A hybrid system health management method and system

Abstract

The application discloses a kind of hybrid power system health management method and system, belong to power system health management field.This application selects the characteristic signal of engine, battery and vehicle in hybrid power system to establish hybrid power system health degree;When hybrid power system fails, by obtaining characteristic signal and carrying out normalization processing, hybrid power system health degree based on grey correlation analysis method GRA is calculated in real time, and the misjudgment and the missed judgment of fault are reduced by the fault confirmation method based on counting;When hybrid power system fault is confirmed, through the power redistribution health management strategy under system fault based on LSTM-MPC, the change of hybrid power system speed, battery temperature and diesel engine exhaust temperature in limited time domain is predicted, and the predictive maintenance of hybrid power system health degree is carried out in combination with the change prediction result of part of the state of hybrid power system, so that hybrid power system can still operate with fault under the condition of meeting constraint.

Description

TECHNICAL FIELD

[0001] The application relates to a hybrid power system health management method and system, and belongs to the technical field of power system health management. BACKGROUND

[0002] The hybrid power system is one of main power systems for realizing the double-carbon target. In particular, in recent years, the hybrid power system presents a situation of coexistence of multiple types for different application scenarios. Although the systems are different, the composition and operation mode of the hybrid power system are becoming more and more complex, which has become a common technical trend. For a complex system, it is necessary to carry out health management of the system.

[0003] In particular, from the system level, considering the characteristics of the hybrid power system with multiple power sources, that is, each power source can work independently and can be backed up by each other, when the system performance appears to be degraded or deteriorated, a health management strategy meeting the constraint condition can be formulated to improve the system reliability, realize real-time health state evaluation of the system and predictive maintenance after system failure, that is, by reconfiguring the control strategy of the system, the system can still realize fault-tolerant operation under the condition of meeting the constraint condition, which has important significance for the control of the hybrid power system. SUMMARY

[0004] The application aims to provide a hybrid power system health management method and system, which selects characteristic signals of an engine, a battery and a vehicle in the hybrid power system to establish a hybrid power system health degree, obtains the characteristic signals and carries out normalization processing when the hybrid power system fails, calculates the hybrid power system health degree based on a grey correlation analysis method (GRA) in real time, reduces false judgment and missed judgment of the failure through a failure confirmation method based on counting, accurately judges whether the hybrid power system fails according to the degree of deviation of the health degree from the normal value after the failure, and when the hybrid power system failure is confirmed, a power redistribution health management strategy under the system failure based on LSTM-MPC is used to predict the changes of the hybrid power system speed, the battery temperature and the diesel engine exhaust temperature within a limited time range, and the health degree of the hybrid power system is predicted and maintained in a predictive manner in combination with the prediction results of the changes of part of the states of the hybrid power system, so that the hybrid power system can still operate with failure under the condition of meeting the constraint condition. The part of the states of the hybrid power system refers to the vehicle speed, the battery temperature and the diesel engine temperature.

[0005] The application aims to achieve the above-mentioned purposes through the following technical solutions.

[0006] The application discloses a hybrid power system health management method, which comprises the following steps:

[0007] Step one, through the measured signal acquisition and processing module to obtain diesel engine instantaneous analog signal, battery instantaneous analog signal, vehicle instantaneous analog signal, and get the corresponding hybrid power assembly state signal, the state signal of the hybrid power assembly is normalized, and the normalized hybrid power assembly diesel engine instantaneous state signal, battery instantaneous state signal and vehicle instantaneous state signal are obtained.

[0008] Step 1.1: obtaining the original instantaneous analog signal of the vehicle component through the measured signal acquisition and processing module;

[0009] Step 1.2: signal conversion and calibration are performed on the original instantaneous analog signal obtained in step 1.1 to obtain the relationship between the original analog signal and the original digital signal, and filtering processing is performed to obtain the instantaneous state signal of the vehicle;

[0010] Step 1.3: based on the original instantaneous analog signal obtained in step 1.1 and the instantaneous state signal obtained in step 1.2, the hybrid power system state signal is obtained, which includes three types of signals, i.e. diesel engine instantaneous state signal, battery instantaneous state signal and vehicle instantaneous state signal: the diesel engine instantaneous state signal includes intake temperature, intake pressure, exhaust temperature, diesel engine speed and throttle pedal position; the battery instantaneous state signal includes SOC, battery bus voltage, battery bus current and battery temperature; the vehicle instantaneous state signal includes target vehicle speed, actual vehicle speed and vehicle operation mode.

[0011] Step 1.4: signal processing is performed on the state signal of the hybrid power assembly obtained in step 1.3, and normalization processing is performed to eliminate the influence of different dimensions, and the normalized state signal of the hybrid power assembly is obtained.

[0012] Step two, according to the uniform dimension state signal input by the measured signal acquisition and processing module, combined with the fault form and the real vehicle sensor signal, part of the state signal of the engine, battery and vehicle in the hybrid power system is selected as the characteristic signal, the selected characteristic signal includes diesel engine intake pressure, diesel engine intake temperature, diesel engine exhaust temperature, throttle pedal position, battery bus voltage, battery bus current, battery temperature and vehicle speed difference. The feature signal weight is determined by principal component analysis, and the health degree of the hybrid power system based on gray correlation analysis method GRA is calculated in real time. The fault is confirmed by the fault confirmation method based on counting, the fault misjudgment and omission caused by the error and noise of the characteristic signal are reduced, and the fault confirmation signal is obtained.

