Method and apparatus for evaluating temperature measurements from temperature sensors distributed in a device battery

Abstract

The invention relates to a method for monitoring sensor drift of temperature sensors (23) which are arranged distributed in battery modules (21) of a device battery (2), comprising the following steps: - Measuring (S1) temperature readings from the temperature sensors (23); - Check (S2, S3, S4) each temperature sensor (23) for sensor drift; based on the measured temperature values; - if a sensor drift is detected for a temperature sensor (23), store (S5, S6) the sensor drift assigned to the temperature sensor (23) in question; and - Signal (S7, S8) when a sensor drift above a predetermined upper sensor error threshold and / or below a predetermined lower sensor error threshold is detected or predicted for one or more of the temperature sensors (23).

Description

Technical field

[0001] The invention relates to methods for determining signal drift and / or failures of temperature sensors in device batteries. Technical background

[0002] Monitoring the temperature of battery cells in a device battery is important because a temperature increase, especially in a localized area, can indicate a damaged battery cell. A temperature increase can pose a safety risk, as it can specifically point to a thermal runaway failure, which can develop rapidly and lead to a fire in the device battery.

[0003] High-performance batteries, such as those used in vehicles, typically incorporate a number of temperature sensors distributed across battery modules. A battery module often contains approximately eight to twelve cells with two to four spaced-apart temperature sensors, and a single battery can comprise between ten and twenty such modules.

[0004] The temperature measurements of the temperature sensors are monitored by a battery management system of the device battery, and a temperature sensor error is detected if a threshold value is exceeded or fallen below compared to all other temperature sensors in the device battery, or if a temperature sensor fails, for example due to a broken cable.

[0005] For error handling, a substitute value is usually used, which corresponds to an average of the temperature readings of all still functioning temperature sensors, or which corresponds to the temperature reading of another functioning temperature sensor in the same battery module.

[0006] Temperature sensor drift is not currently diagnosed; instead, it is treated by diagnosing an error if a threshold value, usually a lower or upper temperature threshold, is exceeded or fallen below.

[0007] Detecting signal drift from a temperature sensor would have the advantage of allowing for early notification and enabling the replacement of the affected temperature sensor or battery module during routine maintenance. This is because, upon detection of a defective temperature sensor, the battery manufacturer typically restricts the device's battery's lifespan. Such a replacement is particularly necessary because battery manufacturers usually limit the battery's use to a few cycles when a defective temperature sensor is detected, in order to mitigate the risk of failing to detect thermal runaway or other serious malfunctions.

[0008] In particular, increasing sensor drift of the temperature sensor can lead to a falsely detected failure of the temperature sensor and thus limit the usability of the device battery, without there actually being a fault.

[0009] It is therefore desirable to provide a method with which the temperature sensor drift of a temperature sensor in a device battery can be detected very early and very robustly, and a failure time can be determined so that a replacement time can be planned if necessary. Disclosure of the invention

[0010] According to the invention, a method for detecting a temperature sensor drift of a number of temperature sensors distributed in a device battery according to claim 1 and a corresponding device according to the dependent claim are provided.

[0011] Further details are specified in the dependent claims.

[0012] According to a first aspect, a method for monitoring sensor drift of temperature sensors distributed across battery modules of a device battery is provided, comprising the following steps: - Measuring temperature readings from temperature sensors; - Check each temperature sensor for sensor drift based on the measured temperature readings (also referred to as temperature values); - if sensor drift is detected for a temperature sensor, store the sensor drift assigned to the relevant temperature sensor; - Signal when a sensor drift above a predetermined upper sensor error threshold and / or below a predetermined lower sensor error threshold is detected or predicted for one or more of the temperature sensors.

[0013] Device batteries with multiple battery modules, each containing a large number of battery cells, are typically equipped with temperature sensors that allow the temperature distribution across the device battery to be monitored. For example, each battery module might be equipped with two temperature sensors positioned at a distance from each other on the battery cells.

[0014] Monitoring of the device battery with regard to a failure of a battery cell or battery module, or with regard to implausible values ​​of a temperature reading recorded by the temperature sensors, is carried out by means of a threshold comparison with a failure threshold, which only detects whether a temperature reading is too high or too low.

