CN115219931A - Method and device for operating a system for providing a predictive aging state of an electrical energy store - Google Patents
Method and device for operating a system for providing a predictive aging state of an electrical energy store Download PDFInfo
- Publication number
- CN115219931A CN115219931A CN202210395067.4A CN202210395067A CN115219931A CN 115219931 A CN115219931 A CN 115219931A CN 202210395067 A CN202210395067 A CN 202210395067A CN 115219931 A CN115219931 A CN 115219931A
- Authority
- CN
- China
- Prior art keywords
- model
- aging
- data
- state
- aging state
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000032683 aging Effects 0.000 title claims abstract description 366
- 238000000034 method Methods 0.000 title claims abstract description 92
- 230000008859 change Effects 0.000 claims abstract description 39
- 230000008569 process Effects 0.000 claims abstract description 35
- 230000002123 temporal effect Effects 0.000 claims description 91
- 238000012549 training Methods 0.000 claims description 53
- 238000012937 correction Methods 0.000 claims description 42
- 238000012360 testing method Methods 0.000 claims description 11
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000004836 empirical method Methods 0.000 claims description 4
- 238000012986 modification Methods 0.000 claims description 4
- 230000004048 modification Effects 0.000 claims description 4
- 230000000306 recurrent effect Effects 0.000 claims description 4
- 238000009434 installation Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims 1
- 230000006399 behavior Effects 0.000 description 17
- 230000006870 function Effects 0.000 description 14
- 238000011156 evaluation Methods 0.000 description 13
- 210000004027 cell Anatomy 0.000 description 10
- 230000000875 corresponding effect Effects 0.000 description 10
- 230000001419 dependent effect Effects 0.000 description 9
- 238000005457 optimization Methods 0.000 description 7
- 230000000737 periodic effect Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 230000004927 fusion Effects 0.000 description 5
- 230000014759 maintenance of location Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 239000003792 electrolyte Substances 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 239000000446 fuel Substances 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 230000001932 seasonal effect Effects 0.000 description 4
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 3
- 238000004146 energy storage Methods 0.000 description 3
- 229910052744 lithium Inorganic materials 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 241001596784 Pegasus Species 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 2
- 239000011149 active material Substances 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000033228 biological regulation Effects 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000007599 discharging Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000007086 side reaction Methods 0.000 description 2
- 239000002904 solvent Substances 0.000 description 2
- 230000035882 stress Effects 0.000 description 2
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000003679 aging effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013531 bayesian neural network Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 239000006182 cathode active material Substances 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000001143 conditioned effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000008021 deposition Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000011234 economic evaluation Methods 0.000 description 1
- 230000005518 electrochemistry Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000008929 regeneration Effects 0.000 description 1
- 238000011069 regeneration method Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Secondary Cells (AREA)
Abstract
A computer-implemented method of predicting a state of aging of a technical equipment accumulator or modeling a current state of aging, comprising: providing or receiving an energy accumulator operation parameter time variation process; providing a data-based aging state model which is trained to assign the modeled aging state of the energy store according to the time course of the operating variable; checking whether a data gap exists in time in the time change process of the operation parameter of the energy accumulator; when a data gap is determined, generating an operation parameter time change process for the data gap duration by means of a use mode model, wherein the use mode model is constructed to provide the operation parameter time change process of the data gap and/or a load parameter from which the operation parameter can be derived for the data gap duration according to a use mode and calendar time description data; the aging state is predicted or the current aging state is determined by means of an aging state model on the basis of the operational variable time profile supplemented in the data gap by the generated operational variable time profile.
Description
Technical Field
The invention relates to an electrical device, in particular an electrically drivable motor vehicle, in particular an electric or hybrid vehicle, having an electrical energy store, which operates independently of a power grid, and to measures for determining the State of aging (SOH: state of Health) of the electrical energy store. The invention also relates to mobile and stationary electrical energy accumulator systems.
Background
Electrical devices and machines which operate independently of the electrical network, such as electrically driven motor vehicles, are supplied with energy by means of an electrical energy store (usually a device battery or a vehicle battery). The accumulator supplies electrical energy to operate these devices. In this context, energy converters with energy reserve, for example fuel cells with hydrogen storage tanks (Wasserstofftank), are also referred to as electrical energy stores.
The aging state of the energy store decays significantly over its service life, which has an effect on reducing the maximum storage capacity. The measure of the aging of the energy store depends on the individual load of the energy store, i.e. in the case of a vehicle battery of a motor vehicle, on the usage behavior of the driver, the external ambient conditions and the vehicle battery type.
Although the instantaneous aging state of the energy store can be determined on the basis of historical operating state changes by means of a physical aging state model, this model is not accurate in certain cases. This inaccuracy of conventional aging state models makes it difficult to predict the aging state change process. However, the prediction of the aging process of the energy store is an important technical variable, since an economic evaluation of the remaining value (Restwert) of the energy store is achieved by this prediction. Furthermore, the prediction of the course of the aging state of the energy store can be used to operate the energy store in an optimized manner, for example in order to be able to map the sought aging course.
Disclosure of Invention
According to the invention, a method for predicting the aging state of an electrical energy store according to claim 1 and a device in an electrically drivable apparatus according to the independent claims are provided.
Further embodiments are specified in the dependent claims.
According to a first aspect, a computer-implemented method for predicting the state of aging (SOH) of an energy store of a technical installation or for modeling a current state of aging is provided, wherein the method comprises the following steps:
-receiving a time profile of at least one operating variable of at least one energy store;
-providing a data-based aging state model, which is trained for assigning a modeled aging state of the energy store as a function of the time profile of the at least one operating variable;
-testing: whether a temporal data gap (Datenl üke) exists during the temporal change of the at least one operating variable of the energy accumulator;
-generating a temporal course of the at least one operating variable for the duration of the data gap by means of a usage pattern model (nutsungsmasquerodell) in the event of a determination of the temporal data gap, wherein the usage pattern model is designed to provide the temporal course of the at least one operating variable for the temporal data gap and/or at least one load variable for the duration of the data gap from a usage pattern and calendar time specification data (kalendarischen Zeitangabe), wherein the at least one operating variable can be derived from the at least one load variable;
predicting the aging state or determining the current aging state based on the temporal profile of the at least one operating variable and by means of the aging state model, wherein the temporal profile of the at least one operating variable is supplemented in the temporal data gap by the generated temporal profile of the at least one operating variable.
An energy store in the sense of the present description comprises a device battery and an energy converter system, wherein the energy converter system has an electrochemical energy converter with an energy carrier reserve (energy ä gervorrat), for example a fuel cell system with a fuel cell and an energy carrier reserve.
The state of aging of the electrical energy accumulator, in particular of the device battery, is usually not measured directly. This may require a series of sensors in the vicinity of the accumulator which makes the manufacture of such accumulators cost intensive and complex and may increase space requirements. Furthermore, no measurement method suitable for everyday use is available on the market for determining the state of aging in an energy store. The current aging state is therefore usually determined by means of a physical aging model in the device. Such physical aging state models are inaccurate in certain situations and typically have model deviations of up to 5% or more.
Furthermore, due to the inaccuracies of the physical aging model, the physical aging model can only account for the instantaneous aging state of the energy accumulator. In particular, the prediction of the aging state, which is dependent on the operating mode of the energy store, such as in the case of a system battery, on the number and size of the charge flows in and out and thus on the usage behavior and usage parameters, can lead to very inaccurate predictions and is not currently provided at the location of the energy store, for example at a vehicle battery in a vehicle.
In the case of a device battery as an electrical energy store, the State of aging (SOH) is a critical variable for describing the remaining battery capacity or the remaining battery charge. The aging state generally represents a measure for the aging of the electrical energy store, which measure describes the efficiency of the energy store. In the case of a device battery or a battery module or a battery cell, the state of aging can be described as Capacity Retention Rate (SOH-C) or as internal resistance increase (SOH-R). The capacity retention ratio SOH-C is illustrated as the ratio of the measured instantaneous capacity to the initial capacity of a fully charged battery. The relative change in internal resistance SOH-R increases as the battery pack ages.
It is highly desirable to provide user-specific and user-specific modeling and prediction of the aging state of the electrical energy store on the basis of an aging state model which uses the course of the operating variables starting from the time of commissioning in order to adapt the aging state, starting from the aging state at the time of commissioning, in each case time step by time step. The aging state model may be purely data-based, but may also be implemented as a hybrid data-based aging state model. Such an aging state model can be implemented, for example, in a central unit (cloud) and parameterized or trained by means of operating variables of a large number of energy stores of different devices which are connected in communication with the central unit.