[0013] Step 3: A health management strategy for power redistribution under system faults based on LSTM-MPC. The hybrid powertrain signal obtained in Step 1 is received as a predictive maintenance indicator. After receiving the fault confirmation signal obtained in Step 2, a partial prediction model for MPC is obtained through LSTM training. Based on the LSTM-MPC prediction model, changes in partial system states within a finite time domain are predicted. The health of the hybrid power system is then predictively maintained based on the predicted changes in partial hybrid system states. When a fault occurs in the hybrid power system, the conversion relationship between energy consumption and fuel consumption is given. Battery energy consumption is equated to fuel energy consumption, and the comprehensive energy consumption is calculated. By sacrificing comprehensive energy consumption, the target speed following effect is improved, while ensuring that the battery operates within its optimal temperature range and the diesel engine does not exceed the exhaust temperature limit. This ensures that the hybrid power system can still operate with the fault under constrained conditions.

[0014] The power redistribution health management strategy under system faults based on LSTM-MPC has the following rules for the MPC control strategy:

[0015] At the current time k, the LSTM-based prediction model obtains the future time domain k+t by acquiring the current k-time input of the controlled object and information from historical times. p The optimal control sequence is obtained by retaining only the first control variable u of the optimal control sequence. p (k+1) and apply it to the controlled object, ignoring all other control variables, and repeat the above optimization process at the next sampling time. At time k, the optimization objective is represented by equation (1):

[0016]

[0017] In equation (1), J k For the cost function, t p Let x(t) be the prediction duration, x(t) be the state variable at time t, and u(t) be the control variable at time t.

[0018] The constraint is expressed by formula (2):

[0019]

[0020] In equation (2), x min (t) and x max (t) represents the upper and lower limits of the state variable at time t, respectively. min (t) and u max (t) represents the upper and lower limits of the control variable at time t.

[0021] According to the multi-power source characteristics of the hybrid system, the optimization objective is to minimize the speed following error by sacrificing the comprehensive energy consumption. The constraint conditions include the diesel exhaust temperature constraint, the battery temperature constraint, and other theoretical physical constraints of various components. At time k, in the prediction time domain, the cost function of the hybrid system is represented by equation (3):

[0022]

[0023] In equation (3), t p is the prediction length, k p is the weight coefficient of the speed difference, k e is the weight coefficient of the comprehensive energy consumption, u(t) is the control quantity, which is the throttle pedal position correction coefficient here, v a (u(t), t) and v t (t) are the actual speed and target speed, respectively, in km / h.

[0024] is the comprehensive energy consumption, g / s.

[0025] The constraint condition is represented by equation (4):

[0026]

[0027] In equation (4), T e is the diesel exhaust temperature, T e_max is the diesel exhaust temperature limit value; T b is the battery temperature,

[0028] [T b_smin , T b_smax ] is the optimal working temperature range of the lithium-ion battery, with a value of [15, 35] ℃; SOC min , SOC max are the upper and lower limits of the power battery charge and discharge SOC; P dem , P e and P m are the total demand power, diesel engine power, and motor power, respectively, in kW.

[0029] The input of the LSTM prediction model is the health status evaluation result, vehicle operating mode, throttle pedal position, battery power, actual speed, battery temperature, and diesel exhaust temperature. The output is the speed of the hybrid system, the diesel exhaust temperature, and the battery temperature. The speed, battery temperature, and diesel exhaust temperature at future time are predicted by rolling the input information of the historical time period and the current time. The mean absolute error (MAE) is selected as the loss function for optimizing the parameters of the prediction model and measuring the accuracy of the prediction model. The MAE can be represented by equation (5):

[0030]

[0031] In formula (5), Y is a true value, f(x) is a predicted value, and n is a sample number.

[0032] As preferred, according to the conversion relationship between the electric energy consumption and the fuel consumption given in GB / T 19754-2021 “Heavy-duty hybrid electric vehicle energy consumption test method”, the battery energy consumption is equivalent to the fuel energy consumption, the comprehensive energy consumption is calculated, the target vehicle speed following effect is improved by sacrificing the comprehensive energy consumption, meanwhile, the battery is ensured to work in the best temperature range and the diesel engine is ensured not to exceed the exhaust temperature limit, and the hybrid power system is ensured to still be able to operate with faults under the satisfaction of the constraint conditions.

[0033] The application further discloses a hybrid power system health management system for realizing the hybrid power system health management method. The hybrid power system health management system mainly comprises a host computer PC, a hybrid power assembly, a vehicle controller VCU, a signal acquisition and processing module, a health state evaluation module and a power distribution correction module based on LSTM-MPC. The host computer PC and the vehicle controller VCU are connected through a signal line. The hybrid power assembly is connected with the vehicle controller VCU through a signal bus. The host computer PC controls the vehicle controller VCU to receive and process signals from the hybrid power assembly, and performs real-time signal transmission interaction. The signal acquisition and processing module, the health state evaluation module and the power distribution correction module based on LSTM-MPC written in the vehicle controller VCU are used to evaluate the health state of the hybrid power system, and the hybrid power system is corrected according to the health state evaluation result.

[0034] The signal acquisition and processing module receives and processes the hybrid power system state signals, and the state signals include three types of signals, i.e., diesel engine characteristic signals, battery characteristic signals and vehicle characteristic signals. The diesel engine characteristic signals include intake air temperature, intake air pressure, exhaust temperature, diesel engine speed and accelerator pedal position. The battery characteristic signals include SOC, battery bus voltage, battery bus current and battery temperature. The vehicle characteristic signals include target vehicle speed, actual vehicle speed and vehicle operation mode. After the signals are acquired, the state signals are normalized to eliminate the influence of different dimensions, and the health state evaluation module is preprocessed. Through the normalization of the state signals, the health state evaluation efficiency of the hybrid power system is improved on the basis of ensuring the accuracy of characteristic signals of different orders of magnitude.