[0015] A malfunctioning temperature sensor is detected and signaled in the usual way. A missing temperature reading is then simulated as accurately as possible using other methods, such as averaging the temperature readings from the remaining temperature sensors or using a temperature reading from another sensor in the same battery module.

[0016] Another potential source of error is sensor drift in the temperature sensors, which can cause measured temperature values ​​to not correspond to the actual temperature. This can lead to temperature values ​​that are too high or too low when comparing the threshold with the corresponding failure threshold, resulting in a false detection of a temperature fault in a battery module.

[0017] By evaluating the temperature sensor readings according to the procedure described above, a defective battery module can be detected. Furthermore, developing sensor drifts of the temperature sensors and their progression can be identified early, and impending error detections resulting from threshold comparisons with upper and lower sensor error thresholds can be immediately determined. In particular, the early detection of sensor drift in the temperature sensor allows the user to be alerted to a necessary service visit in a timely manner, while the device battery can continue to operate, even if several temperature sensors in the same battery module exhibit significant sensor drifts that do not, however, lead to the detection of exceeding the upper sensor error threshold or falling below the lower sensor error threshold.

[0018] Furthermore, the method can be implemented in existing battery systems without any additional hardware modifications. In addition, the method enables the detection of temperature sensor drift even when there is a typical temperature variation across the installation space within the device battery.

[0019] In detail, the above procedure involves a cyclical or regular check of the temperature sensors in a device battery to monitor the respective sensor drifts and, if necessary, to detect a battery module fault where a battery cell or module heats up beyond a permissible level (deviating from the other battery modules). The detected sensor drifts are stored for each temperature sensor, and if a sensor drift occurs with a deviation above the upper or lower sensor fault threshold, a fault is reported, signaling the need for immediate or prompt replacement of the temperature sensor or battery module.

[0020] It may be possible to determine the sensor drift for one of the temperature sensors by - the measured temperature values ​​of each of the temperature sensors are compared with an upper and / or lower drift threshold for each temperature sensor adjacent to the temperature sensor in question, in order to obtain an indication of potential sensor drift for corresponding temperature sensor pairs if the upper drift threshold is exceeded or the lower drift threshold is not reached, wherein the upper drift threshold and the lower drift threshold depend on the temperature value of the respective adjacent temperature sensor in question. - a sensor drift of one of the temperature sensors is detected if the potential sensor drifts of all evaluated temperature sensor pairs have only one specific temperature sensor in common and no other temperature sensor pairs with potential sensor drift have been detected.

[0021] Sensor drift is detected according to the following procedure. First, the measured temperature values ​​of each temperature sensor are compared with an upper and / or lower drift threshold, which is determined based on a temperature value from a neighboring temperature sensor and its previously determined and stored sensor drift. The upper drift threshold, for example, results from the sum of the temperature value of the neighboring temperature sensor, its sensor drift, and a predefined positive temperature delta. The lower drift threshold, for example, results from the sum of the temperature value of the neighboring temperature sensor, its sensor drift, and a predefined negative temperature delta.The temperature delta is set so high that typical temperature gradients of several tenths of a degree Celsius, which can occur due to the conventional operation of the device battery, are not taken into account when detecting sensor drift. For example, the temperature delta can range between 0.5 °C and 2 °C.

[0022] This results in multiple comparison results for each temperature sensor from the pairwise threshold comparisons, depending on the number of directly adjacent temperature sensors. Potential sensor drift for such a temperature sensor pair is detected if the temperature value of the considered temperature sensor exceeds the upper drift threshold or falls below the lower drift threshold.

[0023] Sensor drift of a specific temperature sensor is detected when the potential sensor drifts of all evaluated temperature sensor pairs have only that particular temperature sensor in common, and no other temperature sensor pairs with potential sensor drift have been identified. It is assumed that sensor drift above the upper drift threshold or below the lower drift threshold can be detected for only one of the temperature sensors per evaluation cycle. Therefore, it is considered highly unlikely that sensor drift can be detected or occur simultaneously in two temperature sensors.

[0024] It may be provided that the positive and / or negative temperature delta is adjusted to the probability of a false detection of a defective temperature sensor.

