Hybrid power system health management capability verification and evaluation method

Abstract

The invention discloses a hybrid power system health management capability verification and evaluation method, and belongs to the field of health management system verification and system evaluation. The platform can collect data from different sensors and information systems in real time, such as operation state, temperature, vibration and the like, and timely process and analyze the data. The real-time performance is helpful for quickly identifying fault symptoms, timely taking preventive measures, and reducing the risk of fault occurrence and the maintenance cost. The simulation test not only can reduce the development and verification cost, but also can avoid the damage to the equipment, and improves the safety of the whole test. Verification can be carried out in the system design stage, the equipment health state can be continuously monitored in the production, operation and maintenance stages, and fault prediction and diagnosis can be carried out. Through combination of historical data and real-time data, the platform can provide effective support for full-life-cycle management of equipment, prolong the service life of the system and optimize operation and maintenance decisions.

Description

Technical Field

[0001] The present invention belongs to the field of verification and system evaluation of health management systems. Specifically, by building a universal health management hardware verification system and a health management software verification system, and using a standard mathematical measurement indicator system to strictly evaluate the prediction and health management performance of the health management system, quantitative performance verification and evaluation of the health management system is achieved. Background Art

[0002] The complexity, informatization and integration of health management systems are constantly increasing, which puts forward higher requirements for system reliability, safety and economy. At present, engines mostly adopt preventive maintenance methods. Faced with the needs of the new generation of engines to "enhance the diagnostic fault warning and diagnosis capabilities, improve the engine health management and auxiliary decision-making capabilities", the existing technology can no longer adapt to future development, and it is urgent to carry out research on fault warning and health management technology.

[0003] A hybrid electric vehicle (HEV) is a vehicle power system that combines a traditional internal combustion engine (usually a gasoline engine) and an electric motor. When the vehicle is started or driving at low speed, it is usually driven by an electric motor, which can save fuel and reduce exhaust emissions. The electric motor is powered by the on-board battery. When driving at high speed or accelerating suddenly, the internal combustion engine and the electric motor work together. The internal combustion engine can provide greater power, while the electric motor can make up for the power lag of the internal combustion engine during acceleration and improve the overall acceleration performance. When decelerating or braking, hybrid electric vehicles are usually equipped with a regenerative braking system. The system converts the kinetic energy generated during braking into electrical energy and stores it in the battery. This not only improves energy efficiency, but also extends the service life of the braking system.

[0004] Verification and system evaluation of health management systems is an important step in the development and deployment of health management. Once a health management system is put into operation, it will cost a lot of manpower, material and financial resources to modify it. There has been no systematic and targeted research and technical application of health management system verification and evaluation technology, which has hindered the further development of health management systems. Therefore, it is necessary to carry out research on general health management system verification and evaluation technology to provide support for health management system verification.

[0005] The present invention proposes a hybrid power system health management capability verification and evaluation method, which supports access to real-time simulation models and health management systems, can provide typical fault data sets and a semi-physical simulation experimental test environment for the health management system, and has typical fault injection and simulation capabilities, health management model verification capabilities and health management system evaluation capabilities. Summary of the invention

[0006] The purpose of this invention is to achieve efficient, accurate and low-cost verification of system health management solutions by integrating multiple technical means. Running models of real hardware in a virtual simulation environment avoids the high risk and high cost of testing directly on actual equipment and reduces equipment loss during the testing process. Ensure the accuracy, reliability and availability of the health management system in all stages and application scenarios, and achieve more efficient and safe equipment management and maintenance.

[0007] A hybrid power system health management capability verification and evaluation method comprises the following steps:

[0008] Step 1. System modeling: In this stage, the goal and scope of fault diagnosis are clarified, and the corresponding mathematical, physical or simulation models are established. These models provide the basis for fault injection and ensure the accuracy and controllability of the fault injection process.

[0009] Step 2. Fault diagnosis algorithm selection and verification: A fault diagnosis algorithm suitable for the system characteristics is selected, which may be a data-driven, model-driven or hybrid-driven approach. Then the algorithm parameters are adjusted to adapt to the actual application environment, and the algorithm is tested and verified through simulation platforms and hardware-in-the-loop experiments.

[0010] Step 3. Fault injection: Fault simulation and testing of the hybrid system are achieved by building a fault injection test set, simulating fault behavior, monitoring system response, and automating injection diagnosis to optimize the system model and improve the efficiency and standardization level of fault diagnosis.

[0011] Step 4. Diagnostic testing: The possible fault types are defined and the key nodes or components for fault injection are identified. Then real-time testing of fault diagnosis is performed to verify the performance of the diagnostic system and record the results of fault injection and diagnosis in detail.

