Energy storage battery state of health management system based on load prediction

Abstract

The application discloses a load prediction-based energy storage battery health state management system, belongs to the technical field of energy storage battery management, and is characterized in that future load prediction and current state parameters of single bodies are deeply fused, an electrical-thermal-mechanical coupling model is used to recursively calculate at each time step, a group-level load curve is finely mapped to electrical, thermal and mechanical stress parameters of each single body at each time step in the future, a multi-dimensional stress time sequence is formed, the sequence is input into a pre-trained aging accumulation model at each time step, a current health state value is taken as a starting point to recursively generate an independent health state attenuation track of each single body, a traditional single health state prediction value is extended into a dynamic curve containing double information of initial difference and future differentiated stress history, and thus active balancing decisions based on a predictive safety boundary of single bodies are realized, charging and discharging strategies are regulated in advance before the health state is predicted to be lower than the boundary, and the service life of the whole group of batteries is prolonged.

Description

Technical Field

[0001] This invention relates to the field of energy storage battery management technology, and more specifically, to an energy storage battery health status management system based on load forecasting. Background Technology

[0002] With the large-scale grid connection of renewable energy and the development of smart grids, energy storage systems are playing an increasingly critical role in scenarios such as peak shaving and valley filling, frequency regulation, and backup power. The battery management system is one of the core components of the energy storage system, responsible for monitoring the voltage, current, temperature and other states of the battery, and implementing equalization management to ensure the safety and lifespan of the battery pack.

[0003] Currently, most commercial BMS adopts a reactive balancing strategy: the balancing circuit is activated or the charging and discharging parameters are adjusted only when the state difference between individual cells exceeds a preset threshold at the current moment. This method cannot predict future load changes (such as short-term high-power charging or discharging peaks) and is difficult to prevent some cells from aging faster due to future load shocks.

[0004] In recent years, some studies have attempted to predict battery health status, such as using historical data for data-driven modeling. However, most prediction technologies output a future health status value or trend, without using the prediction results to actively intervene in the current charging and discharging strategy. In other words, even if the system predicts that a certain cell will age rapidly in the future, traditional BMS still lacks a mechanism to actively adjust the load of that cell at the present moment to delay its degradation.

[0005] Furthermore, the utilization of future loads often remains at the level of energy management (such as peak shaving and valley filling strategies), and there is no deep coupling between load forecasting and the electro-thermal-mechanical aging mechanism of batteries, making it impossible to quantitatively assess the differentiated aging stress caused by different load curves to each individual cell.

[0006] Therefore, there is an urgent need for a new system that can integrate load forecasting, stress mapping and active balancing management to achieve preventive maintenance of the health status of energy storage batteries. Summary of the Invention

[0007] The purpose of this invention is to address practical technical deficiencies, and a health status management system for energy storage batteries based on load forecasting is provided.

[0008] The objective of this invention can be achieved through the following technical solution: a load forecast-based energy storage battery health status management system, comprising:

[0009] The data acquisition and generation module is used to collect the operating parameters of each cell in the energy storage battery pack in real time and obtain the current health status value of each cell. The operating parameters include the current value, temperature value, state of charge value, and internal resistance value of each cell. The module obtains the pack-level current sequence by summing the current of each cell at the same time.

[0010] The load forecasting and allocation module is used to generate the total charge and discharge current curve of the battery pack port within a preset period of time based on the group-level current sequence and future weather information. The total charge and discharge current curve of the battery pack port is combined with the current state of charge and ohmic internal resistance to decompose the expected current sequence of each cell within the preset period of time of time.

[0011] The single-cell multidimensional stress prediction module has a built-in electro-thermal-mechanical coupling model. It uses the expected current sequence of each single cell as excitation, and the current temperature, state of charge, ohmic internal resistance and preset heat dissipation coefficient of the single cell as the initial state. Through time-step recursive calculation, it outputs the electro-stress parameters, thermal stress parameters and mechanical stress parameters of the single cell at each future time step, thereby forming the multidimensional stress time series of the single cell.

[0012] The individual cell decay accumulation module is used to input the stress combination in its multidimensional stress time series into the pre-trained and optimized aging accumulation model, calculate the health status value step by step, and generate an independent health status decay trajectory for each cell.

