Overcharge and Overdischarge Protection Control Method for New Energy Batteries

Abstract

The present invention discloses a method for overcharge and over-discharge protection control of new energy batteries, which specifically relates to the field of electrical engineering and is used to solve the problem of aging of relay contacts of new energy batteries. By collecting contact signals, extracting dual features of recursive modal entropy change and fractal chaos diffusion, and analyzing and quantifying the electrical-mechanical coupling degradation effect of contacts; dynamically adjusting the current limit value and the relay action interval based on the aging rate, and using the gradient descent algorithm to gradually shrink the current margin to achieve real-time matching between the software strategy and the contact state; through the dynamic weight distribution of the main and standby relays and the progressive load transfer mechanism, combining life prediction to adaptively switch redundant nodes and synchronously calibrate parameters to avoid the misoperation risk of traditional threshold protection; when the load deviates from the safe range, trigger hardware switching and reset the strategy to ensure the stable operation of the system during the full life cycle of the contacts, improve the life and reliability of new energy battery relays in frequent charge and discharge scenarios, and reduce the risk of thermal runaway.

Description

Technical Field

[0001] The present invention relates to the field of electrical engineering, and more specifically, to a method for protecting new energy batteries against overcharging and over-discharging. Background Art

[0002] Under frequent start-stop or energy recovery conditions of electric vehicles, the battery management system (BMS) dynamically adjusts the charging and discharging power thresholds through software to avoid overcharging / over-discharging, and relies on hardware redundancy protection such as relays as the ultimate defense. When there is a short-term large current fluctuation in the battery (such as emergency braking after rapid acceleration), the BMS frequently triggers the software power limit strategy, resulting in the main relay being turned on and off multiple times within a few seconds; and for the hardware protection to reduce the risk of false triggering, a conservative disconnection threshold is often set. In this scenario, the mechanical contacts of the relay are oxidized due to frequent actions, the contact resistance increases, further causing abnormal temperature rise; at the same time, the software is forced to reduce the dynamic adjustment range to avoid hardware failure, ultimately resulting in the compression of the available power of the battery and the decline of the vehicle's dynamic performance.

[0003] The existing cooperative mechanism of hardware redundancy protection and software dynamic adjustment is based on static priorities and fixed thresholds, and does not establish a dynamic mapping relationship between the hardware state (such as relay life, contact resistance) and software protection parameters (such as power limit threshold, response speed). The impact of hardware degradation on software strategies has not been quantitatively modeled. For example, the oxidation of relay contacts leads to an increase in on-off delay, but the software still uses the initial response time parameter, which may cause asynchronous protection actions; conversely, when the software overly limits power to reduce the hardware load, it will cause waste of battery performance. The core contradiction lies in the uncontrollable coupling interference between the attenuation of hardware reliability and the dynamic adjustment requirements of software strategies, and the existing system lacks a cross-layer joint optimization model and cannot achieve the global optimal balance of life-safety-performance.

[0004] To solve the above problems, a technical solution is provided now. Summary of the Invention

[0005] To overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a new energy battery overcharge and over-discharge protection control method. By collecting contact signals, extracting dual features of recursive modal entropy change and fractal chaos diffusion, and analyzing and quantifying the electrical-mechanical coupling degradation effect of the contacts; based on the aging rate, dynamically adjusting the current limit value and the relay action interval, and using the gradient descent algorithm to gradually shrink the current margin, so as to achieve real-time matching between the software strategy and the contact state; through the dynamic weight distribution of the main and standby relays and the progressive load transfer mechanism, combining life prediction to adaptively switch redundant nodes and synchronously calibrate parameters, avoiding the misoperation risk of traditional threshold protection; when the load deviates from the safe range, triggering hardware switching and resetting the strategy to ensure the stable operation of the system within the full life cycle of the contacts, improving the life and reliability of the new energy battery relay in frequent charge and discharge scenarios, and reducing the risk of thermal runaway, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A new energy battery overcharge and over-discharge protection control method, comprising the steps of:

[0008] Step 1: Dynamically collect the relay contact voltage and current waveforms through high-frequency pulses, and construct time-series data including the volatility of the contact resistance and the on-off time interval.

[0009] Step 2: Extract key features from the time-series data, input them into a pre-trained degradation evaluation model, and output the aging rate index and the remaining life prediction value of the contacts.

[0010] Step 3: Dynamically adjust the current upper limit threshold of the software protection layer according to the aging rate index. When the index exceeds the safety threshold, use the gradient descent algorithm to gradually shrink the current margin and extend the minimum time window of the relay action interval.

[0011] Step 4: Combine the remaining life prediction value with the real-time load demand to generate a dynamic weight coefficient for the coordinated operation of the main / standby relays. When the life of the main relay is lower than the critical value, automatically increase the load weight of the backup relay and synchronously calibrate the software current limiting parameters.

[0012] Step 5: Based on the real-time load weight ratio of the main / standby relays, preferentially limit the peak value of the main circuit current at the software layer. If the load ratio continuously deviates from the set interval, trigger a hardware switching instruction and reset the protection strategy parameters.

[0013] In a preferred embodiment, Step 1 includes the following content:

[0014] When the main circuit of the relay is in a non-operating state, inject high-frequency square wave pulses with a predetermined amplitude at both ends of the contacts. The pulse frequency is set within a predetermined frequency band range, and synchronously collect the voltage waveform and the pulse current waveform at both ends of the contacts.