[0035] According to formula (6), the time series data of intake air pressure, intake air temperature, exhaust temperature, accelerator pedal position, bus current, bus voltage, battery temperature and vehicle speed are normalized.

[0036]

[0037] wherein, i = 1, 2, 3, …, n, k = 1, 2, 3, …, m, n is the number of samples, and m is the number of features

[0038] The health state evaluation module acquires the state signal from the signal acquisition and processing module, calculates the health degree through the vehicle controller VCU according to the health degree, judges the health state of the hybrid power assembly, analyzes the hybrid power working condition, and diagnoses whether the system has a fault.

[0039] The health degree is constructed according to the normalized feature signal, that is, constructed according to the intake air pressure, intake air temperature, exhaust temperature, accelerator pedal position, bus current, bus voltage, battery temperature and vehicle speed, and the constructed health degree is embedded into the health state evaluation module.

[0040] The health degree construction method is as follows:

[0041] The diesel-electric hybrid system has the characteristics of high coupling, nonlinearity and multiple power sources, and is comprehensively judged through multiple dimensions and multiple standards. The multi-dimensional state signals collected by the diesel-electric hybrid system are combined, and the grey correlation analysis is used to calculate the correlation degree between the running state and the health state (normal state) of the hybrid system. The correlation degree is used to reflect the deviation degree of the running state of the hybrid system from the health state, and the greater the deviation degree, the smaller the correlation degree.

[0042] Specifically, the health degree is as follows:

[0043] 1) The normalized data output by the signal acquisition and processing module to be measured is used to calculate the grey correlation coefficients one by one;

[0044]

[0045] In formula (7), k = 1, 2, 3, …, m, i = 1, 2, 3, …, n, and p represents a discrimination coefficient, p ∈ (0, 1), the smaller the p, the greater the difference between the correlation coefficients, and the stronger the resolution ability, and the preferred value is 0.5.

[0046] 2) The correlation degree is calculated one by one according to formula (8).

[0047]

[0048] In formula (8), ω i (k) represents the weight coefficient of the feature signal, k = 1, 2, 3, …, m, and i = 1, 2, 3, …, n.

[0049] The health state evaluation of the hybrid system based on GRA needs to calculate the correlation degree. Since the sensitivity of different characteristic signals in the hybrid system to faults is different, directly taking the average value may lead to inaccurate correlation degree calculation. The weight value of the characteristic signal in different dimensions is determined by using the PCA method.

[0050] The power distribution correction module based on LSTM-MPC obtains the output from the health state evaluation module and the to-be-tested signal, obtains the hybrid power system state in the health state evaluation module, calculates the power redistribution health management method based on LSTM-MPC through the vehicle controller VCU, and calculates the accelerator pedal position correction coefficient in advance. After confirming the fault, the correction coefficient is output to correct the accelerator pedal position of the vehicle controller, that is, under the condition of hybrid power system failure, the target speed following effect is improved by sacrificing the comprehensive energy consumption, while the battery is ensured to work in its optimal temperature range and the diesel engine is ensured not to exceed the exhaust temperature limit, so that the hybrid power system can still operate with faults under the condition of meeting the constraint conditions.

[0051] Advantages:

[0052] 1. The hybrid power system health management method and system disclosed by the application can accurately evaluate the health state of the hybrid power system by analyzing the hybrid power system failure, determining the characteristic signal, and using PCA to confirm the weight coefficient of each characteristic signal to obtain the health degree of the hybrid power system, compared with directly taking the average value as the weight coefficient of each characteristic signal.

[0053] 2. The hybrid power system health management method and system disclosed by the application uses the health management strategy of power redistribution under system failure based on LSTM-MPC, fully utilizes the advantages of the hybrid system with multiple power sources, enables the hybrid power system to operate with faults under the condition of meeting the constraint conditions, increases the tolerance of the hybrid power system, and the original vehicle controller VCU of the vehicle can meet the calculation requirements, compared with the health management strategy based on MPC, has the advantages of fast response and strong practicability.

[0054] 3. The hybrid power system health management method and system disclosed by the application integrates the constructed health degree into the health state evaluation module, directly calls the health degree for health state evaluation during hybrid power system health evaluation, has faster evaluation speed, and can improve the health state evaluation efficiency.

[0055] 4, The application discloses a hybrid power system health management method and system, a health management strategy for power redistribution under system failure based on an LSTM-MPC, receives hybrid power assembly signals as predictive maintenance indexes, after receiving a failure confirmation signal, a partial prediction model for MPC is obtained through LSTM training, changes in system partial states within a limited time domain are predicted based on the LSTM-MPC prediction model, and the hybrid power system health degree is predicted in combination with the change prediction result of the hybrid power system partial states. When the hybrid power system fails, the conversion relationship between the electric quantity consumption and the fuel consumption is given, the battery energy consumption is equivalent to the fuel energy consumption, the comprehensive energy consumption is calculated, the target vehicle speed following effect is improved by sacrificing the comprehensive energy consumption, and meanwhile, the battery is ensured to work in its optimal temperature range and the diesel engine is ensured not to exceed the exhaust temperature limit, so that the hybrid power system can still operate with failure under the condition of meeting the constraints.

[0056] 5, The application discloses a hybrid power system health management method and system, which calculates the health degree of the hybrid power system based on a grey correlation analysis method GRA in real time by acquiring characteristic signals and performing normalization processing on the characteristic signals, reduces misjudgment and omission of failure through a failure confirmation method based on counting, accurately judges whether the hybrid power system fails according to the degree to which the health degree deviates from a normal value after failure, and improves the precision and efficiency of health management of the hybrid power system.