[0012] Beneficial effects of the present invention:

[0013] (1) The platform can collect data from different sensors and information systems in real time, such as operating status, temperature, vibration, etc., and process and analyze them immediately. This real-time feature helps to quickly identify signs of failure, take preventive measures in a timely manner, and reduce the risk of failure and maintenance costs.

[0014] (2) Simulation testing can not only reduce the cost of development and verification, but also avoid damage to the equipment and improve the safety of the overall test. Not only can verification be performed during the system design phase, but the health status of the equipment can also be continuously monitored during the production, operation and maintenance phases to predict and diagnose faults. By combining historical data with real-time data, the platform can provide effective support for the management of the entire life cycle of the equipment, extend the service life of the system and optimize operation and maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is the overall architecture design diagram of the present invention;

[0016] Figure 2 is a hybrid power system diagram of the present invention;

[0017] Figure 3 is a fault diagnosis algorithm diagram of the present invention;

[0018] Figure 4 is a fault injection flow chart of the present invention;

[0019] Figure 5 It is a system hardware architecture diagram of the present invention;

[0020] Figure 6 It is a functional diagram of the verification platform of the present invention. DETAILED DESCRIPTION

[0021] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0022] The overall architecture design diagram of the present invention is as follows: Figure 1 ,First, the hybrid system is studied, and then a high-precision simulation model is established, its health management model is designed, and then appropriate indicators are selected to evaluate the health management system. Figure 2 , decomposing the system, from feasibility study to component-level testing, to ensure that the system design meets the requirements and operates reliably.

[0023] Step 1: According to the needs of the specific hybrid power system, clarify the object, diagnosis scope, and verification target of fault diagnosis. Model the hybrid power system to be tested to obtain a hybrid power system model, and use the hybrid power system model to form a simulation platform or an in-the-loop test platform to ensure the accuracy and controllability of fault injection. The operation of the hybrid power system depends on a series of precise models working together. The hybrid power system model includes a vehicle dynamics model, a motor control model, a motor model, a battery control model, a battery model, a generator control model, a generator model, an engine-generator coupling device model, an engine controller model, an engine model, a power coupling device model, and a reduction gearbox model;

[0024] First, the vehicle dynamics model calculates the required power and torque based on the driver's operation and vehicle status. Next, the motor control model adjusts the operating state of the motor according to the power and torque, while the motor model is responsible for actually generating power. The battery control model manages the charging and discharging of the battery to ensure that the motor and generator have a stable power supply, while the battery model describes the physical characteristics and behavior of the battery. When the battery is low, the generator control model starts the generator to supplement the power or directly supply power, and the generator model describes the physical characteristics and power generation behavior of the generator. The engine-generator coupling device model is responsible for the mechanical connection between the engine and the generator, while the engine controller model controls the operation of the engine according to demand. The engine model describes the physical characteristics and operating behavior of the engine. The power coupling device model transmits the power generated by the engine, generator and motor to the reduction box, while the reduction box model converts high-speed, low-torque power into low-speed, high-torque power suitable for driving wheels. The series connection of all models ensures that the hybrid system intelligently distributes and converts power sources under different driving conditions to achieve optimal fuel efficiency and performance.

[0025] Step 2 constructs a classifier-free guided diffusion model based on data-driven, model-driven or hybrid-driven. Adjust and optimize the parameters of the classifier-free guided diffusion model to adapt to the actual application environment and system. In actual work, experimental data faces the problem of data imbalance, that is, the number of fault data is less than that of healthy data. The present invention uses a method for generating fault data using a classifier-free guided diffusion model to solve the data imbalance problem. Figure 3 .

[0026] The classifier-free guided diffusion model uses real fault data, simulates the data generation process, gradually adds noise and performs denoising, and generates synthetic fault data similar to the real fault data. The synthetic fault data is integrated with the real fault data to further obtain all fault types (such as generating mechanical, electrical, sensor, and software faults in the power system, transmission system, and auxiliary system).

[0027] The classifier-free guided diffusion model adds a multi-head self-attention mechanism and jump connection on the basis of the diffusion model to enhance the feature extraction capability, so that the classifier-free guided diffusion model can process time steps and label embedding to generate hybrid power system fault data. The diffusion model includes an encoder, an intermediate feature extractor and a decoder. The classifier-free guided diffusion model takes into account the data imbalance problem and optimizes the performance of generating data by adjusting key hyperparameters.

[0028] Furthermore, synthetic fault data is injected into the simulation platform to test and verify the classifier-free guided diffusion model. Synthetic fault data is injected into the in-the-loop test platform, combined with hardware-in-the-loop experiments or semi-physical experiments, physical fault injection and diagnosis are performed to verify the performance of the classifier-free guided diffusion model in actual scenarios.