[0013] The active balancing decision module compares the health status decay trajectory of each cell with the preset safety boundary. When the decay trajectory of any cell falls below the safety boundary at any future time, it generates a balancing control command for that cell to dynamically adjust its charging and discharging strategy.

[0014] Furthermore, the process of obtaining the total charge / discharge load curve at the battery pack port includes:

[0015] The system acquires group-level current sequences and obtains temperature and irradiance from future meteorological information as input features. These are then input into a pre-trained and optimized time series prediction model, which outputs group-level charge and discharge current curves for a predetermined future time period.

[0016] Furthermore, the process of obtaining the expected current sequence for each individual cell within a predetermined future time period includes:

[0017] Based on the preset weighted allocation rules, the group-level current sequence is decomposed into the expected current sequence of each individual cell. The preset weighted allocation rules include the proportion of the current state of charge of each individual cell, the weighting factor that is inversely proportional to the ohmic internal resistance, and the total current balance constraint. The total current balance constraint means that the sum of the expected currents of all individual cells calculated at each time step is equal to the total current of the group level.

[0018] Furthermore, the electro-thermal-mechanical coupling model is a discrete-time dynamic model based on state recursion. Its inputs include: the expected current sequence of the cell, the current temperature value, the current state of charge value, the current ohmic internal resistance value, and the heat dissipation coefficient at the location of the cell. The heat dissipation coefficient is a preset parameter of the system, which is pre-calibrated according to the thermal design of the battery pack and the location of the cell.

[0019] The model performs the following operations sequentially, starting from the first future time step:

[0020] Read the expected current at this time step, and use the temperature value and state of charge value at the end of the previous time step as the current state;

[0021] Calculate the ohmic heat generation in this step based on the expected current and the current ohmic internal resistance. Update the predicted temperature at the end of this step by combining the current temperature, heat dissipation coefficient and ambient temperature. Use the predicted temperature as the thermal stress parameter.

[0022] Update the state of charge at the end of the step based on the expected current and time step, and use the change in the state of charge within the step as a mechanical stress parameter.

[0023] The charge / discharge rate at the end of this step is updated by dividing the absolute value of the expected current by the rated capacity, and used as the electrical stress parameter.

[0024] Record the electrical stress parameters, thermal stress parameters, and mechanical stress parameters for this step, and pass the updated temperature and state of charge to the next time step. After completing all time steps, the model outputs a stress parameter sequence arranged in chronological order, i.e., a multidimensional stress time series.

[0025] Furthermore, the process of obtaining the individual health status decline trajectory for each unit includes:

[0026] The multidimensional stress time series of each individual cell is input into a pre-trained and optimized data-driven aging accumulation model. The model takes the current health status of the cell as the initial value and reads the stress combination of each time step in sequence. For each time step, the model outputs the health status value at the end of the time step based on the current health status value and the stress combination of the time step, and uses this value as the current health status value of the next time step. This process is repeated step by step to obtain the health status value of each future time step, and finally generates an independent health status decay trajectory for each cell.

[0027] Furthermore, the safety boundary in the active balancing decision module is a function curve that decreases with running time, determined by the battery retirement health threshold and expected remaining lifespan; when the predicted health status value of any single cell is lower than this boundary curve for the first time, balancing control is triggered; if multiple cells trigger simultaneously, balancing control is preferentially performed on the single cell with the largest difference between the predicted health status value and the boundary value.

[0028] Furthermore, the equalization control command includes at least one of the following: reducing the charge / discharge current limit of the cell, increasing its charge / discharge cutoff voltage, reducing its charge / discharge priority in the battery pack, or temporarily removing the cell from the charge / discharge path via a bypass switch.

[0029] Compared with the prior art, the advantages of this invention are:

[0030] 1. This invention obtains the charge-discharge current curve through load prediction and deeply integrates it with the current state of the individual cell (temperature, state of charge, internal resistance, heat dissipation coefficient, and ambient temperature). It uses an electro-thermal-mechanical coupling model to recursively calculate time-step by time, and refines the macroscopic charge-discharge current curve to map the electrical stress, thermal stress, and mechanical stress of each individual cell at each future time step. This forms a multi-dimensional stress time series at the individual cell level that contains multi-physics information, achieving a high-resolution, physically interpretable quantitative characterization of future aging drivers. This provides input features for subsequent health status prediction that are much richer and closer to the actual aging mechanism than traditional single-dimensional data.