[0057] 6, The application discloses a hybrid power system health management method and system, which converts battery energy consumption into fuel energy consumption according to the conversion relationship between the electric quantity consumption and the fuel consumption given in GB / T 19754-2021 "Heavy Hybrid Electric Vehicle Energy Consumption Test Method", and unifies the calculated comprehensive energy consumption into fuel consumption, so that energy consumption calculation and statistics of the hybrid power system are facilitated. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is a schematic diagram of a hybrid power system health management system of the application;

[0059] Figure 2 is a flowchart of a hybrid power system health management method of the application;

[0060] Figure 3 is a schematic diagram of health degree and failure confirmation under a battery internal resistance increase failure of the application;

[0061] Figure 4 is a flowchart of a power redistribution control strategy algorithm based on LSTM-MPC of the application;

[0062] Figure 5 is a vehicle speed following situation diagram with or without power correction under a battery internal resistance failure of the application;

[0063] Figure 6 is a battery temperature and exhaust temperature situation chart with or without power correction under battery internal resistance failure of the present application.

[0064] Figure 7 is a comprehensive fuel consumption situation chart with or without power correction under battery internal resistance failure of the present application. DETAILED DESCRIPTION

[0065] In order to better illustrate the purposes and advantages of the present application, the following further illustrates the content of the application in combination with the drawings and examples.

[0066] Example 1:

[0067] This embodiment takes a certain type of P2 parallel hybrid electric vehicle as the background, and a hybrid power system health evaluation method is composed of a to-be-measured signal acquisition and processing module, a health state evaluation module and a power distribution correction module. Combined with the hybrid power system health evaluation system disclosed in the invention patent, through real-time data interaction with the hybrid power system, specific health evaluation is carried out. The system health evaluation system of the present application can quickly and accurately judge the health state of the hybrid power system, and can effectively utilize the advantages of the hybrid system, and through the method based on LSTM-MPC, the output power ratio of the engine and the battery is redistributed, so that the system can run with failure within a certain constraint condition.

[0068] As shown in Figure 1 The hybrid power system health management system disclosed in the present embodiment is composed of a host computer PC, a hybrid power assembly, a vehicle controller lower computer VCU, a to-be-measured signal acquisition and analysis module, a health state evaluation module and a power distribution correction module. The host computer PC and the vehicle controller lower computer VCU are connected through RS485 signal lines. The hybrid power assembly is connected with the vehicle controller lower computer VCU through CANs bus, and the host computer PC controls the vehicle controller lower computer VCU to receive and process signals from the hybrid power assembly, and performs real-time signal transmission interaction. Based on the to-be-measured signal acquisition and analysis module, the health state evaluation module and the power distribution correction module written in the vehicle controller lower computer VCU, the health state of the hybrid power system is evaluated, and the power distribution of the hybrid power system is corrected according to the health state obtained by evaluation. The working process of the hybrid power system health evaluation method is as shown in Figure 2As shown, firstly, the to-be-tested signal acquisition and analysis module acquires the driving to-be-tested state signals of the engine and the battery in the vehicle through the CANs bus, and performs normalization processing on the signals to eliminate the influence of different dimensions, thereby pre-processing for the subsequent health state evaluation module; the health state evaluation module acquires the state signals from the to-be-tested signal acquisition and analysis module, calculates the health degree through the vehicle controller lower computer VCU, analyzes the hybrid power working condition, and judges whether the system is in a healthy state. The power distribution correction module acquires the output from the health state evaluation module and the hybrid power system state in the to-be-tested signal acquisition and processing module, calculates the power distribution correction control strategy based on LSTM-MPC through the vehicle controller lower computer VCU, continuously reduces the power proportion of the fault subsystem of the hybrid power assembly, and achieves the optimization goal. For the power gap reduced by the fault subsystem, the other normal subsystems in the hybrid power assembly supplement the power gap, obtain a new power distribution of the hybrid power assembly under fault, fully exert the advantages of the hybrid power system, and enable the system to operate with fault within the specified constraint range.

[0069] The to-be-tested signal acquisition and processing module receives and processes hybrid power system state signals, and the state signals include three types of signals, i.e., diesel engine characteristic signals, battery characteristic signals, and vehicle characteristic signals: the diesel engine characteristic signals include intake temperature, intake pressure, exhaust temperature, diesel engine speed, and accelerator pedal position; the battery characteristic signals include SOC, battery bus voltage, battery bus current, and battery temperature; and the vehicle characteristic signals include target vehicle speed, actual vehicle speed, and vehicle running mode. After the signals are acquired, the state signals are normalized to eliminate the influence of different dimensions, thereby pre-processing for the subsequent health state evaluation module. Through unifying the dimensions of the state signals, the accuracy of the characteristic signals under different orders of magnitude is ensured, and the health state evaluation efficiency of the hybrid power system is improved.

[0070] According to formula (6), the time series data of intake pressure, intake temperature, exhaust temperature, accelerator pedal position, bus current, bus voltage, battery temperature, and vehicle speed are normalized:

[0071]

[0072] wherein i = 1, 2, 3, …, n, k = 1, 2, 3, …, m, n is the sample number, and m is the characteristic number

[0073] The health state evaluation module acquires the state signals from the signal acquisition and processing module, calculates the health degree through the vehicle controller lower computer VCU according to the health degree, judges the health state of the hybrid power assembly, analyzes the hybrid power working condition, and diagnoses whether the system is faulty.