[0029] Step 3: All fault types are combined into a fault injection test set, and relevant faults are called from the fault injection test set and injected into the simulation platform or in the loop test platform to simulate the fault behavior of the hybrid power system and its related equipment; the injection of different fault types is achieved in the following ways:

[0030] Method 1: short-circuit or open-circuit some models in the simulation platform or the loop test platform;

[0031] Method 2: Change the component parameters in the simulation platform or some models in the loop test platform;

[0032] Furthermore, testing needs to be performed after each fault injection. During the test, the response of the hybrid system is monitored and relevant data is recorded for analysis. Based on the analysis results, the hybrid system model, test cases and injection methods after the fault injection are adjusted to better simulate real-world conditions. This ensures that the parameters during the fault injection process are controllable, and the injection time and fault intensity can be quantitatively controlled.

[0033] Furthermore, the fault injection process is implemented by using automated injection and automated diagnosis technologies, which reduces manual intervention, quantifies fault injection and diagnosis results, provides a basis for establishing the standardization and replicability of the fault diagnosis system, and improves the efficiency and standardization level of the verification process.

[0034] Step 4: Fault injection flow chart Figure 4 . Connect the health management system to the simulation platform or the in-loop test platform, and perform fault injection according to the method in step 3. After the fault injection, perform a real-time test of fault diagnosis to verify whether the health management system can detect, identify and locate the fault in time when the fault occurs. Record the detailed results of each fault injection and fault diagnosis, including data such as fault type, injection parameters, and diagnostic delay. Quantitatively measure various evaluation indicators of the health management system to form a verification report containing fault injection methods, diagnostic results, comparative analysis, optimization suggestions, etc., as the basis for the final evaluation of the diagnostic system performance. The evaluation indicators selected for the verification evaluation system are as follows:

[0035] Table 1 Evaluation indicators

[0036] Evaluation indicators Introduction Fault isolation rate The proportion of all detected faults that can be successfully isolated and identified Fault detection time The time interval from when a fault occurs to when it is recognized by the detection system Fault isolation time The time required from the time a fault is detected to the time the source of the fault is successfully identified and isolated Fault detection rate The proportion of all faults that can be successfully detected by the detection system Prediction accuracy The proportion of correct predictions among all predictions

[0037] The corresponding calculation formula is:

[0038] 1. Fault Isolation Rate (FIR)

[0039] Definition: The ratio of the number of failures that the system can correctly isolate to specific faulty components to the total number of actual failures that occur.

[0040] Calculation formula: Standard: FIR ≥ 90%

[0041] 2. Mean Time to Failure Detection (MTTD)

[0042] Definition: The average time required for a system to detect a fault from the beginning until it is confirmed.

[0043] Calculation formula: Standard: MTTD≤5 minutes

[0044] 3. Mean time to isolate (MTTI)

[0045] Definition: The average time required for a system to isolate and confirm the faulty component.

[0046] Calculation formula: Standard: MTTI ≤ 10 minutes

[0047] 4. Fault Detection Rate (FDR)

[0048] Definition: The ratio of the number of faults that a system is able to correctly detect to the total number of faults that actually occur.

[0049] Calculation formula: Standard: FDR ≥ 95%

[0050] 5. Prediction accuracy (PA)

[0051] Definition: The proportion of correctly predicted results among all predicted results.

[0052] Calculation formula: Standard: PA ≥ 90%.

[0053] For the health management system, the corresponding demand indicators are formulated to cover the entire life cycle of the health management system design, integration, physical and semi-physical verification and demand changes, which can make the realization efficient, reduce the equipment cost in the entire life cycle and improve economic benefits. When comprehensively evaluating the health management system, it is necessary to combine the performance of the algorithm, the design and implementation of the system, and the management of the entire life cycle to ensure the effectiveness and economy of the system. Establish a comprehensive test plan covering functional testing, performance testing and safety testing to ensure the normal operation of the system under various circumstances. By formulating comprehensive demand indicators, ensure that the system can operate efficiently in all aspects of design, integration, verification and demand change management, thereby achieving the goal of reducing the cost of the entire life cycle and improving economic benefits.

[0054] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above specific embodiments. The above embodiments are merely illustrative and instructive, and are not restrictive. A person of ordinary skill in the art may make various forms under the guidance of this specification and without departing from the scope of protection of the claims of the present invention, all of which belong to the scope of protection of the present invention.

