Battery system health monitoring system

Abstract

A battery system health monitoring system. A monitoring system (120) for a battery system (110) comprising one or more battery cells (110). The monitoring system (120) comprises one or more sensors (125) for measuring a characteristic associated with the battery system (110) during a charging or discharging operation, and a processing device (130) communicatively coupled to the one or more sensors (125). The processing device (130) comprises a pre-processor (135) configured to receive measurement data from the at least one sensor (125), and a neural network (140) configured to receive processed data from the pre-processor (135). Based on the received measurement data, the pre-processor (135) determines a normalized rate of change, with respect to one of the times, of a first measured characteristic or a second measured characteristic associated with the battery system (110) measured during the charging or discharging operation, where the second measured characteristic differs from the first measured characteristic. The neural network (140) is configured to use the normalized determined rate of change from the pre-processor (135) as an input to determine a state of health (SOH) of the battery system (110).

Description

Technical Field

[0001] The present disclosure relates to the field of battery technology, energy cells, and battery monitoring systems and solutions for measuring the state of health (SoH) of a battery system. Specifically, embodiments of the present disclosure relate to monitoring systems, processing devices, methods, and computer-readable media for determining the state of health of a battery system.

Background Art

[0002] Battery systems are used in a wide range of modern power applications. For example, battery systems are used to power electric vehicles and in commercial applications such as industrial power applications, transportation, energy storage, and powering modern electronic devices. Considering the relatively high power requirements in such applications, battery systems often include a plurality of battery cells connected to each other to achieve the required power output. The battery cells can be connected to each other to form a battery pack, and a battery system can include one or more battery packs.

[0003] It is common to monitor the voltage and current of battery cells within a battery pack to measure the operating state of the battery cells. Typically, a battery system is connected to a battery management system that receives a signal indicating the current operating state of the battery cells, and it is ensured that the battery system operates within its safe operating limits (i.e., the combination of voltage, current, and temperature conditions under which the battery system is predicted to operate without self-damage). For further details, interested readers are referred to the following Wikipedia website: https: / / en.wikipedia.org / wiki / Battery_management_system.

[0004] One aspect of a battery system that is highly desirable to monitor is the health of the battery system, or State of Health (SoH). The SoH of a battery system is the state of the battery system at the time of manufacture. charge Battery systems that have aged in comparison to their capacity charge It can be defined as a measure of the percentage of capacity. Typically, SoH is 100% at the time of manufacture and decreases over time and with use. By monitoring the SoH of a battery system, the degradation of the battery system from the time of manufacture can be visualized, and this can be used to implement effective operation and maintenance of the battery system. However, unlike temperature, voltage, or current, SoH cannot be measured directly. Rather, SoH needs to be determined based on the behavior of the battery system during charge and discharge operations.

[0005] Conventional methods for determining the State of Heat (SoH) of a battery system typically involve performing a complete low-speed (low C-speed) charge-discharge cycle on the battery system. charge To measure the volume, and then the measured charge Capacity of the battery system at the time of initial manufacture charge This involves determining the ratio to the capacity. However, this approach is not always feasible or practical. This is because some battery systems are never fully charged or discharged, which can negatively impact the battery system's lifespan or interfere with normal operation in battery power supply applications.

[0006] Directly measuring SoH is often impractical, and other methods have been proposed.

[0007] One proposed approach is to design a model that describes the behavior corresponding to the behavior of the battery system. While such models are typically implemented using software, some very simple models can be computed using dedicated hardware. The model could be an equivalent circuit model, and here, Charge retention capacityThese are, for example, state variables that have been modified to reduce errors in the predicted open-circuit voltage due to the Kalman loop.

[0008] However, using a model-based method, the charge state of the battery system, or what is often called the SoC, is determined by the remaining state of the battery system. charge Capacity, complete battery charge The ratio to capacitance, and other state variables such as internal resistance, are calculated. Charge retention capacity It is difficult to separate them. Furthermore, the model, charge In physical models where capacity is determined by the physical properties of the cell and can change, accurate modeling requires knowledge of many different cell parameters, as well as other contributing factors such as the amount of lithium lost from the cell during use and the amount of lithium available. Most of these cannot be measured directly, and therefore the practicality of model-based methods is limited.

[0009] A different approach is disclosed in CN110824364A, which proposes using deep neural networks to perform SoH prediction. Deep neural networks have multiple hidden layers between the input and output layers, and can model complex nonlinear relationships with fewer data units than shallow neural networks that achieve similar performance and have fewer (typically fewer than 10) hidden layers between the input and output layers.

[0010] However, deep neural networks, such as those described in CN110824364A, require intensive processing power, and the processing requirements needed to implement deep neural networks exceed the processing power of chips in embedded systems that could be used to provide on-site monitoring of battery systems, such as battery systems in electric vehicles. This makes such approaches unsuitable for real-world implementation.

[0011] While advances in cloud processing capabilities mean that some of the processing activities required to implement deep neural networks can be performed in the cloud, implementing deep neural networks remotely presents other challenges. In particular, for SoH prediction, remote processing of data by deep neural networks requires a battery monitoring system to be constantly connected to the remote processing system, and the system must include additional hardware to enable secure transmission of data in real time. These requirements mean that SoH prediction based on deep neural networks, from a practical standpoint, is limited to laboratory environments where these problems can be avoided.