The aging state model for determining the aging state of the electrical energy store can be provided in the form of a hybrid aging state model, i.e. in the form of a combination of a physical aging model and a data-based model. In the case of a hybrid aging state model, a physical aging state can be determined by means of a physical or electrochemical aging model and a correction value (beaufschlagen mit) can be applied to the physical aging state, in particular by addition or multiplication, the correction value being derived from a data-based correction model. The physical aging model is based on electrochemical model equations which characterize the electrochemical state of a system of nonlinear differential equations, are continuously calculated and mapped to the physical aging state for output as SOH-C and/or SOH-R. These computations may typically be performed in the cloud, for example, once per week. The calculation of the physical aging model is based on a time-integration method for solving a system of differential equations.
Furthermore, a correction model of the hybrid data-based aging state model can be constructed using a probabilistic regression model or an artificial intelligence-based regression model, in particular a gaussian process model, and can be trained for correcting the aging state obtained by the physical aging model. For this purpose, there is therefore a data-based correction model for correcting the state of aging of the SOH-C and/or at least one further data-based correction model for correcting the state of aging of the SOH-R. Possible alternatives to the gaussian process are other supervised learning methods, such as those based on random forest models, adaBoost models, support vector machines or Bayes' schen neural networks.
The prediction of the aging state is helpful, for example, if the remaining service life of the energy storage device is to be determined and evaluated, for example, with respect to warranty conditions or CO2 fleet regulations. In this case, the data-based aging state model can be continuously queried in conjunction with a predefined usage pattern, wherein the usage pattern specifies the manner of use and the manner of operation of the electrical energy store. For this purpose, starting from the current time point, a continuous generation of a future time profile of the artificial operating variable is required, wherein the physical aging model requires the future time profile of the artificial operating variable on account of the time integration method for solving the differential in order to model the predicted aging state. In particular, a predictive course of the aging state is determined from the current time. For this purpose, these operating variables are either directly dependent on the identified usage pattern or are generated on the basis of a change in the load variable derived from the usage pattern, wherein the change in the required operating variable is generated by the change in the load variable.
The use-mode model can thus provide a temporal profile of the at least one load variable, wherein the operating variable model is designed to generate a temporal profile of the at least one load variable from the temporal profile of the at least one load variable, wherein the at least one load variable is required on the input side for the aging state model.
Advantageously, this possibility of prediction uses a trained model of the aging state and usage patterns, thereby enabling a more accurate prediction of the aging state than in a pure extrapolation approach.
Since the equation for the electrochemical physical aging state is solved for each time integration method in the system of differential equations, it is necessary to: as input to a physical aging model that exists as a time series (Zeitreihe) or time varying process. In order to model the aging state of the energy store by means of a physical or electrochemical aging state model and to make an optional prediction by means of a data-based correction model (hybrid aging state model), it is necessary to: the time profile (time sequence) of the operating variable is provided at a relatively high frequency. The time profile of these operating variables must also be provided as seamlessly as possible for the required accuracy requirements, i.e. in order to determine the state of aging at the current point in time for the device battery: in particular, the time profile of the battery pack temperature, the battery pack current, the battery pack voltage and the state of charge is provided at the battery level.
The calculation of the electrochemical model together with the correction model is preferably carried out outside the device, since it is very computationally expensive and usually does not make the required processing capacity sufficient in a battery-operated device or in the hardware proximity (hardware-nah) to the battery-operated device or should not be provided for cost reasons. Furthermore, the electrochemical physical aging model can be retrained from time to time (e.g. every 6 months) based on newly generated tags in a central unit (cloud) outside the device with the aid of data of a large number of accumulators.
Since the equations for the physical aging state are solved for each time-integration method in the system of differential equations, the inputs of the electrochemical aging model, which exist as a time series, are required. The inputs may include, for example, temperature, current, and voltage. For capacity reasons, the time profile of the operating variable is transmitted to a central unit outside the device, and the state of aging is determined at the central unit on the basis of a physical aging model (electrochemical model) and, if necessary, a correction model.
However, data gaps in the time series of the operating parameters may occur for a variety of reasons. For example, signal gaps which last longer and which are to be filled in a simple manner may occur as a result of longer parking periods and/or sporadic use of the device. Information about the process of running parameter changes may be lost due to connectivity problems (Konnektivit ä tsproblem), such as due to network communication interference, which results in an incomplete time sequence. In this connection, it is necessary: the temporal course of the operating variables in the occurring signal gaps is filled or reconstructed in order to be able to run the aging state model used for determining the current or future aging state. Thereby, the accuracy and robustness of the modeled and/or predicted aging state may be significantly increased. The method is realized as follows: the time profile of at least one operating variable required for determining the state of aging, in particular the state of charge of the device battery as an energy store, the battery temperature, the battery current and the battery voltage, is completed (Vervollst ä ndiggung).
The reconstruction is carried out by means of self-learning usage pattern models, which lead to an increasing accuracy of the reproduction as the number of data increases. The usage pattern model can fill in the gaps in the time profile of the operating variables by using reasonable assumptions and thus significantly increase the robustness of the aging state model by increasing the Input quality (Input-Qualit ä t) of the aging state model. In combination with a separate observation-based basic aging model for verification for determining the aging state in another way, the post-hoc reconstructed load pattern (lastmaster) can be evaluated and verified model-based.
The usage pattern model can be constructed or trained on the basis of the provided or received temporal course of the at least one operating variable in a device-specific manner.
The above method provides that: using a usage pattern model which generates a time profile of the at least one operating variable or contributes to the generation of the time profile of the at least one operating variable from one or more usage parameters of the usage pattern. The learned (erlernen) usage pattern is characterized by parameterization of the usage pattern model or trained model parameters. The usage pattern is used to model the time sequence of the at least one operating variable in a model-based manner by means of the usage pattern model. Thus, with the usage pattern model: the learned load pattern is subjected to a time-series prediction in the form of a time profile of at least one operating variable and/or at least one load variable, such as current or temperature.
The temporal course of the at least one operating variable is generated by means of a model of the use pattern. The usage pattern model is designed to output a continuous course of the at least one variable as a function of one or more usage parameters, wherein the one or more usage parameters explicitly or implicitly, i.e. by means of a data-based model, specify the usage pattern, wherein the at least one variable comprises at least one load variable which can correspond to the at least one operating variable or from which the at least one operating variable can be generated in a model-based manner. For a device battery as an energy store, the at least one load variable may correspond, for example, to a temporal profile of the battery temperature and/or the battery current. This makes it possible to convert the usage behavior parameterized by the usage pattern into a time profile of the at least one operating variable. The usage model can thus specify the load type of the energy store in the form of a time profile of the at least one load variable and/or of the at least one operating variable.
It can thus be provided that: the usage pattern (in particular predefined by the time series of the at least one load variable) is created on the basis of a data-based usage pattern model by means of historical usage behavior, wherein the created usage pattern is determined, in particular, for predicting the aging state.
The usage pattern model is preferably designed as a data-based model which characterizes the operating mode of the energy store. In this case, an artificial change process of the operating variables which leads to the same load on the energy store and thus to a similar and user-representative aging behavior is generated on the basis of the usage data reflected in the change process of the at least one operating variable with respect to the time specification data (date, clock time, day of the week, season, etc.) on its calendar.
Alternatively, the usage pattern model can be designed as a hybrid usage pattern model with a non-data-based model and a data-based usage correction model, wherein the non-data-based model provides a temporal course of the at least one operating variable for the temporal data gap as a function of operating characteristics and/or the at least one load variable for the time duration as a function of time specification data on a calendar, and the data-based usage correction model is used to correct the at least one load variable for the time duration and/or the temporal course of the at least one operating variable for the temporal data gap as a function of time specification data on the calendar.
The usage pattern is derived from model parameters of a usage pattern model, which can be constructed in a data-based manner, in particular as a recurrent neural network, for example bayesian LSTM (LSTM: long Short Term Memory) or other models using supervised learning methods, such as gaussian processes or alternatively using layers of interest (Attention-Layer). It is trained by: assigning time-specific data on the calendar to the at least one time profile of the at least one load variable and/or the at least one operating variable. The trained usage pattern model then assigns the corresponding course of change of the at least one load variable or of the at least one operating variable to the time specification data on the predefined calendar or generates the corresponding course of change of the at least one load variable or of the at least one operating variable in the case of correspondingly predefined time specification data on the calendar. The time specification data on the calendar may specify calendar dates and clock times and information to be derived therefrom, such as the day, season or month of the week, etc.