[0015] Denoise the original voltage and current signals, calculate the dynamic contact resistance of the contacts based on the denoised signals, select a continuous time period in the pulse steady state stage, calculate the average value of the resistance in the corresponding interval as the contact resistance reference value, and statistically calculate the standard deviation of the resistance values in the corresponding interval as the volatility index for quantifying the surface state of the contacts;

[0016] Record the action timestamps of each contact closing and opening, and calculate the time interval between adjacent closing actions; if the contact is in a continuous closed state, extract the action time intervals of adjacent complete on-off cycles. For the time interval sequence, calculate the ratio of its standard deviation to the mean value to generate a coefficient of variation index reflecting the discrete degree of the mechanical action of the relay.

[0017] In a preferred embodiment, step two includes the following content:

[0018] Extract key features from the time series data, where the key features include the recursive modal entropy change index and the fractal chaos diffusion index.

[0019] In a preferred embodiment, the calculation logic of the recursive modal entropy change index is as follows:

[0020] Integrate the standard deviation of the contact resistance fluctuation and the coefficient of variation time window of the time interval to construct a two-dimensional recurrence matrix , representing the dynamic similarity of the contact state in the parameter space:

[0021] ;

[0022] where is the step function. When the Euclidean distance between two points and is less than the dynamic neighborhood radius , , otherwise it is 0; and are the index numbers of different time points in the time series data window, , is the total number of data points in the window; based on the diagonal length distribution of the recurrence matrix, calculate the temperature and load corrected recursive modal entropy change index :

[0023] ;

[0024] where represents the maximum value of the diagonal length in the recurrence graph, used to limit the diagonal range; is the minimum diagonal length; and are the real-time load current of the main circuit and the nominal value of the rated current of the relay, respectively; and are the ambient temperature and the reference temperature, respectively; is a variable of the diagonal length and a statistic of the diagonal length in the recursive matrix; is an exponential parameter, and its specific value needs to be determined by fitting experimental data to accurately describe the influence of environmental factors on the degradation process.

[0025] In a preferred embodiment, the calculation logic of the fractal chaotic diffusion index is as follows:

[0026] Map the standard deviation sequence of the contact resistance fluctuation within the time window to a phase space trajectory and calculate the correlation dimension to quantify the self-similarity of the sequence:

[0027] ;

[0028] where the correlation integral is the proportion of point pairs with a distance less than in the statistical phase space; the scaling radius in the correlation integral; Combine the transient change of the time interval coefficient of variation and the temperature influence to generate the fractal chaotic diffusion index :

[0029] ;

[0030] where represents the number of differential calculations of the time interval coefficient of variation; is the index number of the differential calculation of the time interval coefficient of variation, , representing the th differential operation; is the temperature coupling coefficient; represents the time interval coefficient of variation after the th differential; represents the time interval coefficient of variation after the th differential; represents the average value of the time interval coefficient of variation; is the hyperbolic tangent function.

[0031] In a preferred embodiment, input the key features into the pre-trained degradation evaluation model to output the aging rate index and the remaining life prediction value of the contact. The specific steps are as follows:

[0032] Input the recursive modal entropy change index and the fractal chaotic diffusion index into the pre-trained spatio-temporal attention network to output the predicted aging rate and the remaining life prediction value ; Correct the prediction deviation of the data-driven model through the Arrhenius equation to improve engineering applicability:

[0033] ;

[0034] where represents the aging rate exponent; is the dynamic gain coefficient; represents the change rate of the standard deviation of resistance fluctuation over time; represents the activation energy; is the Boltzmann constant, a physical constant.

[0035] In a preferred embodiment, step three includes the following:

[0036] S3.1, The initial current upper limit threshold is set based on a certain proportion of the rated current as the base value; subsequently, the actual current limit is determined by subtracting a dynamically adjusted amount, and the adjustment amount is determined by the relative deviation between the aging rate exponent and the safety threshold. When the aging rate exponent exceeds the safety threshold, the adjustment amount is calculated according to the preset current margin multiplied by the deviation ratio; the new current limit is based on the current limit and is corrected by the attenuation coefficient and the square term of the ratio of the aging rate exponent to the safety threshold;

[0037] S3.2, Use the gradient descent algorithm to optimize the current margin, balance degradation suppression and system stability. Based on the gradient, the current margin is gradually updated by the step size controlled by the learning rate.

[0038] In a preferred embodiment, S3.3, The minimum allowable action interval is dynamically adjusted based on the base interval by the aging rate exponent. The adjustment term is determined by multiplying the expansion coefficient by the ratio of the aging rate exponent to the safety threshold and then multiplying by the current action interval; to avoid frequent switching of the action interval, a hysteresis interval is introduced. From the minimum allowable interval to a slightly larger tolerance range, if the actual action interval is less than the minimum allowable interval, the next interval is adjusted to the minimum interval plus the tolerance value; otherwise, it remains unchanged.

[0039] In a preferred embodiment, step four includes the following:

[0040] Normalize the predicted remaining life value of the main relay to the main relay weight coefficient , when the predicted remaining life value approaches the preset life critical value, the weight coefficient decays according to the inverse proportional function, and the standby relay weight increases synchronously; According to the real-time load power , calculate the actual power distribution of the main / standby relays:

[0041] ;

[0042] Set the lower threshold of the weight coefficient. When the weight coefficient is lower than the lower threshold of the weight coefficient, force the trigger of the main relay offline command; when the predicted remaining life value of the main relay does not exceed the life critical value, send a soft shutdown command to the main relay, gradually transfer the load to the standby relay, and the transfer rate is controlled by the load power change rate threshold; based on the historical state data of the standby relay, update its software current limit threshold , calculation formula:

[0043] ;

[0044] where is the cumulative action times of the standby relay, is the rated life times, is the attenuation coefficient; after the switching is completed, reset the life prediction model of the main relay, mark the standby relay as the new main relay, and initialize the monitoring parameters of the new standby node.