[0074] The battery internal resistance in the hybrid assembly is increased to simulate a real fault condition, and the health degree is used to quickly and accurately determine whether the hybrid assembly has failed;

[0075] The health degree is as follows:

[0076] 1) The normalized data output by the signal acquisition and processing module is used to calculate the grey correlation coefficient one by one;

[0077]

[0078] In formula (7), k = 1, 2, 3, …, m, i = 1, 2, 3, …, n, and p represents a discrimination coefficient, p ∈ (0, 1), the smaller p is, the greater the difference in correlation coefficients is, and the stronger the resolution capability is, and the general value is 0.5.

[0079] 2) The correlation degree is calculated one by one.

[0080]

[0081] In formula (8), ω i (k) represents a weight coefficient of the characteristic signal, k = 1, 2, 3, …, m, and i = 1, 2, 3, …, n.

[0082] The weight value of the characteristic signal of different dimensions is determined by using the PCA method.

[0083] The characteristic signal data under normal circulation and hybrid system fault circulation are collected as sample sets, and the weight values of the characteristic signals under different modes are obtained by the PCA method as follows:

[0084] Battery driving:

[0085] 1) Bus current: 15.7

[0086] 2) Bus voltage: 4.7

[0087] 3) Battery temperature: 42.3

[0088] 4) Speed difference: 37.3

[0089] Diesel engine driving:

[0090] 1) Intake pressure: 18.5

[0091] 2) Intake temperature: 19.7

[0092] 3) Exhaust temperature: 33.4

[0093] 4) Throttle pedal: 12.1

[0094] 5) Speed difference: 16.3

[0095] Joint drive:

[0096] 1) Intake pressure: 15.6

[0097] 2) Intake temperature: 16.7

[0098] 3) Exhaust temperature: 29.1

[0099] 4) Accelerator pedal: 10.4

[0100] 5) Bus current: 6.6

[0101] 6) Bus voltage: 1.6

[0102] 7) Battery temperature: 11.3

[0103] 8) Speed difference: 8.7

[0104] The health degree obtained is used to judge the fault state of the hybrid powertrain, and the fault state is accumulated by counting the mode to reduce false positives and false negatives. The upper threshold is set to 127, and the lower threshold is set to -128. The counter is incremented by 1 when the current fault state is 1 (i.e. there is a fault), and decremented by 1 when it is 0 (no fault). When the threshold is reached, the fault confirmation is no longer accumulated. When the count reaches the set threshold upper limit, fault confirmation is performed.

[0105] The health degree and fault confirmation in the case of battery internal resistance increase fault in the hybrid powertrain are as shown in Figure 3 .

[0106] The power distribution correction module based on LSTM-MPC obtains the output from the health state evaluation module and the to-be-tested signal, obtains the hybrid power system state in the health state evaluation module, calculates the power redistribution health management method based on LSTM-MPC through the lower computer VCU of the vehicle controller, and calculates the accelerator pedal position correction coefficient in advance. When the fault is confirmed, the correction coefficient is output to correct the accelerator pedal position of the vehicle controller, that is, under system failure, the target speed following effect is improved by sacrificing a certain comprehensive energy consumption, while ensuring that the battery works in its best temperature range and the diesel engine does not exceed the exhaust temperature limit, ensuring that the system can still operate with faults under the constraint condition. The algorithm flowchart is as shown in Figure 4 .

[0107] When the hybrid powertrain is in the fault of increased battery internal resistance, the speed following condition is poor, and the battery is out of the best working temperature range. After power distribution correction, the speed following condition (power performance) is improved at the expense of a certain comprehensive fuel consumption (economy), and the battery works in the best temperature range and the diesel engine does not exceed the exhaust temperature limit, as shown in Figure 5 , 6, 7.

[0108] The embodiment can evaluate the health state of the hybrid power system, fully exert the advantages of the hybrid power system, and through the power redistribution strategy based on the LSTM-MPC, redistribute the power when a fault occurs, so that the hybrid power system operates with the fault under the condition that the battery is in the optimal working temperature range and the exhaust temperature of the diesel engine does not exceed the limit, and the tolerance of the hybrid power system is increased. Meanwhile, the health management method and system of the hybrid power system disclosed by the embodiment embed the health degree constructed into the health state evaluation module, directly call the health degree index during the health evaluation of the hybrid power system, and the evaluation speed is fast, the health evaluation efficiency is improved, the original vehicle controller VCU can meet the calculation requirements, and the method has the characteristics of fast speed and strong practicability.

[0109] As shown in Figure 2 The embodiment discloses a hybrid power system health management method, and the specific implementation steps are as follows:

[0110] Step one, through the measured signal acquisition and processing module, diesel engine instantaneous analog signals, battery instantaneous analog signals and vehicle instantaneous analog signals are obtained, and corresponding hybrid power assembly state signals are obtained. The hybrid power assembly state signals are normalized to obtain the normalized hybrid power assembly diesel engine instantaneous state signal, battery instantaneous state signal and vehicle instantaneous state signal.

[0111] Step 1.1: obtaining the original instantaneous analog signal of the vehicle component through the measured signal acquisition and processing module;

[0112] Step 1.2: signal conversion and calibration are performed on the original instantaneous analog signal obtained in step 1.1 to obtain the relationship between the original analog signal and the original digital signal, and filtering processing is performed to obtain the instantaneous state signal of the vehicle;

[0113] Step 1.3: based on the original instantaneous analog signal obtained in step 1.1 and the instantaneous state signal obtained in step 1.2, the hybrid power system state signal is obtained, and the state signal includes three types of signals, i.e., diesel engine instantaneous state signal, battery instantaneous state signal and vehicle instantaneous state signal: the diesel engine instantaneous state signal includes intake temperature, intake pressure, exhaust temperature, diesel engine speed and throttle pedal position; the battery instantaneous state signal includes SOC, battery bus voltage, battery bus current and battery temperature; and the vehicle instantaneous state signal includes target vehicle speed, actual vehicle speed and vehicle operation mode.