[0031] 2. Based on the generation of multi-dimensional stress time series at the individual cell level, this series is input into a pre-trained and optimized data-driven aging accumulation model step by step. Starting from the current health state value, the health state value of each future time step is calculated recursively. This generates an independent health state decay trajectory that considers the initial differences between individual cells and the future differentiated stress history. This enables proactive balancing decisions based on the predictive safety boundary of individual cells. Before the health state of any individual cell is expected to fall below the safety boundary, a targeted balancing control command is generated in advance. This achieves a leap from the traditional "passive response balancing" to "predictive proactive balancing", slowing down the decay rate of the cells that age first and improving the consistency, usable capacity and overall service life of the entire battery pack. Attached Figure Description

[0032] Figure 1 This is a block diagram illustrating the system module principle of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0034] Example 1: This invention discloses a battery health status management system based on load forecasting. Please refer to [link / reference]. Figure 1It includes a data acquisition and generation module, a load prediction and allocation module, a single-unit multi-dimensional stress prediction module, a single-unit attenuation accumulation module, and an active balancing decision module.

[0035] The data acquisition and generation module is used to collect the operating parameters of each cell in the energy storage battery pack in real time and obtain the current health status value of each cell. The operating parameters include the current value, temperature value, state of charge value, and ohmic internal resistance value of each cell. The module obtains the pack-level current sequence by summing the current of each cell at the same time. The current health status value of each cell comes from the real-time estimation output of the battery management system.

[0036] The load forecasting and allocation module is connected to the data acquisition and generation module. It is used to generate the total charge and discharge current curve of the battery pack port in the future preset period based on the group-level current sequence and future weather information. The total charge and discharge current curve of the battery pack port is combined with the current state of charge and ohmic internal resistance to decompose the expected current sequence of each cell in the future preset period. This module is used to predict the overall load in the future and decompose the load to each cell.

[0037] The process of obtaining the total charge / discharge load curve at the battery pack port includes:

[0038] The system acquires group-level current sequences and obtains temperature and irradiance from future meteorological information as input features. These are then input into a pre-trained and optimized time series prediction model, which outputs group-level charge and discharge current curves for a preset future time period. The future meteorological information is provided by an external meteorological service.

[0039] The time series forecasting model is a deep learning-based sequence forecasting model. Before deployment, the model is trained offline using historical operating data. The input of the training samples are group-level current and meteorological data within the past window, and the output is the group-level current for the future window. During system operation, the model can be incrementally trained or fine-tuned periodically (e.g., daily or weekly) using newly accumulated operating data to adapt to load pattern drift caused by seasonal changes or system aging.

[0040] The process of obtaining the expected current sequence for each cell within a predetermined future time period includes:

[0041] Based on the preset weighted allocation rules, the group-level current sequence is decomposed into the expected current sequence of each individual cell. The preset weighted allocation rules include the proportion of the current state of charge of each individual cell, the weighting factor that is inversely proportional to the ohmic internal resistance, and the total current balance constraint. The total current balance constraint means that the sum of the expected currents of all individual cells calculated at each time step is equal to the total current of the group level.

[0042] The core of this rule is to calculate the current allocation weight of each cell based on its current state of charge and ohmic resistance at each time step, so that the algebraic sum of the currents of all cells equals the total current of the group, while giving priority to protecting cells in poor current condition (too high / too low state of charge, too large ohmic resistance) to avoid them being subjected to excessive stress.

[0043] The single-cell multidimensional stress prediction module is connected to the load prediction and allocation module and the data acquisition and generation module. It has a built-in electro-thermal-mechanical coupling model, which uses the expected current sequence of each single cell as excitation, and the current temperature, state of charge, ohmic internal resistance and preset heat dissipation coefficient of the single cell as the initial state. Through time-step recursive calculation, it outputs the electrical stress parameters, thermal stress parameters and mechanical stress parameters of the single cell at each future time step, thereby forming the multidimensional stress time series of the single cell.