[0045] In a preferred embodiment, step five includes the following content:

[0046] Obtain the load weight ratio of the main / standby relay in real time, and calculate the maximum allowable current peak value of the main circuit according to the weight coefficient of the main relay in real time , the calculation formula is , where is the rated current of the relay; set the allowable interval of the load ratio. If it is detected that the weight coefficient of the main relay exceeds the interval for multiple consecutive control cycles, it is determined as continuous deviation;

[0047] When the load ratio deviates continuously, send a progressive shutdown command to the main relay and a pre-charge command to the standby relay at the same time; use a ramp function to control the load weights of the main and standby relays, linearly reduce the weight coefficient of the main relay from the current value to zero within a fixed time window, and the weight of the standby relay rises synchronously to 1; after the switching is completed, mark the main relay as standby, reset the parameters of its life prediction model, and re-initialize the maximum current peak value based on the historical state data of the standby relay.

[0048] The technical effects and advantages of the new energy battery overcharge and over-discharge protection control method of the present invention:

[0049] The present invention dynamically collects contact electrical signals through high-frequency pulses and constructs long-term degradation time-series data. By combining the dual feature extraction of recursive modal entropy change and fractal chaos diffusion, it accurately quantifies the electrical-mechanical coupling degradation effect of the contacts. Based on the aging rate index, it dynamically adjusts the current limit value and the relay action interval, and uses the gradient descent algorithm to gradually shrink the current margin and extend the mechanical operation window, realizing the real-time matching of the software protection strategy and the contact degradation state. Through the dynamic weight allocation of the primary and backup relays and the progressive load transfer mechanism, combined with the remaining life prediction value, it adaptively switches the hardware redundant nodes and synchronously calibrates the protection parameters, avoiding the risks of misoperation or refusal to operate caused by traditional fixed-threshold protection. Finally, when the load ratio continuously deviates, it triggers hardware switching and resets the policy parameters to ensure the stable operation of the system throughout the full cycle of contact degradation. This solution solves the technical defects such as the disconnection between hardware redundancy and software strategy, insufficient adaptability of static thresholds, and lag in local overcharge / overdischarge identification. Through data-physics joint drive and cross-layer collaborative optimization, it significantly improves the life and reliability of new energy battery relays in frequent charge and discharge scenarios and reduces the risk of thermal runaway. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic structural diagram of the overcharge and overdischarge protection control method for new energy batteries of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0052] Embodiment 1: Figure 1 The overcharge and overdischarge protection control method for new energy batteries of the present invention is given, including:

[0053] Step 1: Dynamically collect the voltage and current waveforms of the relay contacts through high-frequency pulses, and construct time-series data including the volatility of the contact resistance and the on-off time interval.

[0054] Step 2: Extract key features from the time-series data, input them into a pre-trained degradation evaluation model, and output the aging rate index and the remaining life prediction value of the contacts.

[0055] Step 3: Dynamically adjust the current upper limit threshold of the software protection layer according to the aging rate index. When the index exceeds the safety threshold, use the gradient descent algorithm to gradually shrink the current margin and extend the minimum time window of the relay action interval.

[0056] Step 4: Combine the remaining life prediction value with the real-time load demand to generate a dynamic weight coefficient for the coordinated operation of the main / backup relays. When the life of the main relay is lower than the critical value, automatically increase the load weight of the backup relay and synchronously calibrate the software current limiting parameters.

[0057] Step 5: Based on the real-time load weight ratio of the main / backup relays, preferentially limit the peak current of the main circuit at the software layer. If the load ratio continuously deviates from the set range, trigger a hardware switching instruction and reset the protection strategy parameters.

[0058] In the existing relay protection mechanism, the fragmentation between the hardware redundancy design and the software dynamic adjustment leads to uncontrollable losses in system life and performance. Especially in the scenario of frequent on / off operations, the vicious cycle of accelerated contact degradation and lagging protection strategy is formed. By dynamically monitoring the contact state through high-frequency pulses and constructing a long-term degradation database, the static limitation of traditional threshold protection can be broken through, providing high-precision degradation trajectory input for the coordinated optimization of hardware-software, realizing the closed-loop linkage between contact life prediction and protection strategy, and promoting the paradigm upgrade of relay protection from passive response to active prevention.

[0059] Step 1 includes the following contents:

[0060] When the main circuit of the relay is in the non-operating state, inject high-frequency square-wave pulses with a predetermined amplitude across the contacts. The pulse frequency is set within a predetermined frequency band range, and the duration of a single pulse is controlled to be much less than the magnitude of the main circuit operating cycle. The electrical isolation between the pulse signal and the main circuit load current is achieved through an isolation coupling circuit. Use high-speed voltage sensors and precision current sensors to synchronously collect the voltage waveform across the contacts and the pulse current waveform. The sampling rate is set to a high-speed mode that meets the high-frequency signal capture requirements to ensure complete coverage of the transient processes of the rising edge, steady state, and falling edge of the pulse.