[0012] Another approach is presented in DE 10 2015 016 987 A1, which involves detecting the terminal voltage of the battery cell during the charging or discharging process and the SoC that would have flowed during the charging process, or charge This involves mathematically deriving terminal voltages based on measurements and identifying peak values ​​from the derived dataset. However, this approach also has several drawbacks. The disclosed system is charge Alternatively, it relies on providing an accurate model for deriving SoC values, requiring memory to store the previously derived datasets for later comparison with subsequent datasets. Furthermore, the approach discussed in DE 10 2015 016 987 A1 only provides information about the relative degree of battery system degradation, not absolute measurements of SoH.

[0013] From the above perspective, an alternative approach to monitoring the State of Heat (SOH) of battery systems is needed.

[0014] An objective of at least some embodiments of this disclosure is to address one or more drawbacks of prior art systems. [Overview of the Initiative]

[0015] One aspect of the present disclosure provides a monitoring system for a battery system comprising one or more battery cells. The monitoring system comprises one or more sensors for measuring characteristics associated with the battery system during a charge or discharge operation, and a processing device communicatively coupled to one or more sensors. The processing device comprises a preprocessor configured to receive measurement data from at least one sensor, and a neural network configured to receive processed data from the preprocessor. Based on the received measurement data, the preprocessor, The normalized rate of change of the first measured characteristic to one of the second measured characteristics associated with the battery system, measured over time or during a charge or discharge operation. The second measured characteristic is determined to be different from the first measured characteristic. The neural network is configured to determine the state of health (SOH) of the battery system, using the determined, normalized rate of change from the preprocessor as input.

[0016] Another aspect of this disclosure provides a processing device for use in a battery monitoring system. The processing device comprises a preprocessor configured to receive measurement data from at least one sensor, and a neural network configured to receive processed data from the preprocessor. Based on the received measurement data, the preprocessor The normalized rate of change of the first measured characteristic to one of the second measured characteristics associated with the battery system, measured over time or during a charge or discharge operation. The second measured characteristic is determined to be different from the first measured characteristic. The neural network is configured to determine the state of health (SOH) of the battery system, using the determined, normalized rate of change from the preprocessor as input.

[0017] Another aspect of this disclosure provides a method for determining the state of health (SOH) of a battery system. According to this method, during a charge or discharge operation, measurement data representing one or more characteristics associated with the battery system is received. Based on the received measurement data, The normalized rate of change of the first measured characteristic to one of the second measured characteristics associated with the battery system, measured over time or during a charge or discharge operation. The second measured characteristic is determined to be different from the first measured characteristic. The determined, normalized rate of change is then used as input to a neural network to determine the state of health (SOH) of the battery system.

[0018] A further aspect of the present disclosure provides a non-transitory computer-readable medium for configuring a processing device or a monitoring system.

[0019] Other features of the present disclosure will become apparent from the dependent claims and the following description. Further features of the present disclosure will become apparent from the description following the exemplary embodiments, with reference to the accompanying drawings.

[0020] Further aspects and embodiments will become apparent from the following description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] [Figure 1] Schematic diagrams of a battery system and a monitoring system for determining the soundness of the battery system, according to an embodiment of the present disclosure. [Figure 2] A and B are exemplary graphs showing how the rate of change of charge with respect to voltage (dQ / dV) varies with the soundness of an exemplary battery system. [Figure 3] A and B are exemplary graphs showing how the normalized rate of change of charge with respect to voltage (dQ / dV) varies with the soundness of an exemplary battery system. [Figure 4] Process flowchart showing a method for determining the soundness of a battery system using the monitoring system of FIG. 1. [Figure 5] A and B are exemplary graphs showing how the rate of change of voltage with respect to charge (dV / dQ), measured in a charge-discharge cycle, varies with the soundness of an exemplary battery system. [Figure 6] A and B are exemplary graphs showing how the rate of change of voltage with respect to state of charge (dV / dSoC), measured in a charge-discharge cycle, varies with the soundness of an exemplary battery system. [Figure 7]This graph illustrates the rate of change in voltage relative to the charge state (dV / dSoC) measured during a charge-discharge cycle, and how the health of a battery system changes when charged at multiple different charging speeds. [Figure 8] This is a process flowchart illustrating a method for training a neural network and determining the health of a battery system based on characteristic derivatives measured according to embodiments of this disclosure. [Modes for carrying out the invention]

[0022] Figure 1 is a schematic diagram of a battery system 110 and a monitoring system 120 for determining the State of Heat (SoH) of the battery system 110, according to an embodiment of the present disclosure.

[0023] The battery system 110 may comprise a single battery cell. Alternatively, in other embodiments, the battery system 110 may be a battery pack comprising multiple battery cells. In embodiments where the battery system 110 comprises a single battery cell, the monitoring system 120 is configured to determine the State of Heat (SoH) of that battery cell. In other embodiments where the battery system 110 comprises multiple battery cells, the monitoring system 120 may be configured to determine the overall SoH of the multiple battery cells.

[0024] The monitoring system 120 includes one or more sensors 125 for measuring the characteristics of the battery system 110 during charging and / or discharging operations. The sensors 125 may include one or more sensors configured to measure the voltage and / or current of the battery system 110 during charging and / or discharging operations. In some embodiments, the sensors 125 may also include one or more sensors configured to measure the temperature of the battery system 110.

[0025] The monitoring system 120 further comprises a processing device 130 configured to receive signals and / or data from sensors 125 (e.g., signals and / or data indicating voltage, current, temperature, etc.) that represent the measured characteristics of the battery system 110. In some embodiments, the processing device 130 may be configured to timestamp the received sensor data to facilitate time-series analysis of the battery system's characteristics.