If a data gap in the temporal course of the operating variable is now determined, an artificial temporal course of the operating variable for the data gap is generated on the basis of a usage pattern model previously trained with the aid of the detected temporal course of the at least one load variable and/or the at least one operating variable.
The historical load profile (Lastprofile) in the form of the time profile of the at least one load variable and/or the at least one operating variable is used as training data, i.e. for example in the case of a device battery the current profile according to the time specification data on the calendar or the battery temperature according to the time specification data on the calendar.
If the time profile of the at least one load variable is provided by the use-mode model, an operating variable model can be used which modifies the at least one load variable and provides the modified load variable as the operating variable. The operating parameter model makes it possible to characterize the usage behavior in terms of the load of the energy store in a parametrizable manner and to fit (adapt) the course of the at least one operating parameter during operation. The operating parameter model may include at least one domain model or data-based model to learn non-linearities in usage behavior (nichltinearit ä ten), describe the non-linearities, and enable computation of the non-linearities for future prediction. For example, it is possible to take into account a changing usage behavior or effectiveness effect (wirkunggrad-Effekte) based on a continuously changing or seasonal capacity limitation, for example in the winter season when the user charges more frequently to achieve (absolverien) the same driving power.
By means of the usage pattern model and the operating parameter model, all available time sequences are used in a device-specific manner for training. By means of the usage pattern model, it is thus possible to learn implicitly (implinit) the usage pattern of all devices individually, such as a periodic pattern which enables differentiation, for example, on weekdays or weekends. In the determined data gap, the missing temporal course of the at least one operating variable can be reconstructed therefrom, for example, by means of the operating variable model, the state of charge course of the battery current, the battery voltage and the battery temperature being reconstructed from the artificially generated load variable course during a disturbed or longer standstill Phase. They can be used as inputs for the aging state model in order to use it with the aid of the time profile of the at least one operating variable, wherein the time profile of the at least one operating variable is composed of a time segment in which the time profile of the at least one operating variable is detected and a time segment in which the time profile of the at least one operating variable is artificially generated.
It can be provided that: if the device is inactive (inactiv) for at least one predetermined time duration and/or no information exists about the time profile of the operating variable due to connectivity problems, a data gap is determined. For example, a data gap during the time variation of the operating variable can be determined if the operating variable is not detected for a predetermined time duration, for example two cycles (as may occur without the use of technical equipment) or if a communication interference between the technical equipment and the central unit occurs. This can be recognized, for example, by assigning a time stamp or other operation-relevant variable to the operating variable, such as a kilometer reading in the case of a vehicle, so that discontinuities in the course of the time profile of the operating variable can be determined. During the duration of the shutdown situation, the operating variable is not detected, and even in this shutdown situation, the temporal course of the at least one load variable and/or of the at least one operating variable is reconstructed by the use-mode model. For example, the temperature variation process that also contributes to aging (in conjunction with the SOC state) over the calendar is reconstructed so that the physical aging model can be run as time-series data.
Other aging status models may also be provided in a central unit in connection with a large number of devices. The further aging state model can be designed as an observation (isobachend) in order to determine, based on the course of the time variation of the at least one operating variable, by empirical methods, such as Coulomb counting or internal resistance variation measurement, the path points to which the further aging states are respectively assigned, wherein the aging state trajectories for a plurality of energy stores are modeled together with confidence bands (Konfidenzband) for these path points, wherein the confidence bands describe the estimated accuracy of each path point. Furthermore, the usage pattern model may be adapted when the determined aging state of the further aging state model for the current or considered aging time point lies outside a confidence band for the current or considered aging time point.
The further observed aging state model can be used to verify the plausibility (plausibilieren) and to correct the temporal course of the at least one load variable and/or the temporal course of the at least one operating variable when the aging state of the aging state model for the current or considered predicted evaluation time interval lies outside the confidence band for the current or considered predicted evaluation time interval, i.e. when the reconstructed temporal course of the at least one operating variable and/or the temporal course of the at least one load variable does not match the individual measurement of the observed further aging state model.
In particular, the aging state can be determined by means of the further aging state model on the basis of electrical operating parameters of the energy store or of the device battery, for example during the charging process, in order to determine an aging state trajectory in the central unit on the basis of a large number of data points determined therefrom. The base model uses an observer-based algorithm in the management system of the energy storage and is able to fuse a plurality of historical observations of the aging state into a course of change of the trajectory of the aging state with statistical confidence. Other aging state models can also be constructed such that they can calibrate the aging-and temperature-dependent OCV characteristic (idling voltage characteristic) in the static phase in order to improve the calculation of the aging state.
By means of the physical aging model and, if appropriate, by means of a correction model and a usage pattern model for providing a temporal profile of the operating variable during a data gap, a deviation of the reconstructed profile of the at least one operating variable and/or the temporal profile of the at least one load variable or its simulated effect on the aging profile of the energy store from a real measured aging profile can be evaluated and used for iteratively improving the artificially generated temporal profile of the at least one load variable and/or the at least one operating variable.
In particular, if a deviation between the aging state trajectory modeled for a specific device battery and the aging state trajectory generated on the basis of the base model, which is generally determined for the device battery type, results in a deviation which exceeds a predefined tolerance threshold, a new temporal course of the at least one load variable and/or the at least one operating variable can be generated for the data gap. The generation of the at least one load variable and/or the temporal profile of the at least one operating variable may be determined by Sampling (Sampling) from a confidence band of the reconstructed temporal profile of the operating variable. This can be done by Interpolation if the duration of the data gap is less than the maximum duration and thus reconstructing the time profile of the at least one operating variable (for example, in the case of a device battery, battery temperature, battery current, state of charge and battery voltage). For example, the time profile can be interpolated purely on the basis of data below a maximum duration of, for example, 30s, since the temperature can only change to an insignificant extent in the meantime.
In this case, if the aging during a data gap, which is determined by means of the hybrid aging state model, is determined to be too small, the operating variables can be sampled from the part of the confidence band that leads to a higher aging effect, and vice versa. Preferably, a numerical (number) optimization method, for example a bayesian optimization method, can be used for this purpose in order to be able to reconstruct the reconstructed time course precisely with the intrinsic stress factors (inh ä renew Stressfaktoren). Alternatively, numerical optimization methods, such as Regula Falsi (Pegasus method), can also be used.
Alternatively or additionally, a further aging state model (described) can be provided in an observed manner, which is designed to determine the trajectory points, which are each assigned to an aging time point, by empirical methods, such as Coulomb counting or internal resistance change measurement, on the basis of the time profile of the at least one operating variable in order to model an aging state trajectory for a large number of energy stores, wherein the modeled aging state is determined by fusing the aging state of the aging state model for the current or observed aging time point, which is dependent on the determined aging state of the aging state model, and the other aging states of the further aging state model. The fusion can be performed taking into account the confidence of the aging state model used. In the hybrid aging state model, the confidence corresponds exactly to the uncertainty of the probabilistic correction model or the data-based correction model or to the uncertainty of the gaussian process model. The confidence of the observed other aging state model is estimated empirically from the combination of the individual observations and for the calculated model value (e.g., observed SOHC) taking into account the error propagation of the sensor (e.g., battery pack temperature sensor).
Furthermore, the energy store or the battery of devices can be operated as a function of a predicted course of the modeled state of aging (SOH), wherein the remaining service life of the energy store is signaled as a function of the predicted course of the modeled state of aging (SOH), in particular. Furthermore, the operating strategy of the energy storage device can be modified on the basis of the predicted remaining service life in order to positively influence the remaining service life. Possible measures include: an operational intervention limit, such as a Derating-limit (Derating-Limits) with respect to an allowable temperature of the battery pack, is adapted.
Furthermore, energy accumulators can be used for operating devices, such as motor vehicles, electric power-assisted vehicles, aircraft, in particular unmanned aircraft, tool machines, devices of entertainment electronics, such as mobile telephones, autonomous robots and/or domestic appliances.
Furthermore, the current and future usage behavior can be taken into account by updating the usage pattern model, which is based on the updating of the parameters of the operating parameter model in the case of the parameters of the operating parameter model being updated according to the calculated aging of the other aging state models.