[0114] Step 1.4: signal processing is performed on the hybrid power assembly state signal obtained in step 1.3, and normalization processing is performed to eliminate the influence of different dimensions, so as to obtain the normalized hybrid power assembly state signal.

[0115] Step 2: Based on the unified-dimensional state signal input from the signal acquisition and processing module, and combined with the fault type and actual vehicle sensor signals, select some state signals from the engine, battery, and vehicle in the hybrid system as feature signals. The selected feature signals include diesel engine intake pressure, diesel engine intake temperature, diesel engine exhaust temperature, accelerator pedal position, battery bus voltage, battery bus current, battery temperature, and vehicle speed difference. Principal component analysis is used to determine the feature signal weights, and the hybrid system health status based on Grey Relational Analysis (GRA) is calculated in real time. A counting-based fault confirmation method is used to confirm the fault, reducing false positives and false negatives caused by errors and noise in the feature signals, thus obtaining a fault confirmation signal.

[0116] Step 3: A health management strategy for power redistribution under system faults based on LSTM-MPC. The hybrid powertrain signal obtained in Step 1 is received as a predictive maintenance indicator. After receiving the fault confirmation signal obtained in Step 2, a partial prediction model for MPC is obtained through LSTM training. Based on the LSTM-MPC prediction model, changes in partial system states within a finite time domain are predicted. The health of the hybrid power system is then predictively maintained based on the predicted changes in partial hybrid system states. When a fault occurs in the hybrid power system, the conversion relationship between energy consumption and fuel consumption is given. Battery energy consumption is equated to fuel energy consumption, and the comprehensive energy consumption is calculated. By sacrificing comprehensive energy consumption, the target speed following effect is improved, while ensuring that the battery operates within its optimal temperature range and the diesel engine does not exceed the exhaust temperature limit. This ensures that the hybrid power system can still operate with the fault under constrained conditions.

[0117] The power redistribution health management strategy under system faults based on LSTM-MPC has the following rules for the MPC control strategy:

[0118] At the current time k, the LSTM-based prediction model obtains the future time domain k+t by acquiring the current k-time input of the controlled object and information from historical times. p The optimal control sequence is obtained by retaining only the first control variable u of the optimal control sequence. p (k+1) and apply it to the controlled object, ignoring all other control variables, and repeat the above optimization process at the next sampling time. At time k, the optimization objective is represented by equation (1):

[0119]

[0120] In equation (1), J k For the cost function, t p Let x(t) be the prediction duration, x(t) be the state variable at time t, and u(t) be the control variable at time t.

[0121] The constraint is expressed by formula (2):

[0122]

[0123] In formula (2), x min (t) and x max (t) are the upper and lower limits of the state variable at time t, respectively, u min (t) and u max (t) are the upper and lower limits of the control variable at time t, respectively.

[0124] According to the multi-power source characteristics of the hybrid power system, the optimization objective is to minimize the speed following error by sacrificing the comprehensive energy consumption. The constraint conditions include the diesel exhaust temperature constraint, the battery temperature constraint, and the theoretical physical constraints of other components. At time k, in the prediction time domain, the cost function of the hybrid power system is represented by formula (3):

[0125]

[0126] In formula (3), t p is the prediction length, k p is the weight coefficient of the speed difference, k e is the weight coefficient of the comprehensive energy consumption, u(t) is the control variable, which is the accelerator pedal position correction coefficient, v a (u(t), t) and v t (t) are the actual speed and target speed, respectively, km / h. is the comprehensive energy consumption, g / s.

[0127] The constraint condition is represented by formula (4):

[0128]

[0129] In formula (4), T e is the diesel exhaust temperature, T e_max is the diesel exhaust temperature limit, which is derived from the characteristic index provided by the diesel engine manufacturer; T b is the battery temperature, [T b_smin , T b_smax ] is the optimal working temperature range of the lithium ion battery, with a value of [15, 35], ℃; SOC min , SOC max are the upper and lower limits of the power battery charge and discharge SOC, respectively, taking 0.36 and 0.6; P dem , P e , and P m are the total demand power, diesel power, and motor power, respectively, kW.

[0130] The input and output of the LSTM prediction model are selected according to common application scenarios of the hybrid power system. The input quantity is the health state evaluation result, vehicle operation mode, accelerator pedal position, battery power, actual vehicle speed, battery temperature and diesel engine exhaust temperature. The output quantity is the vehicle speed, diesel engine exhaust temperature and battery temperature of the hybrid power system. The vehicle speed, battery temperature and diesel engine exhaust temperature at future time are predicted by rolling the input information of historical time period and current time. The first 85% of offline data is selected as the training set, and the last 15% is selected as the test set for training. The mean absolute error (MAE) is selected as the loss function for optimizing the parameters of the prediction model and measuring the accuracy of the prediction model. The MAE can be expressed by equation (5):

[0131]

[0132] In equation (5), Y is the true value, f(x) is the predicted value, and n is the number of samples.

[0133] As a preferred, according to the conversion relationship between electric energy consumption and fuel consumption given in GB / T 19754-2021 “Heavy Hybrid Electric Vehicle Energy Consumption Test Method”, the battery energy consumption is equivalent to fuel energy consumption, the comprehensive energy consumption is calculated, the target vehicle speed following effect is improved by sacrificing the comprehensive energy consumption, at the same time, the battery works in its best temperature range and the diesel engine does not exceed the exhaust temperature limit, which ensures that the hybrid power system can still operate with faults under the constraint conditions.