[0026] In some embodiments, the processing device 130 may be connected to the sensor 125. In other embodiments, a communicative connection exists between the processing device 130 and the sensor 125, allowing the sensor data from the battery system to be relayed to the processing device 130.

[0027] The complexity of the neural network required to derive the SoH measurement from the measured characteristics of the battery system is determined by preprocessing the measured characteristics. charge The applicants understand that the processing of normalized differential data (i.e., the normalization rate of change data) can be significantly reduced by processing normalized differential data (i.e., the normalization rate of change data) instead of raw measured values ​​such as voltage and temperature. The degree of this simplification is such that it is possible to derive SoH measurements using a shallow neural network instead of using a deep neural network as previously proposed. This reduces the computational load on the processing device 130 to determine the SoH of the battery system 110, making it possible to determine SoH measurements with limited processing power, for example, that may be available in embedded systems. By providing a shallow neural network, the possibility of the neural network overfitting to the training data during training, which could lead to errors during verification, is also reduced.

[0028] Furthermore, the applicants understand that SoH measurements can be performed based on such determined differential data for a limited portion of charge-discharge cycles without significant loss of accuracy. Therefore, by preprocessing data representative of the measured characteristics to determine and normalize the differential data of the rate of change of such characteristics, SoH measurements can be determined for a wide range of battery systems, including battery systems in which the battery cells or battery packs are never fully charged or discharged.

[0029] To this end, in this embodiment, the processing device 130 includes a preprocessor 135 configured to determine and normalize the rate of change of data of measured characteristics received from the sensor 125, and a neural network 140 that is processed by the preprocessor 135 and trained to determine the State of Heat (SoH) from the measured characteristics of the battery system.

[0030] Normalizing the rate of change of the measured characteristics and feeding this normalized rate of change to a properly trained neural network 140 offers several advantages over systems where the neural network processes the data as is. Utilizing the normalized rate of change has the potential to improve prediction accuracy, further reduce the size of the neural network required to determine the State of Heat (SoH), reduce the training data requirements for training the neural network, and assist the trained system in deriving SoH for a wider range of battery cells and cell types.

[0031] Figures 2A, 2B, 3A, and 3B illustrate why using normalized data change rates, rather than raw data change rates, can be advantageous.

[0032] Figures 2A and 2B show the voltage of two 18650 LFP chemical cells. Charge relative toThese are a pair of graphs showing how the rate of change (dQ / dV) changes under various charging and degradation methods. On the other hand, Figures 3A and 3B show, For the corresponding voltage Normalized charge These are a pair of corresponding graphs showing how the rate of change (dQ / dV) is determined to change.

[0033] To make it clear, the graphs in Figures 2A and 2B show the voltage corresponding to the battery's SoH. Charge relative to It shows characteristic fluctuations in the rate of change (dQ / dV). However, voltage against Normalized charge The graphs in Figures 3A and 3B show that the rate of change (dQ / dV) exhibits a higher level of consistency across cells.

[0034] For example, Figures 3A and 3B show voltage against Normalized charge The peak position of the rate of change (dQ / dV) clearly shifts to the right in the graph as the battery ages. In contrast, Figures 2A and 2B show the voltage against Just as it is charge In the case of the rate of change (dQ / dV), a pattern exists, but it is not very clear.

[0035] By preprocessing the data rate of change and supplying the normalized data rate of change to the neural network, the complexity of the neural network required and the level of training data needed to train the neural network and derive health measurements are reduced. This is because normalizing such data removes fluctuations in peak values ​​(for example, the fluctuation of 7.2 to 6.5 in the peak values ​​shown in Figures 2A and 2B), and instead provides the neural network with the task of deriving the SoH value based only on horizontal shift, rather than horizontal shift and amplitude change.

[0036] Furthermore, in addition to reducing the complexity of the neural network required to derive the SoH value, the applicants found that the variation pattern of the normalized rate of change is relatively constant across multiple cell types, due to the fact that the battery systems exhibit similar parabolic differential properties comparable to the trained dataset. This occurs despite the accurate relationship between SoH and the normalized characteristic derivative across different battery system variations.

[0037] Simply put, normalization enables improved pattern recognition compared to processing based on absolute derivatives. Absolute derivatives differ significantly between one cell variation and another; therefore, models trained using absolute derivatives are limited to deriving State of Health (SOH) for a single type of cell variation. However, since most information about cell degradation is derived from the location of the peak rather than the absolute value of such peaks, it is understood that satisfactory results can be obtained by training on normalized derivative data without significant loss of accuracy.

[0038] Therefore, for all the reasons stated above, it is concluded that in addition to processing the sensor measurements as they are to derive the rate of change of the data, it is advantageous to then normalize or rescale such data before feeding it to a neural network in order to convert it into the determined measurement of SoH. This is because it not only reduces the complexity of the neural network required, but also reduces the requirement for training data to train such a network, as the change patterns of the normalized rate of change can be interpolated and compared across multiple chemical cell types.

[0039] Therefore, the approach described is suitable for generating SoH prediction models for novel or unknown cell variations. When adapting a model for use with novel or unknown cell types, an existing general prediction model can be utilized by first adapting it to the specifications of the new cell—for example, thresholds such as voltage, capacitance, and current relative to capacitance—through appropriate changes to the preprocessing. The required adaptations can be identified based on data available from the cell specification sheet. While the performance of such a model may be sufficiently accurate for some applications, the accuracy of such a model can be further improved by subsequently modifying the system based on cell-specific test data and / or test data from similar cells. Advantageously, such an approach makes the SoH model available during the early stages of cell availability when developing new cells and new systems.