[0061] Perform wavelet threshold denoising processing on the original voltage and current signals. Select a preset wavelet basis function to decompose the signal at multiple levels, and eliminate high-frequency noise and power frequency interference through threshold filtering. Calculate the dynamic contact resistance of the contacts based on the denoised signals. Select a continuous time period in the steady state stage of the pulse, calculate the average value of the resistance in this interval as the contact resistance reference value, and calculate the standard deviation of the resistance values in this interval as a volatility index for quantifying the surface state of the contacts.

[0062] Accurately record the action timestamps of each contact closure and disconnection, and calculate the time interval between adjacent closure actions; if the contact is in a continuous closed state, extract the time interval between adjacent actions that include a complete on / off cycle. For the time interval sequence, calculate the ratio of its standard deviation to the mean value to generate a coefficient of variation index that reflects the discrete degree of the relay mechanical actions, and is used to evaluate the fatigue cumulative effect of the mechanical structure.

[0063] The data collected each time is stored as a structured data unit indexed by timestamp, including timestamp, reference value of contact resistance, standard deviation of resistance fluctuation, on-off time interval, coefficient of variation of time interval, ambient temperature, and main circuit load current. The ambient temperature is collected by a temperature sensor, and the load current is monitored in real time by a current sensor. The long-term database organizes data using a sliding window mechanism. Each window contains a set of continuous on-off action data units of a certain length, and the window step advances according to a preset rule, forming a time-series data chain covering the complete degradation cycle of the contact, realizing the dynamic associated storage of multi-dimensional parameters.

[0064] For the dynamic acquisition of contact status based on high-frequency pulse excitation and the construction of a multi-dimensional time-series database, the volatility of contact resistance and the divergence characteristics of on-off intervals are obtained through non-invasive detection, and the ambient temperature and load condition parameters are fused to establish a holographic portrait of the contact degradation process; this step provides a quantifiable and traceable data basis for subsequent degradation rate modeling and dynamic correction of protection strategies, solves the decoupling problem of the hidden attenuation of hardware status and the static setting of software protection parameters, and is the core support for realizing the global optimization of life-safety-performance.

[0065] Step two includes the following:

[0066] Extract key features from the time-series data, where the key features include the recursive modal entropy change index and the fractal chaos diffusion index; the recursive modal entropy change index quantifies the co-degradation effect of electrical characteristics and mechanical actions, capturing the interaction of oxidation, arc erosion, and mechanical fatigue, and the fractal chaos diffusion index is used to evaluate the chaos propagation intensity of contact degradation from local damage to global diffusion.

[0067] The calculation logic of the recursive modal entropy change index is as follows:

[0068] Through the standard deviation of contact resistance fluctuation and the coefficient of variation of time interval The joint analysis quantifies the non-linear coupling effect of contact electro-mechanical degradation and captures the interaction of oxidation, arc erosion, and mechanical fatigue.

[0069] Integrate the standard deviation of contact resistance fluctuation and the time window of the coefficient of variation of time interval to construct a two-dimensional recursive matrix , characterizing the dynamic similarity of the contact state in the parameter space:

[0070] ;

[0071] where is a step function. When the Euclidean distance between two points and is less than the dynamic neighborhood radius , , otherwise it is 0; the dynamic neighborhood radius is adaptively adjusted according to the range of the standard deviation of the resistance fluctuation to ensure sensitive detection in the high-fluctuation range; and are the index numbers of different time points in the time series data window, , is the total number of data points in the window.

[0072] Based on the diagonal length distribution of the recurrence matrix , calculate the recurrence mode entropy change index corrected by temperature and load :

[0073] ;

[0074] Among them, the entropy value weight term quantifies the uncertainty of the diagonal length distribution and reflects the chaotic degree of the degradation mode; represents the maximum value of the diagonal length in the recurrence plot and is used to limit the diagonal range; is the minimum diagonal length and is used to filter short diagonals generated by instantaneous noise; and are respectively the nominal values of the real-time load current of the main circuit and the rated current of the relay; and are respectively the ambient temperature and the reference temperature; is the variable of the diagonal length and is the statistic of the diagonal length in the recurrence matrix; is an exponential parameter, and its specific value needs to be determined by fitting experimental data to accurately describe the influence of environmental factors on the degradation process.

[0075] Temperature correction term Suppresses misjudgment caused by contact hysteresis in low-temperature environments (low temperature makes the temperature correction term approach 1 and retains the original entropy value);

[0076] Load correction Amplifies the contribution of high-load conditions ( close to 1) to the entropy value and enhances the sensitivity to high-current degradation.

[0077] Through multi-physical field coupling correction, the recurrence mode entropy change index simultaneously characterizes the non-linearity, temperature dependence, and load acceleration effect of contact degradation. The larger the value of the recurrence mode entropy change index, the more significant the enhancement of the non-linear coupling effect between the electrical characteristics and mechanical actions of the contact, and the more complex non-linear dynamic behavior is presented in the degradation processes such as contact surface oxidation and arc erosion. Moreover, the synergistic acceleration effect of the load current and ambient temperature on the degradation rate is significant; a decrease in the value indicates that the electro-mechanical coupling effect of the contact degradation process slows down, the fluctuation mode tends to be stable, the influence of temperature and load on degradation weakens, and the contact is in a relatively stable early wear or late-stage stable failure stage.

[0078] The calculation logic of the fractal chaos diffusion index is as follows:

[0079] Based on the standard deviation of the contact resistance fluctuation The fractal characteristics of the sequence and the coefficient of variation of the time interval The transient changes are used to quantify the chaos diffusion intensity of the contact degradation process and identify the coupled propagation mode of surface deterioration and mechanical looseness.