According to a further aspect, an apparatus for predicting the state of aging (SOH) of an electrical energy store of a technical installation or for modeling a current state of aging is provided, wherein the apparatus is designed to:
-providing or receiving a time profile of at least one operating variable of at least one energy store;
-providing a data-based aging state model, which is trained for assigning a modeled aging state of the energy store as a function of the time profile of the at least one operating variable;
-testing: whether a temporal data gap exists during the temporal change of the at least one operating variable of the energy store;
in the event of a determination of the temporal data gap, a temporal profile of the at least one operating variable is generated for the duration of the data gap by means of a usage pattern model, wherein the usage pattern model is designed to provide the temporal profile of the at least one operating variable for the temporal data gap and/or at least one load variable for the duration of the data gap as a function of a usage pattern and calendar time specification data, wherein the at least one operating variable can be derived from the at least one load variable;
predicting the aging state or determining the current aging state based on the temporal profile of the at least one operating variable and by means of the aging state model, wherein the temporal profile of the at least one operating variable is supplemented in the temporal data gap by the generated temporal profile of the at least one operating variable.
Drawings
Embodiments are further elucidated below on the basis of the figures. Wherein:
fig. 1 shows a schematic illustration of a system for providing driver-specific and vehicle-specific operating variables for determining the state of aging of a vehicle battery in a central unit;
FIG. 2 shows a schematic diagram of the functional structure of a hybrid aging state model;
FIG. 3 illustrates a flow diagram showing a method for training a data-based aging state model; and
FIG. 4 shows a schematic diagram of the functional structure of a hybrid aging state model with usage dependent prediction of the aging state;
FIG. 5 shows a flow chart for explaining a method for determining the aging state in the event of a data gap during the time course of the operating variable; and
fig. 6 shows the course of the aging state trajectory at the aging time of the vehicle battery packs of a plurality of vehicles, together with the confidence bands.
Detailed Description
The method according to the invention is described below with reference to a vehicle battery as an electrical energy accumulator in a large number of motor vehicles as a generic device. In a motor vehicle, a data-based aging state model for the respective vehicle battery pack can be implemented in the control unit. The aging state model may be continuously updated or retrained in the central unit based on the operating parameters of the vehicle battery packs from the fleet. The aging state model is run in the central unit and is used to calculate aging and predict aging.
In a representative manner, the above examples represent a large number of fixed or mobile devices with an energy supply that is not dependent on the power grid, such as vehicles (electric vehicles, electric power-assisted vehicles, etc.), appliances, tool machines, household appliances, IOT devices, etc., which are in connection with a central unit (cloud) via a respective communication connection (e.g., LAN, internet, LTE/5G).
Fig. 1 shows a system 1 for collecting fleet data in a central unit 2 for creating and operating and evaluating an aging status model. The aging state model is used to determine the aging state of an electrical energy store, for example a vehicle battery or a fuel cell in a motor vehicle. Fig. 1 shows a vehicle fleet 3 with a plurality of motor vehicles 4.
One of the motor vehicles 4 is shown in more detail in fig. 1. These motor vehicles 4 each have a vehicle battery 41 as a rechargeable electrical energy accumulator, an electric drive motor 42 and a control unit 43. The control unit 43 is connected to a communication module 44 which is suitable for transmitting data between the respective motor vehicle 4 and the central unit 2 (the so-called cloud).
The motor vehicle 4 transmits to the central unit 2 an operating variable F which specifies at least a variable by which the state of aging of the vehicle battery pack is dependent or can be determined. The operating variable F can be a description of the instantaneous battery current, the instantaneous battery voltage, the instantaneous battery temperature and the instantaneous State of Charge (SOC) in the case of a vehicle battery 41, as well as the pack level, the module level and/or the battery level. These operating variables F are detected in a fast time-scale of 2Hz to 100Hz and can be transmitted regularly to the central unit 2 in uncompressed and/or compressed form. For example, the time series may be transmitted to the central unit 2 block by block at intervals of 10 minutes to several hours.
Depending on the operating variable F, operating characteristics M, which relate to the evaluation period, can be generated in the central unit 2 or in other embodiments also already in the respective motor vehicle 4. The evaluation period may be from hours (e.g., 6 hours) to weeks (e.g., one month) for the aging state determination. A new evaluation of the aging state is made for a common value of the evaluation period of one week, i.e. once per week for each vehicle battery.
These operational characteristics may include, for example: features relating to the evaluation period and/or cumulative features and/or statistical variables determined during the total service life. In particular, these operating characteristics may include, for example: electrochemical states, such as SEI layer thickness, changes in the cyclable (zyklisierbar) lithium due to anode/cathode side reactions, rapid absorption of electrolyte solvents, slow absorption of electrolyte solvents, lithium deposition, loss of anode active (aktiv) material and loss of cathode active material, internal resistance, histogram features (histogrammermal), such as temperature at the state of charge, charge current over temperature and discharge current over temperature, in particular multidimensional histogram data on the battery temperature distribution over the state of charge, charge current distribution over temperature and/or discharge current distribution over temperature, current throughput in ampere hours (stromdurcatatz), the total accumulated charge (Ah), average capacity increase during charging (in particular for charging processes in which the charge increases above a threshold share of the total battery capacity (e.g. 20%), charge capacity and differential capacity (differen kappazit 8978) zxq/dft 8978) or the maximum value of the power driven. These variables are preferably scaled in such a way that they characterize the real usage behavior as well as possible. For example, in the case of an accumulated charge per unit time (ladungsruccatz) (Ah), the normalization with SOHR (Normierung) is carried out in order to correctly delineate the poorer battery effectiveness for the same range (in km) for completion (Bew ä multiging). These operating characteristics M may be used in whole or in part for the methods described below.
Further specification data (Angaben) can be derived from the operating characteristic M and from the operating variable F: load patterns over time, e.g. charging and driving cycles, determined by using pattern N, such as fast charging at high amperage or braking process with regeneration or strong acceleration; the age of the vehicle battery pack; charge accumulated over time and discharge accumulated over time; a maximum charging current; a maximum discharge current; a charging frequency; an average charging current; average discharge current; power throughput in charging and discharging; (especially average) charging temperature; a (particularly average) dispersion of the states of charge, and the like. This temporal load pattern characterizes typical temporal usage behavior and can be used for prediction.
The State of Health (SOH) is a key parameter for describing the remaining battery capacity or the remaining battery charge. The aging state represents a measure for the aging of the vehicle battery pack or the battery module or the battery cell and can be described as a Capacity Retention Rate (SOH-C) or as an internal resistance increase (SOH-R). The capacity retention ratio SOH-C is illustrated as the ratio of the measured instantaneous capacity to the initial capacity of a fully charged battery. The relative change in internal resistance SOH-R increases as the battery pack ages.
The central unit 2 has a data processing unit 21, in which the method described below can be carried out, and a database 22 for storing data points, model parameters, etc.
An aging state model, which is based in particular completely or partially on data, is implemented in the central unit 2. The aging state model may be regularly used to determine the instantaneous aging state of the vehicle battery pack 41 based on the operating characteristics and/or the operating parameters. In other words, it is possible that: the state of ageing of the battery pack 41 of the respective vehicle or of a module or an energy store of the battery belonging thereto is determined on the basis of operating variables and/or operating characteristics derived from the course of operating variable changes of one of the motor vehicles 4 of the vehicle fleet 3.
Fig. 2 schematically shows the functional structure of an embodiment of a hybrid data-based aging state model 9, which is embodied in a hybrid manner. The hybrid aging state model 9 includes a physical aging model 5 and a correction model 6. The physical aging model 5 and the correction model 6 obtain the operating parameters F or the operating characteristics M for the current evaluation period/aging time point (the lifetime of the vehicle battery pack from the commissioning operation time point). The operating characteristic M of the current evaluation time period/aging time point is generated in a characteristic extraction block (merkmals extraktionblock) 8 on the basis of the time sequence of the operating variable F.
The operating variables F are fed as time-series data (including temperature and current) directly into a (eingehen) physical aging state model 5, which is preferably embodied as an electrochemical model and describes the corresponding electrochemical state, for example the layer thickness (for example SEI thickness), the change in the recyclable lithium due to anode/cathode side reactions, the rapid consumption of the electrolyte, the slow consumption of the electrolyte, the loss of active material in the anode, the loss of active material in the cathode, etc., by means of nonlinear differential equations.
The physical aging model 5 may correspond to an electrochemical model of the battery cell and cell chemistry. This model determines the internal physical battery state as a function of the operating parameter F in order to determine a physically based state of aging SOHph in the form of at least one dimension of the above-mentioned electrochemical states, which are mapped linearly or nonlinearly to capacity retention rates (SOH-C) and/or internal resistance increase rates (SOH-R) in order to provide them. The physical aging model 5 can be parameterized by means of training data sets of the vehicles 4 from the fleet 3 and/or by means of laboratory data.