[0040] Returning to Figure 1, the preprocessor 135 for processing sensor data and converting such sensor data into a normalized rate of change can take the form of a microprocessor, an embedded processor, or be integrated into a system-on-a-chip (SoC), but is not limited to these. In some embodiments, the preprocessor 135 may comprise fixed-function digital logic circuits. Furthermore, according to some embodiments, the preprocessor 135 may originate from a family of processors manufactured by Intel®, AMD®, Qualcomm®, Apple®, NVIDIA®, etc. The preprocessor 135 may also be based on an ARM architecture, a mobile processor, or a graphics processing unit, etc. The disclosed embodiments are not limited to any particular type of preprocessor.

[0041] In one embodiment, the neural network 140 is a feedforward neural network having multiple processing layers for determining the State of Heat (SoH) of a monitored battery system 110, which has been trained in a manner described later, and for determining the SoH from the characteristic derivative of a specific normalized battery system.

[0042] Preferably, the neural network 140 is a shallow feedforward neural network having a limited number of nodes and layers. In one embodiment, the neural network 140 may have an input layer, a hidden layer, and an output layer. In some embodiments, the neural network 140 may have multiple hidden layers. In some embodiments, the neural network 140 may have 4 to 6 hidden layers. As described above, the confidence of the SoH measurements derived by the shallow neural network is increased by data preprocessing performed by the preprocessor 135, which reduces input variability in the input data and converts such input values ​​into the derived SoH measurements, thereby reducing the complexity of the task of the neural network 140 and making the task suitable for the capabilities of the shallow neural network.

[0043] In some embodiments, the processing device 130 may be configured to determine the State of Heat (SoH) of the battery system 110 by sequentially firing each layer of a plurality of layers of the neural network 140. By firing the layers of the neural network 140 in this manner, the processing resources required to determine the SoH in multiples equal to the number of layers in the neural network 140 are further reduced. By providing a preprocessor 135 that derives a normalized rate of change of data from the measurements generated by the sensor 125, the determination of SoH by the monitoring system 120, which may have limited processing power, is further facilitated. Moreover, this can be achieved without compromising the accuracy of the SoH determination. The sequential firing of the layers of the neural network 140 in this manner can be initiated by software on the processing device 130. Advantageously, the determination of the battery system's SoH using the disclosed technique can be performed in a single run of the neural network 140 using data collected for a given charge-discharge cycle.

[0044] In some embodiments, the processing device 130 may further include post-processing means configured to remove noise or unexpected values ​​from the output of the neural network. In one embodiment, the post-processing means may include at least one of a low-pass filter or a smoothing filter.

[0045] The processing device 130 can report the determined State of Heat (SoH) of the battery system, and optionally, any additional measured characteristics of the battery system, to a battery management system (BMS) (not shown). The BMS can compare the performance of multiple battery systems (or battery cells). If the performance of a battery system is found to differ from that of other battery systems by more than a predetermined amount, this may indicate a defect in the battery system and prompt the BMS to flag the identified battery system as defective. The predetermined amount may initially be set to a default bind set at the time of manufacture. The predetermined amount may be selected in response to behavior learned from previous battery systems. In some embodiments, the BMS can transmit the collected data to the cloud on a continuous / regular / irregular schedule. The identified differences in battery system performance can be used to optimize the operation and / or maintenance of the battery system.

[0046] Figure 4 is a process flowchart showing a method 200 for determining the health of the battery system 110 using the monitoring system 120 shown in Figure 1.

[0047] As an initial step 210, the battery system 110 undergoes at least a partial charge-discharge cycle, during which the sensor 125 collects measurement data in step 220 that show the characteristics of the battery system 110 as it is being charged / discharged.

[0048] In one embodiment, the collected measurement data may include current and voltage data. In some embodiments, the temperature of the battery system 110 can also be measured.

[0049] In this embodiment, measurement data can be collected by the sensor 125 over a subrange of full charge-discharge cycles in the battery system 110. For example, the measurement data may be collected over a subrange of the battery system (SoC) chargeData can be collected at 20–80% of capacity. In other embodiments, measurement data can be collected within one of the following charge / discharge ranges: 30–70%SoC (or 70–30%SoC), 40–60%SoC (or 60–40%SoC). In one embodiment, the range of charge / discharge operations from which measurement data is collected is one of the following variations of SoC: 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 100%. While the described methods can provide sufficient SoH measurements using smaller ranges of charge / discharge data, the accuracy and effectiveness of these methods in variations of the battery system (e.g., battery systems with different designs or cell chemical properties) can be improved by extending the range of charge / discharge operations from which measurement data is collected. In some embodiments, the charge / discharge range may be a range that occurs naturally in the application of the battery system. In other embodiments, subranges may be measured based on the properties of the battery system to avoid overcharging or over-discharging the battery system and extend the battery system's lifespan. Embodiments of this disclosure are not limited to a single charge / discharge range. In some embodiments, the monitoring system 120 may measure different charge / discharge ranges for different charge / discharge cycles.