[0080] Map the sequence of the standard deviation of the contact resistance fluctuation within the time window into a phase space trajectory and calculate the correlation dimension To quantify the self-similarity of the sequence:

[0081] ;

[0082] where the correlation integral is the proportion of point pairs with a distance less than in the statistical phase space, reflecting the aggregation degree of the sequence; The scaling radius in the correlation integral;

[0083] The correlation dimension : The smaller the correlation dimension, the stronger the regularity of the sequence of the standard deviation of the contact resistance fluctuation (such as the periodic fluctuation caused by oxidation), and the larger the correlation dimension, the stronger the chaos (such as the random damage caused by arc).

[0084] Distinguish the deterministic mode and the random mode of contact degradation through the fractal dimension of the standard deviation of the contact resistance fluctuation.

[0085] Combined with the transient changes and temperature effects of the coefficient of variation of the time interval Generate the fractal chaos diffusion index :

[0086] ;

[0087] where is the number of differential calculations of the coefficient of variation of the time interval; is the index number of the differential calculation of the coefficient of variation of the time interval, , indicating the th differential operation; is the temperature coupling coefficient (dimensionless), determined by fitting experimental data, and used to adjust the influence weight of temperature on the fractal chaos diffusion index; represents the coefficient of variation of the time interval after the th differential; represents the coefficient of variation of the time interval after the th differential; represents the average value of the coefficient of variation of the time interval; is the hyperbolic tangent function.

[0088] Transient difference term: Quantify the mutation frequency and amplitude of the coefficient of variation sequence of the quantization time interval (such as the sudden change of the interval caused by mechanical jamming);

[0089] Temperature coupling term: Balance the accelerated diffusion at high temperature and the diffusion inhibition at low temperature.

[0090] Characterize the evolution intensity of the contact degradation from local damage to global diffusion through the product relationship between the fractal dimension and the mechanical mutation (the difference of the coefficient of variation of the time interval).

[0091] An increase in the value of the fractal chaos diffusion index indicates that the contact degradation process has strong chaotic characteristics and a fast diffusion trend, with frequent and significant mutations in the mechanical action interval, surface damage spreading from local pitting to the global, and material fatigue and lubrication failure entering an accelerated deterioration period; a decrease in the value indicates a reduction in the chaos and diffusivity of the degradation process, the mechanical action interval tends to be regular, the surface damage is limited to a local area and no chain expansion occurs, and the contact is in a state of slow and uniform wear or damage inhibition.

[0092] Input the key features into the pre-trained degradation evaluation model to output the aging rate index and the predicted remaining life value of the contact. The specific steps are as follows:

[0093] Input the recursive modal entropy change index and the fractal chaos diffusion index into the pre-trained spatio-temporal attention network to output the predicted aging rate and the predicted remaining life value :

[0094] ;

[0095] where refers to the degradation evaluation model architecture based on such as the spatio-temporal attention network; represents the parameter set of the pre-trained degradation evaluation model, including neural network weights, biases, etc., obtained by training with laboratory accelerated aging data.

[0096] Model architecture: The temporal convolutional layer extracts local degradation features, the attention mechanism focuses on key degradation stages, and the fully connected layer maps to the life prediction value;

[0097] Pre-trained data: Trained based on laboratory accelerated aging data (such as high-temperature high-load cycling, low-temperature intermittent testing, etc.), covering the degradation path of the contact throughout its life cycle.

[0098] Correct the prediction bias of the data-driven model through the Arrhenius equation to improve the engineering applicability:

[0099] ;

[0100] wherein represents the aging rate index; is the dynamic gain coefficient, which is used to adjust the contribution weight of the resistance fluctuation change rate to the aging rate; represents the change rate of the standard deviation of the resistance fluctuation over time, characterizing the transient deterioration speed of the contact surface state; represents the activation energy, the energy barrier of the material chemical reaction, which determines the influence intensity of temperature on the degradation rate; is the Boltzmann constant, a physical constant.

[0101] Dynamic gain term: Amplify the degradation rate in the acceleration stage of the resistance fluctuation (such as the arc burst period);

[0102] Arrhenius correction term: Introduce the physical constraint of temperature on the material activation energy to suppress high-temperature misjudgment.

[0103] Constrain the prediction range of the data model through physical equations to ensure that the aging rate index complies with the material failure mechanism.

[0104] is the aging rate index, which characterizes the comprehensive aging rate of the contact material after physical constraint correction. Its value quantifies the real-time speed of contact degradation under specific working conditions (temperature, load, resistance fluctuation). The larger the aging rate index, the faster the degradation processes such as chemical oxidation, arc erosion, and mechanical fatigue of the contact caused by high temperature, large current, or severe resistance fluctuation, the higher the deterioration rate of the material performance, and the significantly shorter the remaining life; the smaller the aging rate index, the slower the material damage accumulation and the longer the remaining life. This parameter realizes the multi-dimensional dynamic calibration of the degradation rate by fusing data-driven prediction of the aging rate and the Arrhenius physical equation, capturing the actual degradation trend and following the temperature dependence law of the material activation energy.