[0025] The values ​​of the assumed positive and / or negative temperature delta can therefore be applied to increase the sensitivity of the method. This allows the applicable upper or lower drift threshold to be adjusted across devices. Initially, the drift threshold can be chosen to be sufficiently large to avoid false positives from a defective temperature sensor. During operation, the drift threshold can be reduced by adjusting the positive and / or negative temperature delta accordingly, once the first confirmed cases of sensor drift from temperature sensors occur. The drift threshold is designed so that a defined false-positive probability, such as 0.0003%, is not exceeded.

[0026] If sensor drift is detected for a temperature sensor, it is calculated by averaging the measured temperature values ​​of all neighboring temperature sensors, subtracting their assigned sensor drift. The sensor drift is then determined as the difference between the temperature value of the temperature sensor in question and the average of the temperature values ​​of the neighboring temperature sensors, adjusted for sensor drift. The sensor drift calculated in this way is assigned to the temperature sensor in question and stored.

[0027] The above method takes into account, in particular, possible temperature gradients that may occur in battery cells and battery modules during the operation of the device battery by selecting the upper and / or lower drift threshold.

[0028] Furthermore, the method also enables the detection of a battery module in which an excessive temperature increase occurs in one or more battery cells, even if at least two temperature sensor pairs, where threshold comparisons have each revealed a potential sensor drift, do not share a common temperature sensor. This is because a temperature increase leads to the heating of adjacent battery cells, resulting in multiple temperature sensors measuring a higher temperature value, for which potential sensor drifts are then detected.

[0029] In particular, a warning or instruction for action can be issued to the user of the device battery if a cell fault is detected.

[0030] Furthermore, a probabilistic data-based prediction model can be provided that is trained to map input variables to a sensor drift specification, wherein the input variables are an evaluation time point related to the start of the lifetime or the age of the temperature sensor in question, as well as one or more variables that include a cumulative temperature load, a temperature load variance, derived reliability variables and / or other aggregated state variables, in particular an aggregated power throughput, wherein the sensor drift specification either directly specifies the sensor drift for the temperature sensor in question or provides a value dependent on it, wherein a sensor drift specification is predicted depending on the probabilistic data-based prediction model, signalingif this results in a predicted sensor drift for one or more of the temperature sensors above a predetermined upper sensor error threshold and / or below a predetermined lower sensor error threshold.

[0031] It is therefore possible to evaluate the temporal dependence of the sensor drifts of individual temperature sensors using a probabilistic prediction model. The prediction model can be a data-driven regression model, specifically a Bayesian neural network or a Gaussian process model. The prediction model can be trained to map input variables to a sensor drift measurement. Input variables can include features such as the evaluation time relative to the start of the sensor's lifespan or its age, cumulative loads (e.g., the average temperature or the 50th percentile of the temperature load), and variance (e.g.,...).represented by the standard deviation or the 10% and / or 90% quantile of the temperature loads, as well as derived reliability parameters that can be provided based on domain knowledge of sensor aging and can take into account other aggregated state variables.

[0032] The probabilistic, data-driven prediction model can either directly determine or predict the sensor drift for the relevant temperature sensor or correct a drift value from a domain model, such as a physical sensor drift model used to calculate or predict sensor drift. The physical sensor drift model can preferably be implemented as a rainflow model, in which loads are counted by amplitudes or load spectra and derived as stress factors, which then serve as model input.

[0033] The sensor drift information obtained from the evaluation of the probabilistic prediction model can then correct the physical sensor drift model, e.g. additively or multiplicatively, and in the case of a Gaussian process model as a prediction model, it utilizes the positive property that, in the case of extrapolation, it returns to its prior of zero.

[0034] It may also be provided that an uncertainty in the prediction of the sensor drift value is determined by the probabilistic data-based prediction model, whereby an instruction to a user of the device battery or a restriction of the operation or function of the device battery is signaled depending on the sensor drift value and the uncertainty.

[0035] In addition to regression, the probabilistic prediction model can also provide a confidence level that indicates the uncertainty in the predicted sensor drift value. This uncertainty represents the risk that the prediction will not be accurate. Based on the probability density function, or the model confidence level, a probability can be directly calculated by integrating over the density function. This probability quantifies the risk that, at a specific future evaluation date, the upper or lower sensor error threshold will be reached due to the sensor drift of a given temperature sensor, necessitating its replacement or the replacement of the affected battery module. Thus, the risk of miscalculation—that a prediction using the prediction model will not accurately reflect the actual sensor drift—can be assessed.