[0044] The electro-thermal-mechanical coupling model is a discrete-time dynamic model based on state recursion. Its inputs include: the expected current sequence of the cell, the current temperature value, the current state of charge value, the current ohmic internal resistance value, and the heat dissipation coefficient of the cell's location. The heat dissipation coefficient is a preset parameter of the system, which is pre-calibrated according to the thermal design of the battery pack and the cell location. The ambient temperature can also be input synchronously. The ambient temperature can be obtained in real time through a temperature sensor or set to a fixed value.

[0045] The model performs the following operations sequentially, starting from the first future time step:

[0046] Read the expected current at this time step, and use the temperature value and state of charge value at the end of the previous time step as the current state;

[0047] The ohmic heat generation in this step is calculated based on the expected current and the current ohmic internal resistance. The predicted temperature at the end of this step is updated by combining the current temperature, heat dissipation coefficient, and ambient temperature. This predicted temperature is used as a thermal stress parameter. The specific operation is as follows: First, calculate the ohmic heat generation in this step by multiplying the square of the expected current by the current ohmic internal resistance and then by the time step. Second, calculate the heat dissipation by multiplying the difference between the current temperature and the ambient temperature by the heat dissipation coefficient and then by the time step. Third, subtract the heat dissipation from the ohmic heat generation to obtain the net heat. Divide the net heat by the preset battery heat capacity to obtain the temperature change in this step. Based on the temperature value and temperature change at the end of the previous time step, update the predicted temperature at the end of this step.

[0048] The state of charge at the end of the step is updated based on the expected current and time step. The state of charge decreases during discharge and increases during charging. The magnitude of the change in the state of charge within the step is used as a mechanical stress parameter.

[0049] The charge / discharge rate at the end of this step is updated by dividing the absolute value of the expected current by the rated capacity, and used as the electrical stress parameter.

[0050] Record the electrical stress parameters, thermal stress parameters, and mechanical stress parameters for this step, and pass the updated temperature and state of charge to the next time step;

[0051] After all time steps are completed, the model outputs a sequence of stress parameters arranged in chronological order, i.e., a multidimensional stress time series.

[0052] This module maps the expected current sequence of each cell to a multidimensional stress time series (electric, thermal, and mechanical stress), which includes multi-physics information. This enables a high-resolution, physically interpretable quantitative characterization of future aging drivers, providing input features that are far richer and closer to the actual aging mechanism than traditional single-dimensional data for subsequent health status prediction.

[0053] The single-unit decay accumulation module is connected to the single-unit multidimensional stress prediction module. It is used to input the stress combination in its multidimensional stress time series into the pre-trained and optimized aging accumulation model, calculate the health status value step by step, and generate an independent health status decay trajectory for each single unit.

[0054] The process of obtaining the individual health status decline trajectory of each cell includes:

[0055] The multidimensional stress time series of each cell is input into a pre-trained and optimized data-driven aging accumulation model. The model takes the current health status of the cell as the initial value and reads the stress combination of each time step in sequence. For each time step, the model outputs the health status value at the end of the time step based on the current health status value and the stress combination of the time step, and uses this value as the current health status value of the next time step. In this way, the health status value of each future time step is obtained by recursively calculating the health status value of each future time step, and finally generating an independent health status decay trajectory for each cell.

[0056] The aging accumulation model is a pre-trained data-driven model (e.g., neural network, random forest, or Gaussian process regression). Its training data comes from battery cycle aging experiments. Each training sample contains the stress combination (electrical stress, thermal stress, mechanical stress) within a time step, the health status value at the beginning of that time step, and the measured health status value at the end of that time step. By learning the above mapping relationship, the model can achieve the following recursive function in practical applications:

[0057] For the first future time step of a single unit, the model receives the current health state value and the stress combination of the first step, and outputs the health state value at the end of the first step.

[0058] For the kth time step, the model receives the health state value output from the previous step and the stress combination of the kth step, and outputs the health state value at the end of the kth step.

[0059] This process is repeated until all time steps are processed, resulting in a sequence of health status values ​​changing over time, i.e., the health status decay trajectory.

[0060] The model can use independently trained parameter sets for different cells to reflect differences in cell materials, initial aging conditions, or operating conditions.

