ultra-high-speed protection for EHV transmission lines,buses,DC transmission lines and faulty line selection for non-solid earthed networks are presented.Tests show that the methods presented can determine fault characteristics with ultra-high-speed(5 ms)and that the new principles of fault discrimination can satisfy the demand of EHV systems within a smart grid.展开更多
As the structures of multiple branch lines(MBLs)will be widely applied in the future flexible DC distribution network,there is a urgent need for improving system reliability by tackling the frequent non-permanent pole...As the structures of multiple branch lines(MBLs)will be widely applied in the future flexible DC distribution network,there is a urgent need for improving system reliability by tackling the frequent non-permanent pole-to-pole(P-P)fault on distribution lines.A novel fault restoration strategy based on local information is proposed to solve this issue.The strategy firstly splits a double-ended power supply network into two single-ended power supply networks through the timing difference characteristics of a hybrid direct current circuit breaker(HDCCB)entering the recloser.Then,a method based on the characteristic of the transient energy of fault current is proposed to screen the faulty branch line in each single-ended power supply network.Also,a four-terminal flexible DC distribution network with MBLs is constructed on PSCAD to demonstrate the efficacy of the proposed strategy.Various factors such as noise,fault location,and DC arc equivalent resistance are considered in the simulation model for testing.Test results prove that the proposed strategy for fault restoration is effective,and features high performance and scalability.展开更多
基金Project supported by the Shanghai Pujiang Talent Program,China(No.21PJD026)。
文摘Hardware transient faults are proven to have a significant impact on deep neural networks (DNNs), whose safety-critical misclassification (SCM) in autonomous vehicles, healthcare, and space applications is increased up to four times. However, the inaccuracy evaluation using accurate fault injection is time-consuming and requires several hours and even a couple of days on a complete simulation platform. To accelerate the evaluation of hardware transient faults on DNNs, we design a unified and end-to-end automatic methodology, A-Mean, using the silent data corruption (SDC) rate of basic operations (such as convolution, addition, multiply, ReLU, and max-pooling) and a static two-level mean calculation mechanism to rapidly compute the overall SDC rate, for estimating the general classification metric accuracy and application-specific metric SCM. More importantly, a max-policy is used to determine the SDC boundary of non-sequential structures in DNNs. Then, the worst-case scheme is used to further calculate the enlarged SCM and halved accuracy under transient faults, via merging the static results of SDC with the original data from one-time dynamic fault-free execution. Furthermore, all of the steps mentioned above have been implemented automatically, so that this easy-to-use automatic tool can be employed for prompt evaluation of transient faults on diverse DNNs. Meanwhile, a novel metric “fault sensitivity” is defined to characterize the variation of transient fault-induced higher SCM and lower accuracy. The comparative results with a state-of-the-art fault injection method TensorFI+ on five DNN models and four datasets show that our proposed estimation method A-Mean achieves up to 922.80 times speedup, with just 4.20% SCM loss and 0.77% accuracy loss on average. The artifact of A-Mean is publicly available at https://github.com/breatrice321/A-Mean.
基金This work was supported by National Natural Science Foundation of China(Grant No.51707004)the Fundamental Research Funds for the Central Universities(Grant No.YWF20BJJ522)National Defense Science and Technology Foundation Enhancement Program,and Major Program of the National Natural Science Foundation of China(Grant No.51890882).
文摘To enhance the fault transient performance of aerospace multiphase permanent magnet synchronous motor(PMSM)system,an adaptive robust speed control is proposed regardless of the phase open-circuit(OC)and short-circuit(SC)fault in this paper,which can be applied for both the redundant motor system and fault tolerant motor system.For aerospace multiphase PMSM system,besides external load disturbance and system parameter perturbation,there inevitably exists the electromagnetic torque ripple in fault transient process,which can degrade the system performance and even cause the system instability.To cope with this issue,the electromagnet torque ripple of the multiphase PMSM system in fault transient process is first analyzed.Then,by considering the electromagnet torque fluctuation caused by fault transient as a system uncertainty,a novel adaptive robust speed control scheme is proposed,while the adaptive law is constructed to emulate the total system uncertainty bound,which include the load disturbance,the parameter variation,and the electromagnetic torque fluctuation due to fault transient.The resulting control can ensure the speed control performance even in fault transient process regardless of the uncertainty,in which no prior estimation of the uncertainty bound is required.In addition,the proposed adaptive robust speed control is demonstrated by a six-phase PMSM experimental platform.The novelty of this research is to explore a novel adaptive robust speed control to strengthen the fault tolerance performance of multiphase PMSM system even in fault transient process,which requires no prior estimation of the uncertainty bound.