[0036] Instructions for replacing a battery module or a temperature sensor can then be provided depending on a sensor drift estimate for a given prediction horizon and the uncertainty of the estimate.

[0037] The data-driven prediction model can be trained, for example, using a large number of temperature sensors from a large number of device batteries. This preferably takes place in a central unit external to the device, to which the corresponding temperature values ​​from the temperature sensors are regularly transmitted.

[0038] Based on the result of the procedure, an action recommendation can be issued. For example, immediate maintenance can be recommended if the currently stored sensor drift of one of the temperature sensors exceeds or falls below the predefined upper or lower drift threshold. Furthermore, it can be indicated that there is no sensor drift exceeding or falling below the predefined upper or lower drift threshold and that no immediate replacement or repair is necessary.

[0039] Alternatively, the system can indicate that the temperature sensors are functioning correctly, but that the sensor drift prediction shows that an upper / lower sensor error threshold will be exceeded / fall below a specified uncertainty within a given prediction horizon, and therefore maintenance or a module replacement should be scheduled soon or within a specified timeframe. The timing of the maintenance or module replacement can be determined by the point at which a temperature sensor's drift exceeds / falls below the upper / lower sensor error threshold.

[0040] It may be provided that a temperature fault in the device battery is detected if potential sensor drifts of all evaluated temperature sensor pairs have a specific temperature sensor in common and further temperature sensor pairs with potential sensor drift that do not include the specific temperature sensor have been detected, whereby in particular the temperature fault is signaled as an instruction to a user of the device battery or as a restriction of the operation or function of the device battery. Brief description of the drawings

[0041] The embodiments are explained in more detail below with reference to the accompanying drawings. These show: Fig. 1 a schematic representation of a technical device with a device battery and a battery management system for monitoring temperature sensors; Fig. 2 a schematic representation of a device battery with battery modules, each containing several battery cells; and Fig. 3 A flowchart illustrating a procedure for monitoring the temperature sensors in the device battery. Description of embodiments

[0042] Fig. Figure 1 shows a schematic representation of a technical system 1 with a device battery 2. The device battery 2 is connected to a battery management system 3 or a control unit to carry out the procedure described below.

[0043] As in Fig. As shown in Figure 2, the device battery 2 comprises battery cells 22, such as lithium-ion cells, arranged in several battery modules 21. The battery modules 21 are electrically connected to the device battery 2 by parallel and / or series connection in a manner known per se and are compactly interconnected. Each battery module 21 has several temperature sensors 23, in this example two temperature sensors 23, to monitor the temperature of the battery modules 21 and the battery cells 22.

[0044] The temperature sensors 23 are connected to the battery management system 3, which monitors the temperature of the device battery 2. The battery management system 3 checks the temperatures in the device battery 2 at regular intervals, such as every 5 to 1,000 seconds. For this purpose, the temperature values ​​from each of the temperature sensors 23 are read. The battery management system 3 is configured to signal a fault in a battery module 21 if a temperature sensor 23 located in the battery module 21 indicates a temperature value that exceeds a predefined failure threshold.

[0045] To avoid false detections, it is necessary to distinguish temperature errors in the battery modules 21 from sensor errors, which manifest themselves through sensor drift and which can lead to a false detection of a temperature error.

[0046] Based on the read-out temperature values, the procedure for monitoring the sensor drift of the temperature sensors 23 is executed. This procedure is described using the flowchart of the Fig. 3 described in more detail. The method can be implemented in hardware and / or software, in particular near the battery modules 21 in a control unit or the battery management system 3.

[0047] In step S1, as described above, the temperature values ​​of all temperature sensors 23 are read or recorded at a current time.