[0134] The above specific description further describes the purpose, technical scheme and beneficial effects of the application. It should be understood that the above description is only a specific embodiment of the application and does not limit the protection scope of the application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the application should be included in the protection scope of the application.

Claims

1. A method of hybrid system health management, characterized by: Comprising the following steps, Step one, through the measured signal acquisition and processing module to obtain diesel engine instantaneous analog signal, battery instantaneous analog signal, vehicle instantaneous analog signal, and get the corresponding hybrid power system state signal, the state signal of the hybrid power system is normalized, and the normalized hybrid power system diesel engine instantaneous state signal, battery instantaneous state signal and vehicle instantaneous state signal are obtained; Step two, according to the uniform dimension state signal input by the measured signal acquisition and processing module, combined with the fault form and the real vehicle sensor signal, select part of the state signal of the engine, battery and vehicle in the hybrid power system as the characteristic signal, the selected characteristic signal includes diesel engine intake pressure, diesel engine intake temperature, diesel engine exhaust temperature, accelerator pedal position, battery bus voltage, battery bus current, battery temperature and vehicle speed difference; The weight of the characteristic signal is determined by principal component analysis, and the health degree of the hybrid power system based on gray correlation analysis method GRA is calculated in real time; The fault is confirmed by the fault confirmation method based on counting, the fault misjudgment and omission caused by the error and noise of the characteristic signal are reduced, and the fault confirmation signal is obtained; Step three, the health management strategy of power redistribution under system fault based on LSTM-MPC, receiving the hybrid power system signal obtained in step one as the predictive maintenance index, after receiving the fault confirmation signal obtained in step two, the partial prediction model for MPC is obtained through LSTM training, the change of part of the state of the system in the limited time domain is predicted based on the LSTM-MPC prediction model, and the predictive maintenance of the health degree of the hybrid power system is carried out combined with the prediction result of the change of part of the state of the hybrid power system. When the hybrid power system fails, according to the conversion relationship of the given electric power consumption and fuel consumption, the battery energy consumption is equivalent to the fuel energy consumption, the comprehensive energy consumption is calculated, the target vehicle speed following effect is improved by sacrificing the comprehensive energy consumption, and the battery is ensured to work in its best temperature range and the diesel engine is ensured not to exceed the exhaust temperature limit, so that the hybrid power system can still operate with fault under the condition of meeting the constraints.

2. The method of claim 1, wherein: Step one is implemented by, Step 1.1: obtaining the original instantaneous analog signal of the vehicle components through the measured signal acquisition and processing module; Step 1.2: signal conversion and calibration are performed on the original instantaneous analog signal obtained in step 1.1 to obtain the relationship between the original analog signal and the original digital signal, and filtering is performed to obtain the instantaneous state signal of the vehicle; Step 1.3: based on the original instantaneous analog signal obtained in step 1.1 and the instantaneous state signal obtained in step 1.2, the hybrid power system state signal is obtained, which includes the following three types of signals: diesel engine instantaneous state signal, battery instantaneous state signal and vehicle instantaneous state signal; the diesel engine instantaneous state signal includes intake temperature, intake pressure, exhaust temperature, diesel engine speed and accelerator pedal position; the battery instantaneous state signal includes SOC, battery bus voltage, battery bus current and battery temperature; the vehicle instantaneous state signal includes target vehicle speed, actual vehicle speed and vehicle running mode; Step 1.4: Signal processing is performed on the hybrid power assembly state signal obtained in step 1.3, and the hybrid power assembly state signal is normalized to eliminate the influence of different dimensions, thereby obtaining a normalized hybrid power assembly state signal.

3. The method of claim 2, wherein: In step three, The rule of the MPC control strategy of the system fault-based power redistribution health management strategy based on LSTM-MPC is as follows: At the current time k, the prediction model based on LSTM obtains the optimal control sequence of the future time domain k+t by obtaining the current k time input and historical time information of the controlled object p , only retains the first control amount u p (k+1) of the optimal control sequence and applies it to the controlled object, and all the remaining control amounts are ignored, and the above optimization process is repeated at the next sampling time; at the k time, the optimization objective is represented by formula (1): In formula (1), J k is a cost function, t p is a prediction time length, x(t) is a state quantity at time t, and u(t) is a control quantity at time t. The constraint condition is represented by formula (2): In formula (2), x min (t) and x max (t) are respectively upper and lower limits of the state variable at time t, u min (t) and u max (t) are respectively upper and lower limits of the control variable at time t; According to the characteristics of the multi-power source of the hybrid power system, the optimization objective is determined to be to minimize the vehicle speed following error by sacrificing the comprehensive energy consumption; the constraint conditions include diesel engine exhaust temperature constraints, battery temperature constraints, and other theoretical physical constraints of various components; at time k, in the prediction time domain, the cost function of the hybrid power system is represented by formula (3): In formula (3), t p is a prediction time length, k p is a weight coefficient of the vehicle speed difference, k e is a weight coefficient of the comprehensive energy consumption, u(t) is a control amount, which is a throttle pedal position correction coefficient here, v a (u(t), t) and v t (t) are actual and target vehicle speeds, respectively, km / h; is a comprehensive energy consumption, g / s; The constraint condition is represented by formula (4): In formula (4), T e is the diesel engine exhaust temperature, T e_max is the diesel engine exhaust temperature limit value; T b is the battery temperature, [T b_smin ,T b_smax ] is the optimal working temperature range of lithium-ion battery, taking the value [15, 35], ℃; SOC min , SOC max are the upper and lower limits of the SOC of the power battery during charging and discharging, respectively; P dem , P e and P m are the total demand power, diesel engine power and motor power, respectively, kW; The input of the LSTM prediction model is the health state evaluation result, the vehicle operating mode, the accelerator pedal position, the battery power, the actual vehicle speed, the battery temperature, and the diesel engine exhaust temperature; the output is the vehicle speed, the diesel engine exhaust temperature, and the battery temperature of the hybrid system; the vehicle speed, the battery temperature, and the diesel engine exhaust temperature at future time points are predicted by rolling the input information of the historical time period and the current time point; the mean absolute error MAE is selected as the loss function for optimizing the parameters of the prediction model and measuring the accuracy of the prediction model; the MAE is represented by formula (5): In formula (5), Y is the true value, f(x) is the predicted value, and n is the number of samples.