文摘Slack-Decode Simultaneously and Redundantly Threaded （SD-SRT） is proposed for detecting transient faults in processors. SD-SRT boosts the previously proposed SRT performance via definitely eliminating redundant instructiou fetches. First, the fetch stage is moved out of the Spheres of Replication （SoR）, and a unified instruction-fetch-queue （IFQ） is exploited by both the leading and trailing threads. Second, a scheme called slack-decode cooperates with the unified IFQ to harmonize proceeding of the two threads. The simulations show that SD-SRT outperforms original SRT in terms of IPC by 15%, and decreases I-cache access by 42%. Meanwhile, SD-SRT leads to a lessened size and complexity for hardware structures such as load-value-queue and store-buffer.
文摘The amount of electrical energy produced by wind mills is constantly increasing.Nowadays detailed analyzes considering the impact of wind energy integration on the transmission system are required.The goal of this study is to investigate the dynamic response of a wind turbine with doubly fed induction generator connected to the power system during grid disturbance.The current and future wind power situation is modeled as two cases and a transient fault is simulated.In order to analyze the impact of wind energy integration in electrical power grid,a power system model has been developed,integrated with wind turbine using doubly fed induction generator and transient analysis are performed.Here,an attempt has been made to compare the impact,in terms of voltages,currents,total harmonic distortion,etc.,of adding wind turbines into electrical power grid.
基金supported by National Natural Science Foundation of China under Grant 51977066。
文摘Since the fault dynamic of droop-controlled inverter is different from synchronous generators (SGs), protection devices may become invalid, and the fault overcurrent may damage power electronic devices and threaten the safety of the microgrid. Therefore, it is imperative to conduct a comprehensive fault analysis of the inverter to guide the design of protection schemes. However, due to the complexity of droop control strategy, existing literatures have simplified asymmetric fault analysis of droop-controlled inverters to varying degrees. Therefore, accurate fault analysis of a droop-controlled inverter is needed. In this paper, by analyzing the control system, an accurate fault model is established. Based on this, a calculation method for instantaneous asymmetrical fault current is proposed. In addition, the current components and current characteristics are analyzed. It was determined that fault currents are affected by control loops, fault types, fault distance and nonlinear limiters. In particular, the influences of limiters on the fault model, fault current calculation and fault current characteristics were analyzed. Through detailed analysis, it was found that dynamics of the control loop cannot be ignored, the fault type and fault distance determine fault current level, and part of the limiters will totally change the fault current trend. Finally, calculation and experimental results verify the correctness of the proposed method.
文摘This series of papers report on relay protection strategies that satisfy the demands of a strong smart grid.These strategies include ultra-high-speed transient-based fault discrimination,new co-ordination principles of main and back-up protection to suit the diversification of the power network,optimal co-ordination between relay protection and auto-reclosure to enhance robustness of the power network.There are also new development in protection early warning and tripping functions of protection based on wide area information.In this paper the principles,algorithms and techniques of single-ended,transient-based and ultra-high-speed protection for EHV transmission lines,buses,DC transmission lines and faulty line selection for non-solid earthed networks are presented.Tests show that the methods presented can determine fault characteristics with ultra-high-speed(5 ms)and that the new principles of fault discrimination can satisfy the demand of EHV systems within a smart grid.
基金supported by the National Natural Science Foundation of China(No.51877174)。
文摘As the structures of multiple branch lines(MBLs)will be widely applied in the future flexible DC distribution network,there is a urgent need for improving system reliability by tackling the frequent non-permanent pole-to-pole(P-P)fault on distribution lines.A novel fault restoration strategy based on local information is proposed to solve this issue.The strategy firstly splits a double-ended power supply network into two single-ended power supply networks through the timing difference characteristics of a hybrid direct current circuit breaker(HDCCB)entering the recloser.Then,a method based on the characteristic of the transient energy of fault current is proposed to screen the faulty branch line in each single-ended power supply network.Also,a four-terminal flexible DC distribution network with MBLs is constructed on PSCAD to demonstrate the efficacy of the proposed strategy.Various factors such as noise,fault location,and DC arc equivalent resistance are considered in the simulation model for testing.Test results prove that the proposed strategy for fault restoration is effective,and features high performance and scalability.