[0048] In step S2, the temperature sensors 23 are examined sequentially. The temperature reading of each temperature sensor 23 is compared with corresponding upper and lower drift thresholds for each temperature sensor 23 that is directly adjacent to it in terms of the configuration of the device battery 2. The upper and lower drift thresholds are derived from the temperature reading of the adjacent temperature sensor. The upper drift threshold is calculated by adding the temperature reading of the adjacent temperature sensor to its associated sensor drift and a predefined positive temperature delta. Similarly, the lower drift threshold is calculated by adding the temperature reading of the adjacent temperature sensor to its associated sensor drift and a corresponding negative temperature delta. Different values ​​of the positive and negative temperature deltas can also be used for the upper and lower drift thresholds.Negative temperature deltas are planned.

[0049] The system identifies which pairings of the considered temperature sensor 23 and the adjacent temperature sensor 23 result in an exceedance of the upper drift threshold or a fall below the lower drift threshold. Potential sensor drifts are then determined for these identified pairings.

[0050] In step S3, it is checked whether a potential sensor drift has been detected for at least one temperature sensor pair. If this is the case (alternative: Yes), the procedure continues with step S4. Otherwise (alternative: No), it returns to step S1.

[0051] In step S4, it is checked whether all temperature sensor pairs for which a potential sensor drift has been detected include a specific temperature sensor. If this is the case (alternative: Yes), the specific temperature sensor exhibits sensor drift and the procedure continues with step S5.

[0052] If at least two temperature sensor pairs exist that do not share a common temperature sensor (alternative: No), a temperature fault in one battery cell can be inferred, as a heating of a region of the device battery 2 can be assumed. In this case, the procedure continues with step S10.

[0053] In a subsequent step S5, the sensor drift for the specific temperature sensor 23, for which sensor drift has been detected, is determined. The sensor drift results from calculating the temperature difference, which is the difference between the temperature value of the specific temperature sensor 23 and the mean temperature value of the neighboring temperature sensors, adjusted for their respective sensor drift. The assigned sensor drift corresponds to the sensor drift stored for the respective neighboring temperature sensor.

[0054] Subsequently, in step S6, the sensor drift determined in this way is reassigned to the temperature sensor and saved accordingly.

[0055] Furthermore, the determined sensor drifts for each of the 23 temperature sensors can be provided as training datasets for training a probabilistic, data-driven prediction model. A training dataset includes an evaluation time point related to a commissioning point (BoL) or an age of the respective temperature sensor, the cumulative loads to which the temperature sensor was exposed, such as an average temperature, a 50th percentile of the temperature load, information on the variance of the temperature loads, and other aggregated operating parameters of the device battery, such as Ah throughput, and derived reliability parameters, etc. The stored sensor drift is also assigned as a label to the training dataset. Training can be performed regularly, ensuring that an up-to-date prediction model is always available.

[0056] The probabilistic prediction model can be trained as a Gaussian process model or a Bayesian neural network at regular intervals based on sensor drift from a large number of temperature sensors from a large number of device batteries, determined at specific evaluation times. This results in a certain predictable behavior of the sensor drift of the temperature sensors based on their loads.

[0057] By using the probabilistic prediction model in step S7, a future value of the sensor drift can be predicted with a quantifiable uncertainty (confidence). Based on the predicted value of the sensor drift, it can then be determined when the sensor drift exceeds a predefined sensor error threshold in magnitude. For this purpose, the prediction time can be successively shifted into the future as an input variable of the prediction model, and the corresponding sensor drift and the uncertainty of the prediction can be determined.

[0058] In step S8, depending on the prediction time, the predicted sensor drift, and the associated uncertainty, an instruction for action can be signaled.

[0059] The instruction may specify a time at which the temperature sensor in question, or the battery module in which the temperature sensor is installed, should be replaced. This time may be determined based on the predicted time, the predicted sensor drift, and the associated uncertainty.

[0060] In step S10, a warming of an area of ​​the device battery 2 was detected. A corresponding instruction is then signaled, indicating a fault in the battery module. This instruction may include a request to immediately shut down the device or to reduce its power output. Alternatively, the immediate shutdown of the device or the reduction in power output can be signaled and implemented automatically.

[0061] The positive and / or negative temperature delta can be fixed or adaptively adjusted to increase the sensitivity of the method for detecting sensor drift. Starting with an initially high temperature delta of, for example, between 1 °C and 3 °C, the temperature delta is gradually reduced over the course of the device's battery operation until the first instances of detected temperature sensor drift are observed. The temperature delta is designed to ensure that a defined false-positive probability, for example, between 0.001% and 0.01%, is not exceeded.