[0050] In step 220, the preprocessor 135 receives measurement data from the sensor 125 that indicates characteristics related to the battery system 110. Next, in step 230, the preprocessor determines, based on the received measurement data, the rate of change of a characteristic for any given time, or another characteristic of the battery system 110 during a charge / discharge operation.

[0051] In one embodiment, the preprocessor 135, based on the measurement data received from the sensor 125, determines the battery system 110 charge It may be configured to determine the rate of change of voltage (dV / dQ) relative to . In the above embodiment, the received measurement data may include current and voltage measurement data associated with the battery system 110, spanning the duration of the charge-discharge operation.

[0052] Figures 5A and 5B show measurements taken during charge-discharge cycles. For electric charge This is an exemplary graph showing how the rate of change of voltage (dV / dQ) changes with the health of an exemplary battery system. The rate at which a battery system is charged or discharged is known as the C-rate. The C-rate of a charge / discharge operation can be determined based on the charge / discharge current passing through the battery system, divided by the theoretical draw current that the battery system delivers its nominal rated capacity in one hour. Figure 5A shows the dV / dQ measured in battery system 110 during a charge operation at the C-rate described in 3.6, while Figure 5B shows the dV / dQ measured in battery system 110 during a charge operation at the C-rate described in 4.8. As can be seen in the figures, a clear trend is discernible as battery system 110 degrades from 100% to 85% SoH. Using the graphs shown, for example, by interpolation / extrapolation, it is possible to identify the point in the graph where a battery system with an unknown SoH would fit in order to predict the SoH value of the battery system. The x-axis of the figure is derived from Coulomb counting across the range starting from the open-circuit voltage at Vmin, using current measurement data from the battery system 110. charge This represents the calculated SoH results obtained using dV / dQ analysis in the disclosed method, which have proven to be highly accurate, for 0.35Ah to 0.70Ah. charge An absolute dV / dQ relationship, determined from measurement data collected across a range, is used to achieve an RMSE of less than 1%. Therefore, unlike prior art, the disclosed solution can accurately determine the State of Heat (SoH) from partial charge-discharge cycles using measurement data at faster C speeds.

[0053] In another embodiment, the preprocessor 135 is configured to determine the rate of change of voltage (dV / dSoC) of the battery system 110 relative to the SoC, based on measurement data received from the sensor 125. In this embodiment, the received measurement data may include current and voltage data associated with the battery system 110, spanning the duration of a charge-discharge operation. The preprocessor 135 may be configured to determine the SoC from the received measurement data using any preferred method well known in the art.

[0054] Figures 6A and 6B show the charge state measured during a charge-discharge cycle. against This is an exemplary graph showing how the rate of change of voltage (dV / dSoC) changes with the health of an exemplary battery system. Figure 6A shows the dV / dSoC measured in battery system 110 during a charging operation at the C speed of 3.6, while Figure 6B shows the dV / dQ measured in battery system 110 during a charging operation at the C speed of 4.8. The x-axis of the figure represents the SOC of battery system 110, calculated as the ratio of available A Ah to the maximum Q Ah achieved at the end of each charging cycle. Using the dV / dQ method described above, the figure shows an identifiable trend as battery system 110 degrades from 100% to 85% SoH. SoH simulations using dV / dSoC in the described method achieved an RMSE of less than 1% using measurement data collected over a charging range of 25-80% SoC.

[0055] In many applications, it is desirable to vary the C-speed of charge and discharge operations based on the SoC of the battery system. For example, in some cases, the battery system 110 may be charged at a high C-speed when the SoC is below 50%, and then at a low C-speed when the SoC is above 80%. When the charging speed transitions from high C-speed to low C-speed, a voltage discontinuity or step response can be observed across the battery system 110 due to the charging characteristics of the internal resistance and the chemical properties of the battery system. This results in discontinuity in the dV / dSOC measurements, as shown in Figure 7. Unlike prior art SoH techniques, the combination of characteristic differential analysis of the battery system and trained neural network pattern recognition used in the disclosed method has been proven effective in determining the SoH of the battery system from measurement data collected across multiple charge-discharge cycles involving different charging speeds.

[0056] Figure 7 shows the charge state measured during a charge-discharge cycle. against This is an exemplary graph illustrating the rate of change in voltage (dV / dSoC) and how the health of a battery system changes when charged at multiple different charging speeds. The graph shows the determined dV / dSoC for a battery system 110 that was charged to 40% SoC at C speed 6, and then to 80% at C speed 4. By using an additional processing step to mask absolute dV / dSOC values ​​below a fixed threshold (which may be based on the battery system configuration), it becomes possible to determine SoH using measurement data that includes multiple charging speeds during a single charging cycle. Therefore, by further processing the resulting characteristic differential data of the battery system, it is ensured that the determination of SoH is not distorted by interference caused by swapping charging speeds.

[0057] In another embodiment, the preprocessor 135 is configured to determine a rate of change including at least one of dV / dt, dV / dT, dQ / dT, dQ / dSoC, dQ / dt, and dSoC / dT. With respect to embodiments in which the rate of change of the battery system's characteristics with respect to time is determined, the preprocessor 135 may first timestamp the received measurement data in order to analyze the battery system's characteristics with respect to the time of charge-discharge operations. Each characteristic derivative of a battery system has its own advantages and disadvantages associated with that characteristic derivative.