[0105] Step three includes the following:

[0106] S3.1, The core objective of dynamically adjusting the current upper limit threshold is to protect the relay contacts from overload and degradation through real-time monitoring and adjustment. The following are the specific steps:

[0107] The initial current upper limit threshold is set based on a certain percentage of the rated current as the base value. Subsequently, the actual current limit is determined by subtracting a dynamically adjusted amount. This adjustment amount is determined by the relative deviation between the aging rate index and the safety threshold. When the aging rate index exceeds the safety threshold, the adjustment amount is calculated by multiplying the preset current margin (a part of the rated current) by the deviation ratio. The purpose is to reduce the damage to the contacts caused by high current and slow down the degradation process by gradually reducing the current margin.

[0108] To avoid system instability caused by sudden changes in the current limit, an exponential decay factor is introduced in the adjustment process. The new current limit is based on the current limit and is corrected by the decay coefficient and the square term of the ratio of the aging rate index to the safety threshold. The decay coefficient is adjusted according to the system stability requirements, and the design of the square term makes the current limit shrink faster when the aging rate is higher. This way, the load is quickly reduced when the degradation is severe, and smooth operation is maintained when the change is gentle.

[0109] S3.2, Use the gradient descent algorithm to optimize the current margin and balance degradation suppression and system stability. The following are the key steps:

[0110] Define an objective function to measure the balance between the deviation of the aging rate index from the safety threshold and the adjustment amplitude. The first part promotes it to approach the safety value through the square of the deviation of the ratio of the aging rate index to the safety threshold; the second part restricts the adjustment amplitude from being too large through the product of the regularization coefficient and the square of the current margin. The regularization coefficient prevents excessive fluctuations in the contraction of the current margin and ensures a smooth adjustment process. For example, the objective function is defined as follows:

[0111] ;

[0112] where is the regularization coefficient, which suppresses excessive fluctuations in the contraction of the current margin; is the aging rate index; is the safety threshold; is the preset current margin, which is used to control the current contraction amplitude when the degradation accelerates.

[0113] The gradient of the objective function consists of two parts: one is the influence of the aging rate deviation on the current margin, which depends on the sensitivity of the aging rate index to the current margin; the other is the constraint of the regularization term on the current margin, which is directly related to the size of the current margin. Based on the gradient, the current margin is gradually updated by the step size controlled by the learning rate, and the learning rate determines the speed of adjustment. This iterative process finds the optimal point between degradation suppression and system stability through gradual optimization.

[0114] S3.3, Extending the relay operation interval aims to reduce the mechanical shock frequency, thereby slowing down the contact wear. The following is the implementation method:

[0115] The minimum allowable action interval is dynamically adjusted based on the base interval by the aging rate exponent. The adjustment term is determined by multiplying the expansion coefficient by the ratio of the aging rate exponent to the safety threshold and then multiplying by the current action interval. The expansion coefficient controls the speed of interval extension. When degradation accelerates, this method reduces mechanical wear by extending the action interval.

[0116] To avoid frequent switching of the action interval, a hysteresis interval is introduced, ranging from the minimum allowable interval to a slightly larger tolerance range. If the actual action interval is less than the minimum allowable interval, the next interval is adjusted to the minimum interval plus the tolerance value; otherwise, it remains unchanged. This mechanism prevents oscillations through the hysteresis interval and ensures the stability of relay operation.

[0117] Step four includes the following:

[0118] Remaining life mapping weight: Normalize the predicted remaining life value of the main relay to the main relay weight coefficient , when the predicted remaining life value approaches the preset life critical value, the weight coefficient decays according to the inverse proportional function, and the weight of the standby relay increases synchronously to satisfy .

[0119] Load demand adaptation: According to the real-time load power , calculate the actual allocated power of the main / standby relay:

[0120] ;

[0121] Weight boundary constraint: Set the lower threshold of the weight coefficient (such as 0.2). When the weight coefficient is lower than this value, force the main relay to receive an offline instruction.

[0122] Through the joint mapping of life and load, seamless transition of the main and standby loads is achieved, avoiding load jumps or unbalanced distribution during the switching process.

[0123] Main relay offline trigger: When the predicted remaining life value of the main relay does not exceed the life critical value, send a soft shutdown instruction to the main relay, gradually transfer the load to the standby relay, and the transfer rate is controlled by the load power change rate threshold (such as reducing the load by 10% per second).

[0124] Standby relay parameter calibration: Based on the historical state data of the standby relay (contact resistance, number of operations), update its software current limiting threshold , calculation formula:

[0125] ;

[0126] where is the cumulative number of operations of the standby relay, is the rated life times, is the attenuation coefficient.

[0127] System global parameter reset: After the switchover is completed, reset the main relay life prediction model, mark the standby relay as the new main relay, and initialize the monitoring parameters of the new standby node.

[0128] By means of progressive load transfer and parameter dynamic calibration, avoid the switchover transient risk and ensure the continuous and safe operation of the system.

[0129] Step five includes the following:

[0130] Weight ratio monitoring: Obtain the load weight ratio of the main / standby relay in real time ( : ), and calculate the maximum allowable current peak value of the main circuit in real time according to the weight coefficient of the main relay , and the calculation formula is , where is the rated current of the relay. is the rated current of the relay.

[0131] Current clamping control: In the scenario of load mutation (such as motor startup), if the instantaneous value of the main circuit current exceeds the maximum current peak value, force the current to be limited below the maximum current peak value through PWM modulation or torque command scaling.

[0132] Deviation interval determination: Set the allowable interval of the load ratio (such as ), if it is detected that the weight coefficient of the main relay exceeds the interval for multiple consecutive control cycles (such as 3 cycles), it is determined as continuous deviation.