However, the model values for the physical state of aging SOHph provided by the electrochemical model are not accurate in certain cases and therefore specify: the model value is modified by a modification parameter k. The correction variable k is provided by a data-based correction model 6, which is trained by means of a training data set of the vehicles 4 from the fleet 3 and/or by means of laboratory data.
In order to determine the corrected state of aging SOH to be output, the outputs SOHph, k of the physical aging model 5 and of the correction model 6 are applied to one another, wherein the correction model is preferably implemented as a gaussian process model. In particular, these outputs SOHph, k may be added or multiplied (not shown) in an addition block 7 in order to obtain the modeled state of aging SOH to be output at the current evaluation time period or aging time point. The confidence of the gaussian process can be used in the case of addition also as the confidence of the modified aging value SOH to be output of the hybrid model.
The correction model 6 receives an operating characteristic M on the input side, wherein the operating characteristic is determined by the course of the change in the operating variable F and may also comprise one or more of the internal electrochemical states of the system of differential equations of the physical aging model. In addition, the correction model 6 can obtain the physical aging state SOHph obtained from the physical aging model 5 on the input side. The operating characteristic M of the current evaluation period is generated in the characteristic extraction block 8 on the basis of the time series of the operating variable F. These operating characteristics M also include: the internal states from the state vector, the electrochemical physical aging model 5 and advantageously the physical aging state (SOHph).
Other embodiments of the data-based aging state model are likewise possible, for example, the data-based aging state model can be designed as an unmixed, purely data-based model, which is designed on the basis of a probabilistic regression model or a regression model based on artificial intelligence, in particular a gaussian process model, or on the basis of a bayesian neural network. The model is trained to provide a modeled state of aging SOH on the basis of operating characteristic points determined by a current operating characteristic M of a current evaluation time period/aging time point, which operating characteristic M is determined in a characteristic extraction block 8 on the basis of a time series of the operating variable F.
For scaling (Skemering) and dimensionality reduction of the operating features, PCA (Principal component Analysis) can be used in order to correspondingly reduce the redundant linearly dependent information in the feature space before the modified model is trained (in an unsupervised manner). Alternatively, kernel-PCA (Kernal-PCA) can also be used, in order to be able to delineate non-linear effects also in the data complexity reduction. Normalization (normalization) of the total operating feature space (or principal component space) is carried out not only before dimension reduction but also specifically afterwards, for example in a Min/Max scale (Min/Max-Scaling) or Z-transformation.
The calculation of the state of aging and the prediction of the state of aging can therefore be carried out for an energy store having at least one electrochemical unit, for example a battery cell. The method can also be applied to the overall system of energy stores by means of rule-based and/or data-based mapping. Taking the battery as an example, the aging prediction can therefore also be applied directly to the module level and the pack level, in addition to the battery level.
A training data set is defined for training the hybrid aging state model. These training data sets can be collected from a large number of vehicles in the central unit 2. The physical state of aging model can be parameterized in a manner known per se by measuring the vehicle battery or its battery cells, while the correction model 6 can be trained in view of the deviation between the actual state of aging and the state of aging SOHph modeled by the physical state of aging model 5. The training data points thus map the operating characteristics of the operating characteristic points to the corresponding deviations between the actual state of aging and the state of aging SOHph modeled by the physical state of aging model.
Fig. 3 shows a flow chart for explaining a further method for training a hybrid aging state model in the central unit 2. To this end, the training data sets are divided into a training set and a test set. The training set is used to train the hybrid aging state model, while the test set is used to validate the hybrid aging state model by means of new, unknown data.
The determination of the aging state as a tag can be carried out in a manner known per se under defined tag-generating load conditions and ambient conditions by evaluating the operating variable progression in the vehicle or in the central unit 2 with an additional model, for example in the workshop on a test stand or in a diagnostic mode or tag-generating mode, which represents an operating mode and ensures compliance with predetermined operating conditions of the vehicle battery, for example constant temperature, constant current, etc. In this regard, other models for determining the state of aging may be used, for example based on an analysis of the charge and/or discharge phases identified in use of the battery pack. The SOH-C estimation is preferably performed by coulomb counting or by forming a current integral over time during the charging process, wherein said current integral over time is divided by the shift in state of charge (Hub) between the beginning and the end of the relevant charging and/or discharging phase. In this case, the idle voltage characteristic curve is advantageously calibrated in the quiescent phase in order to calculate the state of charge course in the central unit. If, during the charging process, the vehicle battery reaches a fully charged state from a fully discharged state of charge starting from a defined relaxed (relax) state under reproducible load and ambient conditions, sufficiently reliable specification data can be obtained for the state of aging used as a label, for example. The maximum charge thus detected may be correlated to an initial maximum charge capacity of the vehicle battery pack. The state of aging (SOH-R value) associated with the resistance can also be calculated from the voltage change associated with the current change. Typically they are associated with a defined time interval.
The training data set for the vehicle battery pack is therefore derived from the aging state signatures determined in each case at a time and the operating variable profile for the relevant vehicle battery pack up to this time. A plurality of training data sets at different points in time can be determined for the vehicle battery pack, wherein these points in time are preferably specified with respect to the beginning of the service life. The training data set is collected and provided for a large number of vehicles.
The aging state model can be trained in a conventional manner using the training data set, i.e. with the preservation of the physical aging model 5, the training data set is evaluated by means of the hybrid aging state model 9 and error measures, for example RMSE (relative mean square error) between the output values of the modeled aging state SOH of the training data set under consideration and the associated labels (loss functions), are used in a manner known per se for adapting and training the correction model 6. Provision is made here for: training is carried out in view of the residual error (residual) of the physical model, so that the correction model can be adapted accordingly to the data-driven (datengetrieben) in the data case (datenrange) that is allowed with sufficient confidence. The training data set then represents the training set.
Alternatively, the training is performed by: the training data set is divided into a training set and a testing set. The training set is used to train the hybrid aging state model, and the test set is used to validate the hybrid aging state model with new, not used for training, unknown data of the test set. Preferably, a third data set, the validation data set, is used in order to optimize the hyper-parameters of the modified model. Finally, the hybrid aging state model 9 is always tested on the basis of new data, wherein the Performance (Performance) of the hybrid aging state model 9 is verified on the basis of such an independent data set before Deployment (Deployment) in the central unit 2.
In step S1, the physical aging model 5 is parameterized according to a first part of the training set, in particular by parameter optimization by means of least squares or the like. The physical aging state SOHph is assumed here as the aging state of the corresponding training data set as the output of the physical aging model 5.
In step S2, the physical aging model is applied to the total training set of the hybrid model, i.e. the number of training data sets comprises at least the set of training data sets that have been used to parameterize the physical aging model or even exceeds the set of training data sets that have been used to parameterize the physical aging model. The error of the physical aging model 5 is accordingly rated in terms of the total error in terms of said residual error as histogram of model deviations (evaluieren). This residual in contains, in conjunction with the operating characteristic M or the operating variable F, all relevant information about the systematic weaknesses of the physical aging model 5. Information is also derived about how the physical aging model 5 behaves with respect to new sets of training data that are not parameterized for the physical aging model 5.
In a next step S3, the data-based modified model 6 is trained in view of the complete training set of hybrid models. Said training set of hybrid models comprises at least the training set of physical models corresponding to step S1. In order to train the correction model 6, not only the operating characteristics M are extracted from the operating variables F but also the internal states of the physical aging model 5 are used as a subset of M in order to map all operating characteristics to the error between the model prediction of the physical aging model (physical aging state) and the labeled aging state corresponding to the training data set. The correction model 6 can thus learn the weaknesses of the physical aging model 5, in order to thus be able to carry out a correction of the physical aging state in a correction block.
Training of the data-based correction model 6 may be performed using cross-validation and sequential bagging (Bootstrap aggregation) in order to improve robustness and accuracy. If the correction model is trained, the trained hybrid aging state model can be verified in step S4 by means of the test set, so that the overall performance calculated for the aging state can be verified.
The trained hybrid aging state model can now be used to determine the aging state based on the operating parameter F.
Training of the hybrid aging state model can always be initiated if new tagged data is available, particularly if the data contains new and relevant information. During operation in the central unit based on the fleet data, it is thus possible to continuously retrain the mixed aging state model for determining the aging state and predicting the aging state.
In order to predict the aging state from usage data, such as the usage pattern N of the vehicle driver, a model as shown in fig. 4 may be used. The trained hybrid aging state model according to fig. 2 may thus also comprise a feature extraction block as well as a dimension reduction block (e.g. using principal component analysis: PCA).