Claims

1. A hybrid power system health management capability verification and evaluation method, characterized in that: Here are the steps: Step 1: According to the requirements of the specific hybrid power system, the object, diagnosis scope, and verification target of the fault diagnosis are clarified; the hybrid power system to be tested is modeled to obtain a hybrid power system model, and a simulation platform or an in-loop test platform is constructed using the hybrid power system model to ensure the accuracy and controllability of the fault injection; the hybrid power system model includes a vehicle dynamics model, a motor control model, a motor model, a battery control model, a battery model, a generator control model, a generator model, an engine-generator coupling device model, an engine controller model, an engine model, a power coupling device model, and a reduction box model; Step 2: construct a classifier-free guided diffusion model based on data-driven, model-driven or hybrid-driven; Adjust and optimize the parameters of the non-classifier guided diffusion model to adapt to the actual application environment and system; A method to generate fault data using a classifier-free guided diffusion model to solve the data imbalance problem; The classifier-free guided diffusion model uses real fault data, simulates the data generation process, gradually adds noise and performs denoising, and generates synthetic fault data similar to the real fault data; integrates the synthetic fault data with the real fault data to further obtain all fault types; The classifier-free guided diffusion model adds a multi-head self-attention mechanism and a jump connection on the basis of the diffusion model to enhance the feature extraction capability, so that the classifier-free guided diffusion model can process time steps and label embeddings to generate hybrid power system fault data; the diffusion model includes an encoder, an intermediate feature extractor and a decoder; the classifier-free guided diffusion model takes into account the problem of data imbalance and optimizes the performance of generating data by adjusting key hyperparameters; Step 3: All fault types are combined into a fault injection test set, and relevant faults are called from the fault injection test set and injected into the simulation platform or in the loop test platform to simulate the fault behavior of the hybrid power system and its related equipment; the injection of different fault types is achieved in the following ways: Method 1: short-circuit or open-circuit some models in the simulation platform or the loop test platform; Method 2: Change the component parameters in the simulation platform or some models in the loop test platform; Step 4: Connect the health management system to the simulation platform or the in-loop test platform and perform fault injection in the same manner as in step 3; after the fault injection, perform a real-time test of fault diagnosis to verify whether the health management system can detect, identify and locate the fault in a timely manner when the fault occurs; Record the detailed results of each fault injection and fault diagnosis, including fault type, injection parameters, and diagnostic delay data; quantitatively measure various evaluation indicators of the health management system to form a verification report that includes fault injection methods, diagnostic results, comparative analysis, and optimization suggestions as the basis for the final evaluation of the diagnostic system performance.

2. A hybrid power system health management capability verification and evaluation method according to claim 1, characterized in that: In the step 1, first, the vehicle dynamics model calculates the required power and torque according to the driver's operation and the vehicle state; then, the motor control model adjusts the operating state of the motor according to the power and torque, and the motor model is responsible for actually generating power; the battery control model manages the charging and discharging of the battery to ensure that the motor and the generator have a stable power supply, and the battery model describes the physical characteristics and behavior of the battery; when the battery is low on power, the generator control model starts the generator to supplement the power or directly supply power, and the generator model describes the physical characteristics and power generation behavior of the generator; The engine-generator coupling device model is responsible for the mechanical connection between the engine and the generator, while the engine controller model controls the operation of the engine according to demand; the engine model describes the physical characteristics and operating behavior of the engine; the power coupling device model transfers the power generated by the engine, generator and motor to the reduction box, while the reduction box model converts high-speed, low-torque power into low-speed, high-torque power suitable for driving wheels; the series connection of all models ensures that the hybrid system intelligently distributes and converts power sources under different driving conditions to achieve optimal fuel efficiency and performance.

3. A hybrid power system health management capability verification and evaluation method according to claim 1, characterized in that: In the step 2, synthetic fault data is injected into the simulation platform to test and verify the classifier-free guided diffusion model; synthetic fault data is injected into the in-the-loop test platform, and physical fault injection and diagnosis are performed in combination with hardware-in-the-loop experiments or semi-physical experiments to verify the performance of the classifier-free guided diffusion model in actual scenarios.

4. A hybrid power system health management capability verification and evaluation method according to claim 1, characterized in that: In step 3, a test needs to be performed after each fault injection; during the test, the response of the hybrid system is monitored and relevant data is recorded for analysis; based on the analysis results, the hybrid system model, test cases and injection method after the fault injection are adjusted to better simulate real-world conditions; To ensure that the parameters in the fault injection process are controllable, the injection time and fault intensity can be quantitatively controlled.

5. A hybrid power system health management capability verification and evaluation method according to claim 1, characterized in that: In step 3, automated injection and automated diagnosis techniques are used to implement the fault injection process, reduce manual intervention, and quantify fault injection and diagnosis results.

6. A hybrid power system health management capability verification and evaluation method according to claim 1, characterized in that: In step 4, the evaluation indicators include: fault isolation rate, fault detection time, fault isolation time, fault detection rate, and prediction accuracy; the calculation process is as follows: Fault isolation rate: Standard: FIR ≥ 90%; Mean time to detect failure: Standard: MTTD ≤ 5 minutes; Mean time to isolate fault: Standard: MTTI ≤ 10 minutes; Fault detection rate: Standard: FDR ≥ 95%; Prediction accuracy: Standard: PA ≥ 90%.

Images