[0133] Dynamically limit the main circuit current through the weight ratio to avoid the accelerated degradation of the main relay due to overload, and at the same time provide a buffer window for the hardware switchover.

[0134] Switching instruction generation: When the load ratio is continuously deviated, send a progressive turn-off instruction to the main relay, and at the same time send a pre-charge instruction to the standby relay to ensure no current impact during the switchover instant.

[0135] Load transfer control: Use a ramp function to control the load weights of the main and standby relays, linearly reduce the weight coefficient of the main relay from the current value to zero within a fixed time window, and the weight of the standby relay rises synchronously to 1, and the transfer rate is adaptively adjusted according to the load power (such as reducing the speed at high power).

[0136] Protection parameter reset: After the switchover is completed, mark the main relay as standby, reset the parameters of its life prediction model (such as clearing the action times statistics, realizing the seamless switchover of the main and standby relay roles through progressive weight transfer and parameter reset, and ensuring the continuous and safe operation of the system), and re-initialize the maximum current peak value based on the historical state data of the standby relay.

[0137] The present invention dynamically acquires the contact electrical signals through high-frequency pulses and constructs long-term degradation time-series data. By combining the dual feature extraction of recursive modal entropy change and fractal chaos diffusion, it accurately quantifies the electrical-mechanical coupling degradation effect of the contacts. Based on the aging rate index, it dynamically adjusts the current limit value and the relay action interval, and uses the gradient descent algorithm to gradually shrink the current margin and extend the mechanical operation window, realizing the real-time matching of the software protection strategy and the contact degradation state. Through the dynamic weight allocation of the primary and backup relays and the progressive load transfer mechanism, combined with the remaining life prediction value, it adaptively switches the hardware redundant nodes and synchronously calibrates the protection parameters, avoiding the risk of misoperation or refusal to operate caused by the traditional fixed-threshold protection. Finally, when the load ratio continuously deviates, it triggers the hardware switch and resets the strategy parameters to ensure the stable operation of the system throughout the full cycle of contact degradation. This solution solves the technical defects such as the disconnection between hardware redundancy and software strategy, the insufficient adaptability of static thresholds, and the lag in the identification of local overcharging / overdischarging. Through data-physics joint drive and cross-layer collaborative optimization, it significantly improves the lifespan and reliability of new energy battery relays in frequent charge and discharge scenarios and reduces the risk of thermal runaway.

[0138] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0139] Only some exemplary embodiments of the present invention have been described by way of illustration. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0140] It should be noted that in this article, if there are relational terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including the element.

[0141] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the claims described above.

Claims

1. A new energy battery overcharge and over-discharge protection control method, characterized in that: Includes steps: Step 1: Dynamically collect the voltage and current waveforms of the relay contacts through high-frequency pulses to construct time series data including contact resistance fluctuations and on-off time intervals; Step 2: Extract key features from time series data, input into the pre-trained degradation assessment model, and output the aging rate index and remaining life prediction value of the contact; Step 3: Dynamically adjust the current upper limit threshold of the software protection layer according to the aging rate index. When the index exceeds the safety threshold, use the gradient descent algorithm to gradually shrink the current margin and extend the minimum time window of the relay action interval. Step 4: Combine the remaining life prediction value with the real-time load demand to generate a dynamic weight coefficient for the coordinated operation of the main / backup relays. When the main relay life is lower than the critical value, the backup relay load weight is automatically increased and the software current limiting parameters are calibrated synchronously. Step 5: Based on the real-time load weight ratio of the main / backup relays, the main circuit current peak is limited preferentially at the software layer. If the load ratio continues to deviate from the set range, the hardware switching instruction is triggered and the protection strategy parameters are reset.

2. The new energy battery overcharge and over-discharge protection control method according to claim 1 is characterized in that: Step 1 includes the following: When the main circuit of the relay is in a non-operating state, a high-frequency square wave pulse of a predetermined amplitude is injected into both ends of the contact, the pulse frequency is set within a predetermined frequency band, and the voltage waveform and pulse current waveform at both ends of the contact are synchronously collected; The original voltage and current signals are denoised, and the dynamic contact resistance of the contacts is calculated based on the denoised signals. The continuous time period of the pulse steady-state phase is selected, and the average value of the resistance in the corresponding interval is calculated as the contact resistance reference value. The standard deviation of the resistance value in the corresponding interval is calculated as a volatility indicator to quantify the contact surface state. Record the timestamp of each contact closing and opening action, and calculate the time interval between adjacent closing actions; If the contact is in a continuously closed state, the adjacent action time intervals containing a complete on-off cycle are extracted. For the time interval sequence, the ratio of its standard deviation to the mean is calculated to generate a coefficient of variation index that reflects the discrete degree of the mechanical action of the relay.

3. The new energy battery overcharge and over-discharge protection control method according to claim 2 is characterized in that: Step 2 includes the following: Key features are extracted from time series data, including recursive modal entropy change index and fractal chaos diffusion index.

4. The new energy battery overcharge and over-discharge protection control method according to claim 2 is characterized in that: The calculation logic of the recursive modal entropy change index is as follows: The standard deviation of contact resistance fluctuation is integrated with the time window of coefficient of variation of time interval to construct a two-dimensional recursive matrix , characterizing the dynamic similarity of contact states in parameter space: ； in is a step function, when two points and The Euclidean distance is less than the dynamic neighborhood radius hour, , otherwise 0; and are index numbers for different time points in the time series data window. , is the total number of data points in the window; the diagonal length distribution based on the recursive matrix , calculate the temperature and load corrected recursive modal entropy change index : ； in Indicates the maximum length of the diagonal line in the recursive graph, which is used to limit the range of the diagonal line; Minimum diagonal length; and They are the real-time load current of the main circuit and the nominal value of the rated current of the relay respectively; and are ambient temperature and reference temperature respectively; is the variable of diagonal length, which is the statistic of diagonal length in the recursive matrix; It is an exponential parameter, and its specific value needs to be determined by fitting experimental data to accurately describe the impact of environmental factors on the degradation process.