[0061] By inputting the multidimensional stress time series at the individual unit level into the pre-trained and optimized data-driven aging accumulation model step by step, and recursively calculating the health state value at each future time step starting from the current health state value, it can generate an independent health state decay trajectory that considers the initial differences of the individual unit (initial state of charge, temperature, internal resistance, heat dissipation location) and the future differentiated stress history (rate, temperature evolution, state of charge fluctuation), thereby realizing proactive equilibrium decision-making based on the predictive safety boundary of the individual unit.

[0062] The active balancing decision module is connected to the single-cell attenuation accumulation module. It compares the attenuation trajectory of each single cell with the preset safety boundary. When the attenuation trajectory of any single cell is lower than the safety boundary at any future time, it generates an balancing control command for that single cell to dynamically adjust its charging and discharging strategy.

[0063] The safety boundary in the active balancing decision module is a function curve that decreases with running time, determined by the battery retirement health threshold and expected remaining lifespan. When the predicted health status value of any single cell first falls below this boundary curve, balancing control is triggered. If multiple cells trigger simultaneously, balancing control is performed on the single cell with the largest difference between the predicted health status value and the boundary value.

[0064] The equalization control command includes at least one of the following: reducing the charge and discharge current limit of the cell, increasing its charge and discharge cutoff voltage, reducing its charge and discharge priority in the battery pack, or temporarily removing the cell from the charge and discharge path by means of a bypass switch.

[0065] It can answer both "how much will each individual cell decay to in the future" and "at what rate will the decay occur and when will it cross the safety line";

[0066] Traditional methods can only output a single numerical value indicating "what the future health status of a certain cell is," failing to explain "why it decays faster." This invention addresses the problem that health status decay trajectories are cumulative codes of "initial state vector + future stress history," rather than a single numerical value. This clearly reveals whether high-rate charging (electrical stress), poor heat dissipation leading to temperature rise (thermal stress), or drastic fluctuations in state of charge (mechanical stress) are driving the accelerated aging of the cell. This interpretability provides a basis for proactive balancing decisions: for example, for cells dominated by thermal stress, priority is given to adjusting their current to reduce heat generation; for cells dominated by mechanical stress, priority is given to adjusting their charge / discharge depth range.

[0067] In summary, this approach deeply integrates future load forecasts with the current state parameters of individual units. By using an electro-thermal-mechanical coupling model to iteratively advance the data step by step, the group-level load curves are refined and mapped to the electrical, thermal, and mechanical stress parameters of each individual unit at each future time step. This forms a multi-dimensional stress time series, enabling a high-resolution, physically interpretable quantitative characterization of future aging drivers. This provides input features for subsequent health status prediction that are far richer and closer to the actual aging mechanism than traditional single-dimensional data.

[0068] The sequence is then input into a pre-trained aging accumulation model step by step. Starting from the current health status value, the model recursively generates an independent health status decay trajectory for each cell. This expands the traditional single health status prediction value into a dynamic curve that includes both initial differences and future differentiated stress history information. This enables proactive balancing decisions based on the predictive safety boundary of each cell, allowing for early adjustment of the charging and discharging strategy before the health status is expected to fall below the boundary, thereby extending the service life of the entire battery pack.

[0069] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto; any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and its improved concept, should be covered within the scope of protection of the present invention.

Claims

1. A battery health status management system based on load forecasting, characterized in that: include: The data acquisition and generation module is used to collect the operating parameters of each cell in the energy storage battery pack in real time and obtain the current health status value of each cell. The operating parameters include the current value, temperature value, state of charge value, and internal resistance value of each cell. The module obtains the pack-level current sequence by summing the current of each cell at the same time. The load forecasting and allocation module is used to generate the total charge and discharge current curve of the battery pack port within a preset period of time based on the group-level current sequence and future weather information. The total charge and discharge current curve of the battery pack port is combined with the current state of charge and ohmic internal resistance to decompose the expected current sequence of each cell within the preset period of time of time. The single-cell multidimensional stress prediction module has a built-in electro-thermal-mechanical coupling model. It uses the expected current sequence of each single cell as excitation, and the current temperature, state of charge, ohmic internal resistance and preset heat dissipation coefficient of the single cell as the initial state. Through time-step recursive calculation, it outputs the electro-stress parameters, thermal stress parameters and mechanical stress parameters of the single cell at each future time step, thereby forming the multidimensional stress time series of the single cell. The individual cell decay accumulation module is used to input the stress combination in its multidimensional stress time series into the pre-trained and optimized aging accumulation model, calculate the health status value step by step, and generate an independent health status decay trajectory for each cell. The active balancing decision module compares the health status decay trajectory of each cell with the preset safety boundary. When the decay trajectory of any cell falls below the safety boundary at any future time, it generates a balancing control command for that cell to dynamically adjust its charging and discharging strategy.