[0058] For example, while dV / dt has been found to provide high SoH accuracy for a single battery system design, dV / dt does not perform as well as other derivatives (dV / dQ and dV / dSoC) when used in variations of the battery system (for example, dV / dt trained on cell chemical feature A may not provide as accurate SoH prediction as an untrained cell chemical feature B). Advantageously, using dV / dt also results in simpler implementations, as it allows for direct sampling of raw, time-stamped voltage measurements without further data manipulation. The specific battery system characteristic derivative used in the disclosed method can therefore be selected based on the specific application.

[0059] In step 240, the determined rate of change is then rescaled / normalized. Rescaling and / or normalization can be performed on the rate of change using any preferred method known in the art. As already mentioned above, rescaling / normalizing the determined rate of change improves the effectiveness of determining SoH from modified battery systems and associated measurement data.

[0060] For example, as shown in Figures 6A and 6B, the determined range for dV / dSoC can be rescaled to a range of 1 to -1, and represented as 1 to -1 instead of the original value using, for example, 32 bits, which facilitates easier weighting and supports higher resolution. Furthermore, the low accuracy of the measurement data from sensor 125 can be compensated for by reducing the sampling interval of the measurement data used to determine dV / dSOC. In one embodiment, a sampling interval of 0.5 to 5% SOC can be used.

[0061] In step 250, the processed (i.e., rescaled / normalized) rate of change is used as input to the neural network 140, which, using pattern matching analysis on the received normalized rate of change, outputs the State of Home (SoH) for the battery system 110 in step 260.

[0062] In some embodiments, in addition to the determined, normalized rate of change, the true average current measured by the battery system over the duration of the charge-discharge operation can also be used as input to the neural network 140. Including the average current can improve the accuracy of the State of Heat (SoH) and the adaptability to variations in the battery system. This is because the normalized derivatives, which independently include dV, dt, dT, and dQ, do not provide an index of the rate of change for a given step. Although the average rotational current can be used in some embodiments, this may introduce errors if the charge-discharge rate changes during the charge-discharge cycle. Including the average current in the machine learning model based on the normalized derivative values ​​can also offset any loss of accuracy resulting from the normalization step and prevent the neural network from deriving information from the absolute value rather than the normalized derivative value.

[0063] In some embodiments, in addition to the determined, normalized rate of change, the temperature of the battery system over the duration of the charge-discharge operation can also be used as input to the neural network 140. In one embodiment, the true average temperature of the battery system over the charge-discharge operation can also be used as input to the neural network 140. Including temperature can also improve the accuracy of SoH by compensating for changes in battery system behavior resulting from temperature differences. Specifically, temperature fluctuations affect both the internal resistance of the battery system and the extent to which the battery system can be charged.

[0064] In some embodiments, post-processing can be performed on the output of the neural network 140 to remove outliers and unexpected noise spikes. In some embodiments, values ​​outside the expected parameters are removed. In one embodiment, where the trough of the SoH curve is outside the expected 40-60% SoC, a post-processing step can be implemented to remove this data from the curve, as shown in Figure 6B. By post-processing the SoH output in this manner, the loss of accuracy that occurs when using measurement data collected across a partial charge-discharge range can be compensated for. It will be understood that the post-processing step can also perform standard deviation analysis, threshold levels, and comparison with historical / training data at a particular point in time, as well as removal of outliers / noise from the determined dataset.

[0065] The determined State of Heat (SoH) of the battery system 110 can then be used to modify the effective operation of the battery system 110 (for example, by changing the charge / discharge operation to maintain the lifespan of the battery system 110), and to determine when the battery system 110 (or one or more of its components) needs to be replaced.

[0066] Figure 8 is a process flowchart illustrating a method 600 for determining the health of a battery system based on normalized characteristic derivatives measured according to embodiments of the present disclosure, using a neural network for training. Step 610 involves acquiring training data, which includes measurement data representing the battery system and associated battery system characteristics, during charge-discharge operations. The training data may include the rate of change expected in a field application.

[0067] In one embodiment, the neural network 140 can be trained using normalized battery system characteristics sampled from the MIT dataset. The training data set is available in the public domain -MIT-https: / / data.matr.io / 1 / .

[0068] Using the acquired training data, in step 620, the rate of change of the battery system's characteristics with respect to time or another characteristic of the battery system, observed across charge-discharge operations, is determined. Next, in step 630, the determined rate of change is rescaled / normalized using known techniques to standardize the training process and improve the efficiency of the neural network 140 in modified battery systems. Specifically, as already discussed, rescaling trains the neural network 140 to focus on the determined rate of change patterns, reducing the use of large values ​​in the implementation, thereby enabling higher-resolution fixed-point conversion. In some embodiments, a filtering step may be implemented to remove unexpected / outlier values, as described above.

[0069] In step 640, the neural network 140 is trained using the determined, normalized rate of change, weighting and setting parameters for handling the rate of change to determine the State of Heat (SoH) of the battery system. In some embodiments, parameter training can be set by Bayesian ruled backpropagation.

[0070] The neural network 140 can have a configuration size based on the application requirements. In one embodiment, the configuration of the base network may be [18 15 10 10 5 3]. In one embodiment, the neural network 140 uses the Tanh function to determine the State of Health (SoH) of a battery system. Activation / transfer function: Tanh - enables rapid convergence with a limited number of nodes, but other known activation functions may be used. The vanishing gradient mitigates the risk of exponential errors in untrained datasets. Alternative activation functions used in regression problems, such as the normalized linear unit (ReLU) function, require complex neural networks with an order of magnitude larger number of nodes to match the accuracy of predictions obtained from Tanh. Similar problems are expected with Leaky ReLU and switch activation functions. The limited range of Tanh (-1 to 1) allows for very precise conversions in fixed-point operations.