5. The new energy battery overcharge and over-discharge protection control method according to claim 4 is characterized in that: The calculation logic of the fractal chaos diffusion index is as follows: Map the standard deviation sequence of contact resistance fluctuations within the time window into a phase space trajectory and calculate the correlation dimension To quantify the self-similarity of a sequence: ； The correlation integral is the distance in the statistical phase space less than Point-to-point ratio; Scaling radius in correlation integrals; coefficient of variation of combined time intervals The transient change and temperature influence of the generated fractal chaos diffusion index : ； in The number of difference calculations for the coefficient of variation of the time intervals; The index number for calculating the coefficient of variation of the time interval differences. , indicating the sub-difference operation; is the temperature coupling coefficient; Indicates coefficient of variation of the time interval after sub-difference; Indicates coefficient of variation of the time interval after sub-difference; represents the mean value of the coefficient of variation of the time interval; is the hyperbolic tangent function.

6. The new energy battery overcharge and over-discharge protection control method according to claim 5 is characterized in that: Input the key features into the pre-trained degradation assessment model and output the aging rate index and remaining life prediction value of the contact. The specific steps are as follows: The recursive modal entropy change index and fractal chaos diffusion index are input into the pre-trained spatiotemporal attention network to output the predicted aging rate. Remaining life prediction Correcting the prediction bias of data-driven models through the Arrhenius equation: ； in represents the aging rate index; Dynamic gain coefficient; It represents the rate of change of the standard deviation of resistance fluctuation over time; represents the activation energy; Boltzmann constant, a physical constant.

7. The new energy battery overcharge and over-discharge protection control method according to claim 6 is characterized in that: Step three includes the following: S3.1, the initial current upper limit threshold is set based on a certain proportion of the rated current as a base value; then, the actual current limit is determined by subtracting a dynamic adjustment amount, which is determined by the relative deviation between the aging rate index and the safety threshold. When the aging rate index exceeds the safety threshold, the adjustment amount is calculated based on the preset current margin multiplied by the deviation ratio; the new current limit is based on the current limit and is corrected by the attenuation coefficient and the square of the ratio of the aging rate index to the safety threshold; S3.2, uses the gradient descent algorithm to optimize the current margin, balancing degradation suppression and system stability. Based on the gradient, the current margin is gradually updated by the step size controlled by the learning rate.

8. The new energy battery overcharge and over-discharge protection control method according to claim 7 is characterized in that: S3.3, the minimum allowable action interval is based on the basic interval and is dynamically adjusted through the aging rate index. The adjustment term is determined by multiplying the expansion coefficient by the ratio of the aging rate index to the safety threshold, and then multiplying it by the current action interval. To avoid frequent switching of the action interval, a hysteresis interval is introduced, from the minimum allowable interval to a slightly larger tolerance range. If the actual action interval is less than the minimum allowable interval, the next interval is adjusted to the minimum interval plus the tolerance value; otherwise it remains unchanged.

9. The new energy battery overcharge and over-discharge protection control method according to claim 8, characterized in that: Step 4 includes the following: Normalize the remaining life prediction value of the main relay to the main relay weight coefficient When the remaining life prediction value approaches the preset life critical value, the weight coefficient decays according to the inverse proportional function, and the standby relay weight Synchronous increase; according to the real-time load power , calculate the actual power distribution of the main / backup relay: ； Set the lower limit threshold of the weight coefficient. When the weight coefficient is lower than the lower limit threshold, the main relay offline command is triggered forcibly. When the remaining life prediction value of the main relay does not exceed the life critical value, send a soft shutdown command to the main relay to gradually transfer the load to the backup relay. The transfer rate is controlled by the load power change rate threshold. Based on the historical status data of the backup relay, update its software current limiting threshold , calculation formula: ; in The cumulative number of actions for the standby relay. is the rated life times, is the attenuation coefficient, The reference current value in the initial or design state; After the switch is completed, the main relay life prediction model is reset, the backup relay is marked as the new main relay, and the monitoring parameters of the new backup node are initialized.

10. The new energy battery overcharge and over-discharge protection control method according to claim 9, characterized in that: Step five includes the following: Obtain the load weight ratio of the main / backup relay in real time, according to the main relay weight Real-time calculation of the maximum current peak allowed in the main circuit , the calculation formula is ,in is the rated current of the relay; Set the load ratio allowable range. If the main relay weight coefficient is detected to be out of the range for multiple consecutive control cycles, it is judged as continuous deviation; When the load ratio continues to deviate, a gradual shutdown command is sent to the main relay, and a pre-charge command is sent to the standby relay at the same time; a ramp function is used to control the load weights of the main and standby relays, and the weight of the main relay is linearly reduced from the current value to zero within a fixed time window, and the weight of the standby relay is synchronously increased to 1; after the switching is completed, the main relay is marked as a standby, its life prediction model parameters are reset, and the maximum current peak is reinitialized based on the historical status data of the standby relay.

Images