2. The energy storage battery health status management system based on load forecasting according to claim 1, characterized in that: The process of obtaining the total charge / discharge load curve at the battery pack port includes: The system acquires group-level current sequences and obtains temperature and irradiance from future meteorological information as input features. These are then input into a pre-trained and optimized time series prediction model, which outputs group-level charge and discharge current curves for a predetermined future time period.

3. The energy storage battery health status management system based on load forecasting according to claim 2, characterized in that: The process of obtaining the expected current sequence for each cell within a predetermined future time period includes: Based on the preset weighted allocation rules, the group-level current sequence is decomposed into the expected current sequence of each individual cell. The preset weighted allocation rules include the proportion of the current state of charge of each individual cell, the weighting factor that is inversely proportional to the ohmic internal resistance, and the total current balance constraint. The total current balance constraint means that the sum of the expected currents of all individual cells calculated at each time step is equal to the total current of the group level.

4. The energy storage battery health status management system based on load forecasting according to claim 3, characterized in that: The electro-thermal-mechanical coupling model is a discrete-time dynamic model based on state recursion. Its inputs include: the expected current sequence of the cell, the current temperature value, the current state of charge value, the current ohmic internal resistance value, and the heat dissipation coefficient of the cell's location. The heat dissipation coefficient is a preset parameter of the system, which is pre-calibrated according to the thermal design of the battery pack and the location of the cell. The model performs the following operations sequentially, starting from the first future time step: Read the expected current at this time step, and use the temperature value and state of charge value at the end of the previous time step as the current state; Calculate the ohmic heat generation in this step based on the expected current and the current ohmic internal resistance. Update the predicted temperature at the end of this step by combining the current temperature, heat dissipation coefficient and ambient temperature. Use the predicted temperature as the thermal stress parameter. Update the state of charge at the end of the step based on the expected current and time step, and use the change in the state of charge within the step as a mechanical stress parameter. The charge / discharge rate at the end of this step is updated by dividing the absolute value of the expected current by the rated capacity, and used as the electrical stress parameter. Record the electrical stress parameters, thermal stress parameters, and mechanical stress parameters for this step, and pass the updated temperature and state of charge to the next time step. After completing all time steps, the model outputs a stress parameter sequence arranged in chronological order, i.e., a multidimensional stress time series.

5. The energy storage battery health status management system based on load forecasting according to claim 4, characterized in that: The process of obtaining the individual health status decline trajectory of each cell includes: The multidimensional stress time series of each individual cell is input into a pre-trained and optimized data-driven aging accumulation model. The model takes the current health status of the cell as the initial value and reads the stress combination of each time step in sequence. For each time step, the model outputs the health status value at the end of the time step based on the current health status value and the stress combination of the time step, and uses this value as the current health status value of the next time step. This process is repeated step by step to obtain the health status value of each future time step, and finally generates an independent health status decay trajectory for each cell.

6. The energy storage battery health status management system based on load forecasting according to claim 5, characterized in that: The safety boundary in the active balancing decision module is a function curve that decreases with running time, determined by the battery retirement health threshold and expected remaining lifespan. When the predicted health status value of any single cell first falls below this boundary curve, balancing control is triggered. If multiple cells trigger simultaneously, balancing control is preferentially performed on the single cell with the largest difference between the predicted health status value and the boundary value.

7. The energy storage battery health status management system based on load forecasting according to claim 1, characterized in that: The equalization control command includes at least one of the following: reducing the charge / discharge current limit of the cell, increasing its charge / discharge cutoff voltage, reducing its charge / discharge priority in the battery pack, or temporarily removing the cell from the charge / discharge path via a bypass switch.

Images