[0071] Next, the results of the trained neural network can be validated and tested in step 650 against battery system data using techniques known in the art. In some embodiments, validation of the neural network 140 can be performed on a portion of the entire MIT dataset (e.g., 15%). After the testing and validation steps, the neural network 140 can be optimized in step 660 by adjusting the weights and / or parameters to minimize errors in the battery system SoH determination.

[0072] The implemented algorithm, trained on dV / dQ values, has been proven to predict the State of Heat (SoH) of lithium iron phosphate (LFP) cells (A123 APR18650M1A 1.1Ah) with an RMSE of less than 1%, using absolute dV / dQ calculations obtained from Q Ah in the range of 0.35Ah to 0.70Ah. Retrained on the same MIT dataset using rescaled dV / dSoC v SoC data enabled accurate SoH prediction for other LFP cells (LithiumWerks APR18650M1B 1.2Ah), with a prediction error of less than 3% RMSE. The rescaled dV / dQ or dV / dSoC provides a more general solution due to its greater accuracy across all types of battery systems.

[0073] Advantageously, models generated in the manner described above can derive SoH measurements without needing to know the battery's previous state, as required in some prior art systems such as the DE 10 2015 016 987 A1. This reduces the system's requirement for non-volatile memory to store such state data, or other state variables needed to update the model. Historical data is not used in training. Rather, normalized data is directly linked to SoH values, along with average current and / or temperature, if required.

[0074] In one embodiment, the neural network can be trained in the cloud. The trained neural network 140 can then be transferred to an embedded device and function independently of the cloud. While training the neural network 140 in this way eliminates the requirement of continuous connection to the cloud for SoH decisions, in certain embodiments, the cloud connection can be used to assist in SoH decisions.

[0075] Although the above embodiments describe techniques for determining the State of Health (SOH) of a battery system defined with respect to capacity, it will be understood that such techniques are equally applicable to other definitions of SOH that are well known in the art. Specifically, in some embodiments, SOH can be defined as one of the following: -Power health can be defined as the ratio of the current maximum power that a cell can deliver during a given time to the maximum power that the cell can deliver at the time of manufacture; -The health of the fast charging capacity can be defined as the ratio of the current maximum charging speed to the maximum charging speed of a new cell; and - The health of a cell in terms of its internal resistance can be defined as the ratio of the internal resistance that has increased over time to the internal resistance at the time of manufacture.

[0076] The descriptions of the various embodiments of this disclosure are presented for illustrative purposes only and are not intended to be exclusive to or limit the scope of this disclosure. Many modifications and variations of the disclosed embodiments will be apparent to those skilled in the art without departing from the scope of this disclosure. The terms used herein to disclose the embodiments of this disclosure have been selected to best represent the principles of the embodiments, their practical applications or technical improvements to market-found technologies, or to enable those skilled in the art to understand the embodiments disclosed herein.

[0077] For clarity, certain features of the Disclosure described in the context of separate embodiments may also be provided in a combination of a single embodiment. Conversely, for brevity, various features of the Disclosure described in the context of a single embodiment may also be provided separately, in any preferred subcombination, or as preferred in any other embodiment of the Disclosure described. Certain features described in the context of various embodiments should not be considered essential features of those embodiments unless the embodiments would not function without those elements.

[0078] As this disclosure has been described in conjunction with its specific embodiments, it is evident that many alternatives, modifications, and variations will be apparent to those skilled in the art. Therefore, it is intended to encompass all such alternatives, modifications, and variations that fall within the spirit and broad scope of the appended claims. The inventions disclosed herein include the following: [Aspect 1] A monitoring system for determining the state of health (SOH) of a battery system comprising one or more battery cells, During charging or discharging operations, at least one sensor for measuring at least one characteristic associated with the battery system, A processing device that is communicatively connected to at least one of the sensors, The system receives measurement data from at least one of the sensors, and Based on the received measurement data, The normalized rate of change of the first measured characteristic to one of the second measured characteristics associated with the battery system, measured over time or during the charging or discharging operation. The preprocessor is configured to determine that the second measured characteristic is different from the first measured characteristic. The processing device comprises, A neural network configured to determine the state of health (SOH) of the battery system using the normalized rate of change determined above, A monitoring system equipped with the following features. [Aspect 2] The monitoring system according to embodiment 1, wherein the at least one sensor comprises at least one of a temperature sensor, a voltage sensor, a current sensor, or a combination of the voltage sensor, the current sensor, and / or the temperature sensor. [Aspect 3] The first characteristic to be measured is voltage, charge A monitoring system according to any one of embodiments 1 to 2, which includes one of the following: a charging state. [Aspect 4] The second characteristic to be measured is, charge A monitoring system according to any one of embodiments 1 to 3, including one of the following: charge status or temperature. [Aspect 5] A monitoring system according to any one of embodiments 1 to 4, wherein the first characteristic to be measured is voltage, and the second characteristic to be measured is the charge state. [Aspect 6] The first characteristic to be measured is voltage, and the second characteristic to be measured is charge A monitoring system as described in any one of the three embodiments. [Aspect 7] The monitoring system according to any one of embodiments 1 to 6, wherein the neural network comprises multiple layers, and the processing device determines the health of the battery system by sequentially firing each of the multiple layers. [Aspect 8] A battery system comprising a monitoring system as described in any one of embodiments 1 to 7. [Aspect 9] A processing device for use in a monitoring system described in any one of the embodiments 1 to 7, At least one sensor receives measurement data indicating at least one characteristic associated with the battery system during a charge or discharge operation, and Based on the received measurement data, The normalized rate of change of the first measured characteristic to one of the second measured characteristics associated with the battery system, measured over time or during the charging or discharging operation. The preprocessor is configured to determine that the second measured characteristic is different from the first measured characteristic, A neural network configured to determine the state of health (SOH) of the battery system using the normalized rate of change determined above, The processing device comprising the above. [Aspect 10] A method for using a computer to determine the health of a battery system, i.e., State of Health (SOH), At least one sensor receives measurement data indicating at least one characteristic associated with the battery system during a charging or discharging operation, Based on the received measurement data, The normalized rate of change of the first measured characteristic to one of the second measured characteristics associated with the battery system, measured over time or during the charging or discharging operation. The determination is that the second measured characteristic is different from the first measured characteristic, Using a neural network, the normalized and determined rate of change is used as input data to the neural network to determine the health of the battery system, i.e., the State of Health (SOH). Methods that include... [Aspect 11] The method according to embodiment 10, further comprising normalizing the determined rate of change. [Aspect 12] The method according to embodiment 10 or 11, wherein the neural network comprises a plurality of layers, and the method further comprises sequentially firing each of the plurality of layers to determine the health of the battery system. [Aspect 13] The first characteristic to be measured includes one of voltage, charge, or charge state, and the second characteristic to be measured is, charge The method according to any one of embodiments 10 to 12, comprising one of the following: charge state and temperature, preferably the first characteristic to be measured is voltage and the second characteristic to be measured is charge or charge state. [Aspect 14] A computer-readable medium that, when executed by a processor, includes instructions causing the processor to perform the method described in any one of embodiments 10 to 13.