Fig. 4 shows the overall architecture of a system for determining the state of aging in the central unit 2, wherein the overall architecture is based on the hybrid state of aging model of fig. 2, wherein the operating variable model 10 is additionally used in order to generate, optionally manually, the temporal profile of the operating variables, for example the battery voltage U, the state of charge SOC, the battery current I and the battery temperature T, in the event of data gaps in the provision of the operating variables of the vehicle 4. The operating variable model 10 needs to be set, since the aging state model 9 requires a time series or a time profile of all operating variables F. The determination of the operating variable course of all operating variables is ensured by the operating variable model 10 and is carried out as a function of the state of aging SOH of the vehicle battery 41 and the temporal course of the load variable. The operating variable model may comprise at least one domain model or data-based model for determining at least one part of the operating variable from at least one or more of the load variables, in order to learn non-linearities in the behavior of use, to describe the non-linearities and to be able to calculate the non-linearities for future predictions. For example, it is thus possible to take into account the changing usage behavior or effect of the degree of effectiveness on the basis of continuously changing or seasonal capacity restrictions, for example in the winter season when the user charges more frequently to achieve the same driving power.
The operating parameter model 10 is generally designed to model the battery current and the battery temperature, as well as the battery voltage and the state of charge, taking into account the state of equilibrium of the electrochemistry (gleichgewichtstsund). The operating parameter model 10 uses an artificially generated course of changes of one or more load variables L, which characterize the load of the vehicle battery 41 as a result of the operation of the vehicle battery 41 over a longer period of time.
The artificially generated course of the one or more load variables L (here, the battery current and the battery temperature) is predefined by the usage pattern model 13. In order to generate a course of the load variable L for determining or predicting the state of aging, the usage pattern model 13 converts a predefined usage pattern N into a temporal course of the load variable L on the basis of the calendar time specification data, wherein the load variable reflects the load to which the vehicle battery pack 41 is subjected in the case of the usage and operating mode specified by the usage pattern. These usage patterns N thus result in the output of a temporal profile of the battery current I and the battery temperature T as load variables L by using the pattern model 13, with which the set of operating variables F with the profile of the battery voltage U and the state of charge SOC is completed by means of the operating variable model 10.
The usage pattern can be validated by operating the parameter model and, if necessary, iteratively improved in conjunction with the usage pattern model 13, in order to be able to reflect non-linearities in the predicted usage behavior, for example, a usage behavior that changes due to a changed degree of effectiveness or a particularly small capacity.
The usage pattern N is defined by usage parameters N which are learned in a vehicle-specific manner by the usage pattern model 13, preferably by means of a data-based method, and which are used to simulate usage behavior with respect to the drive train of the vehicle battery 41 in question or the user.
The usage pattern model 13 can be designed as a recurrent neural network, for example as LSTM (Long Short Term Memory) or GRU, in particular as a bayesian LSTM network, and is trained on the basis of a change in a load variable or operating variable which specifies the manner of use of the energy store. The load variable or operating variable change to be taken into account in this case is to be based on the same operating mode of the battery pack and the time period of the same use mode.
The operating mode model 13 may alternatively be designed as a hybrid model. For this purpose, a non-data-based model can be specified, wherein the non-data-based model provides the at least one load variable and/or the temporal course of the at least one operating variable as a function of one or more cumulative operating characteristics, such as the Ah throughput and the calendar time specification data, wherein the temporal course is modified accordingly by modifying the model based on the use of the data with regard to the operating characteristics. The data-based usage correction model can be designed as a recurrent neural network, such as LSTM or GRU, in particular as a bayesian LSTM network, and is trained on the basis of a change in the load variable or operating variable which specifies the manner of use of the energy store, for correspondingly correcting the temporal change in the at least one load variable and/or the at least one operating variable in accordance with the non-data-based model.
The usage parameters which describe the usage pattern then correspond to the model parameters of the usage pattern model, i.e. in the case of a neural network, the deviation values and the weights for the individual neurons. Furthermore, the prior distribution and the posterior distribution and the probability of being conditioned (kondationiern) based on the observation according to bayes theory can be considered as the correlation parameters.
The usage pattern is derived by training the usage pattern model 13 on the basis of known courses of change of the load variables and/or operating variables with respect to references on their own calendar. That is, the usage pattern model is trained in a manner known per se on the input side with time specification data on a calendar and on the output side with load variables (preferably as time-series currents, temperatures) and/or operating variables assigned to the corresponding time specification data. The manual process of the load variable and/or the operating variable can thus be generated by means of a trained usage pattern model by means of a predefined date and time of day. The time specification data on the calendar may also contain the day of the week, the month and knowledge about holidays and take into account the seasonality, in particular by Feature Engineering (Feature Engineering).
As long as no driver change is detected, the usage pattern model 13 is regularly trained, for example once a month, taking into account new data. Typically, the time segments not used for training but for verifying the usage pattern model are separated from the training data set in the middle of the segmentation. Typically, a bayesian LSTM network is used to characterize the current variation process in an individual way for the driver. Furthermore, bayesian LSTM networks are also used to characterize the temperature variation process in a driver-specific manner. An alternative to the LSTM approach is the ARIMAX model or gaussian process model, in particular for modeling temperature variation processes.
The use-mode model can thus be constructed directly from the raw data of the course of the change of the load variable L and/or of the change of the operating variable F. A typical pattern of the current profile (Stromprofil) is recognized and made reproducible for these vehicle battery packs 41, for example due to repeated commuting segments and typical parking and rest times and loads with respect to temperature ranges.
The usage pattern N can therefore implicitly specify the load type of the energy store, in particular the periodic load.
The usage pattern N can also describe, in particular, ambient conditions and periodic load changes. These ambient conditions can be derived, for example, from a climatic table, which specifies the course of the battery pack temperature variation within a day-night cycle for the season, etc., preferably by means of GPS-related weather data from a central unit (cloud). In this regard, the usage pattern model may be trained with temperature change course specification data in addition to time specification data on a calendar and used. Preferably, predictions of the temperature variation processes associated with GPS can be added (einarbiten) to the usage pattern prediction.
The temperature profile data can be derived from the average temperature over a period of time immediately past, for example over a month, wherein the temperature profile data can be predicted by means of seasonal fluctuations derived from climate tables. The climate table can be derived from position specification data (Geoposition) of the vehicle (vehicle position: position specification data of the most frequently determined vehicle position). The usage pattern model therefore defines a mapping of the calendar time specification data and temperature profile specification data to the change process of the load variable and/or operating variable as input variables and is correspondingly trained.
Furthermore, the usage pattern model 13 may be run according to the modeled state of aging SOH. Thus, for example, in the case of a vehicle operated with a vehicle battery pack: in order to drive the driver's desired distance, the driver is more likely to have to be charged three times per week in the case of an aging battery pack, rather than only twice as originally.
The usage pattern N is trained and predefined in a vehicle-specific manner and characterizes the usage and operating behavior or the mode of operation and use of the respective vehicle battery 41.
The one or more usage parameters of the usage pattern N may thus implicitly describe the load type of the vehicle's individual vehicle battery pack 41, in particular the periodic load with regard to the correlation with the time specification data on the calendar. By training the usage pattern model 13, a time sequence of the load variable L can be derived for the usage parameters of the usage pattern N, wherein the time sequence contains not only weekly periodic effects but also seasonal periodic effects.
By predicting the likelihood of the modeled state of aging SOH, an individual state of aging trajectory for the driver using pattern N can be created. The usage pattern N may be derived from stress factors and/or may be learned in an individual manner by the driver and based on data based on historical fleet data. Preferably, an autoregressive model or alternatively a deep learning method is used for this purpose for pattern recognition.
Next, the method for determining the aging state of the vehicle battery pack by means of the central unit 2 outside the vehicle is further explained according to the flowchart of fig. 5 and the functional block diagram of fig. 4. The method is preferably implemented completely in the central unit 2 and is implemented in the form of software and/or hardware in the central unit. The method is described next with reference to the previous explanation.
In step S11, the operating variable F is transmitted to the central unit 2 at regular time intervals. The operating variables F are transmitted from a plurality of vehicles 4 to the central unit 2 and are buffered in a vehicle-specific manner at the central unit.
In step S12, it is checked: whether the time series to be analyzed has data gaps. By means of the notch detection model 11: whether the time series is continuous or whether the changing process data of the operating parameter F is missing for one or more specific time periods. The identification of data gaps during the time profile of the operating variable F by the gap detection model 11 can be recognized, for example, by monitoring a time sequence of successive operating variables, such as time stamps or consumption variables, such as the distance traveled, in view of discontinuities or by determining sudden changes in the state of charge or other operating characteristics. If a temporal data gap is determined (decision: yes), the method therefore continues with step S13. Otherwise (decision no), the method continues with step S11.