4. A hybrid powertrain health management system for implementing a method of hybrid powertrain health management as claimed in claim 1, 2 or 3, characterized by: The system mainly comprises a host computer PC, a hybrid power assembly, a vehicle controller VCU, a signal acquisition and processing module, a health state evaluation module, and a power distribution correction module based on LSTM-MPC; the host computer PC and the vehicle controller VCU are connected through signal lines; the hybrid power assembly is connected to the vehicle controller VCU through a signal bus; the host computer PC controls the vehicle controller VCU to receive and process signals from the hybrid power assembly, performs real-time signal transmission interaction, evaluates the health state of the hybrid power system based on the signal acquisition and processing module, the health state evaluation module, and the power distribution correction module based on LSTM-MPC written into the vehicle controller VCU, and corrects the power distribution of the hybrid power system according to the health state evaluation result; The signal acquisition and processing module receives and processes the hybrid power system state signal, which includes three types of signals, namely diesel engine characteristic signals, battery characteristic signals, and vehicle characteristic signals; the diesel engine characteristic signals include intake temperature, intake pressure, exhaust temperature, diesel engine speed, and accelerator pedal position; the battery characteristic signals include SOC, battery bus voltage, battery bus current, and battery temperature; The vehicle characteristic signals include target vehicle speed, actual vehicle speed, and vehicle operating mode; after acquiring the signals, the state signals are normalized to eliminate the influence of different dimensions, thereby preprocessing the health state evaluation module; by unifying the dimensions of the state signals, the accuracy of the characteristic signals at different orders of magnitude is ensured, and the efficiency of the health state evaluation of the hybrid power system is improved. According to formula (6), the time series data including intake air pressure, intake air temperature, exhaust temperature, accelerator pedal position, bus current, bus voltage, battery temperature and vehicle speed are normalized: Wherein, i=1, 2, 3, …, n, k=1, 2, 3, …, m, n is the number of samples, and m is the number of characteristics; The health state evaluation module obtains the state signal from the signal acquisition and processing module, calculates the health degree through the vehicle controller VCU according to the health degree, judges the health state of the hybrid power assembly, analyzes the hybrid power working condition and diagnoses whether the system fails; The health degree is constructed according to the normalized characteristic signals, that is, according to the intake air pressure, intake air temperature, exhaust temperature, accelerator pedal position, bus current, bus voltage, battery temperature and vehicle speed, and the constructed health degree is embedded into the health state evaluation module; The health degree construction method is as follows: The diesel-electric hybrid system has the characteristics of high coupling, nonlinearity and multiple power sources, and is comprehensively judged by multiple dimensions and multiple standards; combined with the multi-dimensional state signals collected by the diesel-electric hybrid system, the correlation degree between the running state and the health state of the hybrid system is calculated by using gray correlation analysis; the correlation degree is used to reflect the deviation degree of the running state and the health state of the hybrid system, and the greater the deviation degree, the smaller the correlation degree; The health degree is determined based on the following method: 1) The normalized data output by the signal acquisition and processing module to be tested is used to calculate the gray correlation coefficient one by one; In formula (7), k=1, 2, 3, …, m, i=1, 2, 3, …, n, and p represents the discrimination coefficient, p∈(0, 1), the smaller p is, the greater the difference between the correlation coefficients is, and the stronger the resolution ability is; 2) The correlation degree is calculated one by one according to formula (8); In formula (8), ω i (k) represents a weight coefficient of the characteristic signal, k = 1, 2, 3, …, m, i = 1, 2, 3, …, n. The hybrid system health state evaluation based on the gray relational analysis method GRA needs to calculate the correlation degree; because the sensitivity of different characteristic signals in the hybrid system to faults is different, directly taking the average value may lead to inaccurate correlation degree calculation; the weight values of different dimensional characteristic signals are determined by using the principal component analysis method; The power distribution correction module based on LSTM-MPC obtains the output and the signal to be tested from the health state evaluation module, obtains the hybrid power system state from the health state evaluation module, calculates the power redistribution health management method based on LSTM-MPC through the vehicle controller VCU, and calculates the accelerator pedal position correction coefficient in advance; when the fault is confirmed, the correction coefficient is output to correct the accelerator pedal position of the vehicle controller, that is, under the condition of hybrid power system failure, the target vehicle speed following effect is improved by sacrificing the comprehensive energy consumption, the battery is ensured to work in its best temperature range, and the diesel engine is ensured not to exceed the exhaust temperature limit, so that the hybrid power system can still operate with faults under the condition of meeting the constraints.

5. A hybrid system health management system as recited in claim 4 wherein: P is selected as 0.5.

Images