Claims

1. A monitoring system for determining the state of health (SOH) of a battery system comprising one or more battery cells, During charging or discharging operations, at least one sensor for measuring at least one characteristic associated with the battery system, A processing device that is communicatively connected to at least one of the sensors, The system receives measurement data from at least one of the sensors, and Based on the received measurement data, the system is configured to determine the normalized rate of change of a first measured characteristic with respect to time, or the normalized rate of change of the first measured characteristic with respect to a second measured characteristic, wherein the second measured characteristic is related to the battery system measured during the charging or discharging operation, and the second measured characteristic is different from the first measured characteristic, with respect to the preprocessor. A shallow feedforward neural network configured to determine the state of health (SOH) of the battery system using the normalized rate of change determined above, The processing device comprises, A monitoring system equipped with the following features.

2. The monitoring system according to claim 1, wherein the at least one sensor comprises at least one of a temperature sensor, a voltage sensor, a current sensor, or a combination of the voltage sensor, the current sensor, and / or the temperature sensor.

3. The monitoring system according to claim 1, wherein the first characteristic to be measured includes one of voltage, charge, or charge state.

4. The monitoring system according to claim 1, wherein the second characteristic to be measured includes one of charge, charge state, or temperature.

5. The monitoring system according to claim 1, wherein the first characteristic to be measured is voltage, and the second characteristic to be measured is the charge state.

6. The monitoring system according to claim 1, wherein the first characteristic to be measured is voltage, and the second characteristic to be measured is electric charge.

7. The monitoring system according to claim 1, wherein the shallow feedforward neural network comprises a plurality of layers, and the processing device determines the health of the battery system by sequentially firing each of the plurality of layers.

8. A battery system comprising the monitoring system described in any one of claims 1 to 7.

9. A processing device for use in a monitoring system according to any one of claims 1 to 7, At least one sensor receives measurement data indicating at least one characteristic associated with the battery system during a charge or discharge operation, and Based on the received measurement data, the system is configured to determine the normalized rate of change of a first measured characteristic with respect to time, or the normalized rate of change of the first measured characteristic with respect to a second measured characteristic, wherein the second measured characteristic is related to the battery system measured during the charging or discharging operation, and the second measured characteristic is different from the first measured characteristic, with respect to the preprocessor. A shallow feedforward neural network configured to determine the state of health (SOH) of the battery system using the normalized rate of change determined above, The processing device comprising the above.

10. A method for using a computer to determine the health of a battery system, i.e., State of Health (SOH), At least one sensor receives measurement data indicating at least one characteristic related to the battery system during a charging or discharging operation, Based on the received measurement data, determine the normalized rate of change of a first measured characteristic with respect to time, or the normalized rate of change of the first measured characteristic with respect to a second measured characteristic, wherein the second measured characteristic is related to the battery system measured during the charging or discharging operation, and the second measured characteristic is different from the first measured characteristic. Using a shallow feedforward neural network, the normalized and determined rate of change is used as input data to the shallow feedforward neural network to determine the health of the battery system, i.e., the State of Health (SOH). Methods that include...

11. The method according to claim 10, wherein the shallow feedforward neural network comprises a plurality of layers, and the method further comprises sequentially firing each of the plurality of layers to determine the health of the battery system.

12. The method according to claim 10, wherein the first characteristic to be measured includes one of voltage, charge, or charge state, and the second characteristic to be measured includes one of charge, charge state, or temperature, preferably the first characteristic to be measured is voltage and the second characteristic to be charged or charge state.

13. A computer-readable medium that, when executed by a processor, includes instructions causing the processor to perform the method according to any one of claims 10 to 12.

Images