In step S13, a data gap, which is derived from the operating variable model 10, is supplemented or filled by an artificially generated temporal profile of the operating variable F' by means of the input coupling unit 12. For this purpose, the operating parameter model 10 determines the course over time of the load parameters N (battery current and battery temperature) provided in a driver-specific manner, which are generated in the use-mode model 13.
As described above, the usage parameter N is determined by training the usage pattern model 13 using the past time course of the operating variable detected by the measurement technique, wherein the usage parameters are respectively assigned to the time specification data on the date. The time profile for the load variable L of the vehicle battery is then derived from the learned usage pattern and the time specification data on the calendar of the evaluation time period/time point for the assigned usage characteristic of the vehicle battery.
In step S14, these operating variables are now supplied to the hybrid aging state model. The hybrid aging state model, as described above, includes: a physical aging model 5, a characteristic model 8, which is designed to determine the operating characteristic M from the course of the change of the supplied operating variable F. The provided operating variables F include the operating variables F, F', which fill the data gap as described above.
The modeled state of aging SOH from the physical aging model 5 is corrected by means of correction variables from the data-based correction model 6. The correction model 6 corresponds to a data-based model which, in a correction block 7, corrects a correction variable k for the state of aging SOHph determined by the physical aging model, as a function of an operating characteristic M which results from the temporal course of the supplied operating variable. After filtering the correction variable k: a correction quantity k is applied to the modeled state of aging SOHph. In particular, the correction variable can be low-pass filtered or smoothed, in particular using a PT1 or PT2 element.
In order to verify the usage pattern model 13, an additional aging state model 18 of the observed type can be provided, which additional aging state model 18 evaluates the time profile of the actual measurement of the supplied operating variable F in order to determine an additional aging state SOH' from the operating variable F. When the corresponding operating state of the vehicle battery 41 is identified from the time profile of the operating variable F, the further aging state model 18 can determine, for example, a further aging state as a reference or observer model (beobachtermodel) on the basis of a method such as coulomb counting and/or a method for determining a change in the internal resistance. Data points can therefore be collected for a large number of vehicle battery packs 41, each having an aging state value and an associated aging time point for the vehicle battery pack 41, in particular for all vehicle battery packs 41 of a fleet. After possible smoothing and cleaning of the data points (Bereinigen), the aging state trajectory at the aging time point, which describes the course of further aging states SOH' of a large number of comparable vehicle battery packs for a fleet of vehicles, can be modeled by means of conventional regression methods, together with a confidence interval σ (standard deviation). This change process of the aging state trace is shown in fig. 6 for an aging state that depends on the change of the internal resistance. The confidence in the observed aging state model 18 is empirically calculated for the calculated model values (e.g., the observed SOHC) based on a combination of the individual observations and taking into account the error propagation of the sensors (e.g., the battery pack temperature sensors).
The further aging state model 18 determined and provided in the central unit is evaluated in step S15 in order to obtain a further aging state SOH' at the current aging point in time (life of the vehicle battery from the start of commissioning).
By using the further aging state model 18, the aging state determined by the hybrid aging state model can be validated.
It is therefore possible in step S16 to provide an aging time SOH for the vehicle battery pack considered in the hybrid aging state model, wherein this aging time SOH is compared in a comparison block 19 with the track points of the aging state trajectory at the current aging time considered. If it is determined in the comparison block 16: if the aging state SOH modeled by the hybrid aging state module and the further aging states determined by the further aging state model 15 deviate from one another by more than a confidence interval predefined for the further aging states in each case (decision: yes), the use pattern model 13 is adapted in step S17 in each case so that the reconstructed course of change of the at least one load variable L matches the individual observations of the further aging state model 18 as well as possible and a jump is made back to step S14. The hybrid aging state model can therefore be reused in order to determine the aging state SOH on the basis of a new temporal course of the supplied operating variable F, in which a data gap has been determined on the basis of the adapted usage pattern model 13.
If, for example, the further state of aging SOH' illustrates a smaller state of aging than the modeled state of aging SOH, the one or more, N usage parameters N can be changed in the direction of a smaller load in order to generate operating variables for filling the data gap therefrom. The procedure can be carried out iteratively, with the use of the mode parameters being varied incrementally in each iteration until the aging state SOH modeled by the hybrid aging state model and the further aging states determined by the further aging state model 15 deviate from one another by no more than a predefined confidence interval corresponding to the further aging states. Preferably, numerical optimization methods, for example bayesian optimization methods, can be used for this in order to be able to reconstruct the reconstructed time course precisely with the intrinsic stressors. Alternatively, numerical optimization methods, such as Regula Falsi (Pegasus method), may also be used.
If it is determined in step S16: the modeled state of aging SOH and the further state of aging SOH 'do not deviate from one another by more than the value specified by the confidence interval (decision no), the modeled state of aging SOH and the further state of aging SOH' are fused with one another in a fusion block 20 in step S18. This may be done, for example, by forming an average, forming a weighted average, etc., to provide a fused state of aging SOH Fus 。
In particular, a fusion method on the basis of regulation techniques, such as a particle filter, or a kalman filter, preferably a dual extended kalman filter, can be used in the fusion block 20. In the fusion block 20, two aging state trajectories, namely the trajectory of the modeled aging state from the start of the battery pack of the respective device and the further aging state trajectory, can thus be combined with the respective confidence intervals over time (zusammenf ügen), so that a fused trajectory of the aging state with the least possible uncertainty results.
The above-described method may be performed entirely or partially in the central unit 2.
The fusion trajectory based on the aging state makes it possible to operate the energy store in an optimized manner in order to be able to map out the aging profile sought.
Claims (15)
1. Computer-implemented method for predicting the state of aging (SOH) of an energy store (41) of a technical installation or for modeling a current state of aging (SOH), wherein the method comprises the following steps:
-providing or receiving (S11) a temporal course of at least one operating variable (F) of at least one energy store (41);
-providing an aging state model (9), in particular based on data, which is trained for assigning a modeled aging State (SOH) of the energy store (41) as a function of a time course of the at least one operating variable (F);
-testing: whether a temporal data gap exists during the temporal course of the at least one operating variable (F) of the energy store (41);
-generating a temporal course of the at least one operating variable (F) for the duration of the data gap by means of a usage pattern model (13) if the temporal data gap is determined (S12), wherein the usage pattern model (13) is designed to provide the temporal course of the at least one operating variable (F) for the temporal data gap and/or at least one load variable (L) for the duration of the data gap as a function of a usage pattern (N) and calendar time specification data (Z), wherein the at least one operating variable (F) can be derived from the at least one load variable;
-predicting the aging state or determining the current aging State (SOH) from the temporal course of the at least one operating variable (F) and by means of the aging state model (9) (S13, S14), wherein the temporal course of the at least one operating variable is supplemented in the temporal data gap by the generated temporal course of the at least one operating variable (F).
2. Method according to claim 1, wherein a data gap is determined if the device is inactive for at least one predetermined time duration and/or if no information exists about the time course of the at least one operating parameter (F) due to connectivity problems.
3. The method according to claim 1 or 2, wherein the usage pattern model (13) is configured as a data-based usage pattern model or as a hybrid usage pattern model with a non-data-based model and a data-based usage modification model (13), wherein the data-based usage pattern model is configured as a recurrent neural network, in particular a bayesian LSTM, a gaussian process or a layer of interest, wherein the non-data-based model provides a temporal variation process of the at least one operating parameter for the temporal data gap depending on an operating characteristic (M) and/or provides the at least one load parameter (L) for the duration depending on temporal specification data (Z) on a calendar, wherein the data-based usage modification model (13) is used to modify the at least one load parameter (L) for the duration and/or the temporal variation process of the at least one operating parameter (F) for the temporal data gap depending on the temporal specification data (Z) on the calendar.
4. The method according to one of claims 1 to 3, wherein the usage pattern model (13) is constructed or trained on the basis of a time course of the at least one operating parameter (F) provided or received in a device-specific manner.
5. Method according to one of claims 1 to 4, wherein a mode model (13) is used to provide a temporal course of the at least one load variable (L), wherein the operating variable model (10) is designed to generate a temporal course of the at least one load variable (L) from the temporal course of the at least one load variable (L), wherein the load variable is required on the input side for the aging state model (13).