Claims

[1] Method, in particular a computer-implemented method, for monitoring sensor drifts of temperature sensors (23) which are distributed in battery modules (21) of a device battery (2), comprising the following steps: - measuring (S1) temperature readings from the temperature sensors (23); - Checking (S2, S3, S4) each temperature sensor (23) for sensor drift based on the measured temperature values; - if a sensor drift is detected for a temperature sensor (23), storing (S5, S6) the sensor drift associated with the temperature sensor (23) in question; and - Signaling (S7, S8) when a sensor drift above a predetermined upper sensor error threshold and / or below a predetermined lower sensor error threshold is detected or predicted for one or more of the temperature sensors (23). [2] Method according to claim 1, wherein the sensor drift for one of the temperature sensors (23) is determined by - the measured temperature values ​​of each of the temperature sensors (23) are compared with an upper and / or lower drift threshold value for each temperature sensor (23) adjacent to the respective temperature sensor (23) in order to obtain an indication of a potential sensor drift for corresponding temperature sensor pairs if the upper drift threshold value is exceeded or the lower drift threshold value is undershot, wherein the upper drift threshold value and the lower drift threshold value depend on the temperature value of the respective adjacent temperature sensor (23), - a sensor drift of one of the temperature sensors is detected if the potential sensor drifts of all evaluated temperature sensor pairs have only exactly one specific temperature sensor (23) in common and no further temperature sensor pairs with potential sensor drift have been detected. [3] Method according to claim 2, wherein the upper drift threshold value results from the sum of the temperature value of the respective adjacent temperature sensor (23), its sensor drift and a predetermined positive temperature delta and the lower drift threshold value results from the sum of the temperature value of the respective adjacent temperature sensor (23), its sensor drift and a predetermined negative temperature delta. [4] Method according to claim 3, wherein the positive and / or negative temperature delta is adapted to a probability of false detection of a defective temperature sensor (23). [5] Method according to one of claims 2 to 4, wherein a temperature error in the device battery (2) is detected if potential sensor drifts of all evaluated temperature sensor pairs have a specific temperature sensor (23) in common and further temperature sensor pairs with occurred potential sensor drift, which do not contain the specific temperature sensor, have been detected, wherein in particular the temperature error is signaled as an instruction to a user of the device battery (2) or as a restriction of the operation or function of the device battery (2). [6] Method according to one of claims 1 to 5, wherein, if a sensor drift is detected for one of the temperature sensors (23), this is calculated by averaging the detected temperature values ​​of all temperature sensors (23) adjacent to the temperature sensor (23) in question, less their associated sensor drift, and the sensor drift is determined as the difference between the value of the temperature indication of the temperature sensor in question and the mean value, wherein in particular the sensor drift thus determined is assigned to the temperature sensor in question and stored. [7] Method according to one of claims 1 to 6, wherein a probabilistic data-based prediction model is provided which is trained to map input variables to a sensor drift indication, wherein an evaluation time related to a start of the service life or the age of the temperature sensor (23) in question, as well as one or more variables are used as input variables, which comprise a cumulative temperature load, a dispersion of the temperature load, reliability variables derived therefrom and / or other aggregated state variables, in particular an aggregated power throughput, wherein the sensor drift indication either directly indicates the sensor drift for the temperature sensor in question or provides a value dependent thereon, wherein a sensor drift indication is predicted depending on the probabilistic data-based prediction model, wherein it is signaled,if a predicted sensor drift above a predetermined upper sensor error threshold and / or below a predetermined lower sensor error threshold is predicted for one or more of the temperature sensors. [8] Method according to claim 7, wherein an uncertainty of the prediction of the sensor drift indication is determined by the probabilistic data-based prediction model, wherein an instruction to act is signaled to a user of the device battery (2) or a restriction of the operation or function of the device battery (2) depending on the sensor drift indication and the uncertainty. [9] Device, in particular battery management system, which is designed to carry out the method according to one of claims 1 to 8. [10] Computer program arranged to carry out the method according to one of claims 1 to 8. [11] A machine-readable storage medium on which the computer program according to claim 10 is stored.

Images