6. The method according to one of claims 1 to 5, wherein a further aging state model (18) is provided, which further aging state model (18) is designed to determine, on the basis of the course of the time profile of the operating variable (F), by empirical methods, such as coulomb counting or internal resistance change measurements, the track points to which a further aging state (SOH ') is respectively assigned, wherein the aging state tracks for a plurality of energy stores (41) are modeled together with a confidence band for the track points, wherein the confidence band describes the estimated accuracy of each track point, wherein the usage pattern model (13) is adapted when the determined aging state (SOH') of the further aging state model (18) for a current or considered aging time point lies outside the confidence band for the current or considered aging time point.
7. Method according to one of claims 1 to 6, wherein a further aging state model (18) is provided, which is designed to determine, on the basis of the course of the time variation of the operating variable (F), data points which are each assigned a further aging state (SOH ') by empirical methods, such as coulomb counting or internal resistance variation measurement, in order to model an aging state trajectory for a large number of energy stores (41), wherein the modeled aging state is determined by fusing the aging State (SOH) of the aging state model (9) for a current or considered predicted aging time point, which depends on the determined aging state of the aging state model, and the further aging state (SOH') of the further aging state model (18).
8. Method according to one of claims 1 to 7, wherein the operating variables (F) correspond to the current (I), the voltage (U), the temperature (T) and the state of charge (SOC) of the battery of the device as an energy store (41).
9. The method according to one of claims 1 to 8, wherein the data-based aging state model (9) is configured as a hybrid model and comprises a physical aging model (5) and a trainable data-based correction model (6), in particular in the form of a regression model, preferably a Gaussian process, wherein the physical aging model is based on an electrochemical model equation and is configured for outputting a physical aging state (SOHph), wherein the correction model (6) is trained for correcting the physical aging state (SOHph) and provides a corrected physical aging state, in particular with a quantified uncertainty, as the modeled aging State (SOH).
10. The method of claim 9, wherein the data-based aging state model is trained on the basis of a training data set, wherein the training data set is divided into a training set and an extended total training set, wherein the aging model is parameterized with the training set, and wherein the modification model (6) is trained on the basis of the total training set, wherein the data-based aging state model (9) is tested on the basis of the total training set in order to determine the validity of the data-based aging state model.
11. Method according to one of claims 1 to 10, wherein the plant battery (41) is operated as a function of a predicted course of the modeled state of aging (SOH), wherein the remaining service life of the energy store (41) is signaled in particular as a function of the predicted course of the modeled state of aging (SOH).
12. Method according to any one of claims 1 to 11, wherein the energy accumulator (41) is used for operating devices, such as motor vehicles, electrically assisted vehicles, aircraft, in particular unmanned aerial vehicles, tool machines, devices of entertainment electronics, such as mobile phones, autonomous robots and/or household appliances.
13. Device for predicting the state of aging (SOH) of an energy store of a technical system or for modeling a current state of aging, wherein the device is designed to:
-receiving a time profile of at least one operating variable of at least one energy store;
-providing an aging state model, in particular based on data, which is trained for assigning a modeled aging state of the energy store according to a temporal course of the at least one operating variable;
-testing: whether a temporal data gap exists during the temporal change of the at least one operating variable of the energy store;
if the temporal data gap is determined, a temporal profile of the at least one operating variable is generated for the duration of the data gap by means of a usage pattern model, wherein the usage pattern model is designed to provide the temporal profile of the at least one operating variable for the temporal data gap and/or at least one load variable for the temporal data gap as a function of a usage pattern, wherein the at least one operating variable can be derived from the at least one load variable;
predicting the aging state or determining the current aging state based on the temporal profile of the at least one operating variable and by means of the aging state model, wherein the temporal profile of the at least one operating variable is supplemented in the temporal data gap by the generated temporal profile of the at least one operating variable.
14. Computer program product comprising instructions which, if the program is executed by at least one data processing device, cause the at least one data processing device to carry out the steps of the method according to any one of claims 1 to 12.
15. A machine-readable storage medium comprising instructions which, if executed by at least one data processing apparatus, cause the at least one data processing apparatus to perform the steps of the method according to any one of claims 1 to 12.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102021203729.1 | 2021-04-15 | ||
| DE102021203729.1A DE102021203729A1 (en) | 2021-04-15 | 2021-04-15 | Method and device for operating a system for providing predicted aging states of electrical energy stores using reconstructed time series signals using machine learning methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115219931A true CN115219931A (en) | 2022-10-21 |
Family
ID=83447058
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210395067.4A Pending CN115219931A (en) | 2021-04-15 | 2022-04-15 | Method and device for operating a system for providing a predictive aging state of an electrical energy store |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN115219931A (en) |
| DE (1) | DE102021203729A1 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115510964B (en) * | 2022-09-21 | 2023-05-26 | 浙江省科技项目管理服务中心 | Computer calculation method for liquid chromatograph scientific instrument |
| DE102023200585B4 (en) | 2023-01-25 | 2024-09-26 | Robert Bosch Gesellschaft mit beschränkter Haftung | Method and device for predictive diagnosis of a device battery of a technical device using a trace graph model |
| DE102023201255A1 (en) * | 2023-02-14 | 2024-08-14 | Volkswagen Aktiengesellschaft | Machine learning method and apparatus for predicting aging of a vehicle component |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9210662B1 (en) | 2014-05-29 | 2015-12-08 | Apple Inc. | Adaptive battery life extension |
| DE102018201119A1 (en) | 2018-01-24 | 2019-07-25 | Robert Bosch Gmbh | Method for monitoring the power supply of a motor vehicle with automated driving function |
-
2021
- 2021-04-15 DE DE102021203729.1A patent/DE102021203729A1/en active Pending
-
2022
- 2022-04-15 CN CN202210395067.4A patent/CN115219931A/en active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| DE102021203729A1 (en) | 2022-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11965935B2 (en) | Method and apparatus for operating a system for providing predicted states of health of electrical energy stores for a device using machine learning methods | |
| US11644515B2 (en) | Method and device for operating an electrically drivable motor vehicle depending on a predicted state of health of an electrical energy store | |
| US20220099743A1 (en) | Method and Apparatus for Operating a System for Providing Predicted States of Health of Electrical Energy Stores for a Device Using Machine Learning Methods | |
| CN115219931A (en) | Method and device for operating a system for providing a predictive aging state of an electrical energy store | |
| US20200164763A1 (en) | Predictive model for estimating battery states | |
| KR102297343B1 (en) | Battery Output Voltage Response and State-of-Charge Forecasting Method using Hybrid VARMA and LSTM | |
| KR20160000317A (en) | Method and device to learn and estimate battery state information | |
| JP2013019730A (en) | System, method, and program product for predicting state of battery | |
| US20230305073A1 (en) | Method and apparatus for providing a predicted aging state of a device battery based on a predicted usage pattern | |
| CN115219904A (en) | Method and device for operating a system for providing an electrochemical cell stack model | |
| CN115236541A (en) | Method and device for determining the state of aging of an electrical energy store by providing a state of aging model | |
| US20230306803A1 (en) | Method and apparatus for the user-dependent selection of a battery operated technical device depending on a user usage profile | |
| CN115219932A (en) | Method and device for evaluating the relative aging state of a battery of a device | |
| CN114325442A (en) | Method and device for robustly predicting aging characteristics of an accumulator in a machine | |
| CN116256658A (en) | Method and apparatus for providing device battery pack aging status using correction for status observations based on systematic status and environmental impact | |
| US20240168095A1 (en) | Method and Apparatus for Predictive Diagnosis of a Device Battery of a Technical Device Using a Multivariate Transformer Model | |
| Renold et al. | Comprehensive Review of Machine Learning, Deep Learning, and Digital Twin Data-Driven Approaches in Battery Health Prediction of Electric Vehicles | |
| CN114325397A (en) | Method and apparatus for training artificial intelligence based state models | |
| US20230324463A1 (en) | Method and Apparatus for Operating a System for Detecting an Anomaly of an Electrical Energy Store for a Device by Means of Machine Learning Methods | |
| US20230016228A1 (en) | Method and Apparatus for Operating a System for Providing Predicted Confidence Intervals for States of Health of Electrical Energy Stores for at Least One Device With the Aid of Machine Learning Methods | |
| CN115469224A (en) | Method and device for operating a system for providing a predicted state of ageing of an electrical energy store of a plant by means of a machine learning method | |
| CN116306214A (en) | Method and device for providing an ageing state model for determining the ageing state of an energy store | |
| CN115701545A (en) | Method and device for providing a calculated and predicted state of aging of an electrical energy store | |
| Heinrich | Battery State Estimation of Electric Vehicles using Neural Networks | |
| Mohanty et al. | Electric Vehicle State-of-charge Prediction using Deep LSTM Network Model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |