Predicting the health status of an electrical energy storage device
The method recalibrates lithium-ion battery endurance models using real-world data to improve prediction accuracy by updating model values, addressing the discrepancy between laboratory and operational conditions.
Patent Information
- Application Number
- FR2022014473
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-12-23
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-12-23
AI Technical Summary
Existing methods for predicting the state of health of lithium-ion batteries lack the ability to dynamically recalibrate endurance models during the life of the battery, leading to inaccurate prognoses due to discrepancies between laboratory data and real-world conditions.
A method for recalibrating endurance model parametric tables based on real operational data, using a recalibration coefficient to update the model values when deviations are detected, ensuring precise predictions of battery health.
Enhances the precision of battery health predictions by aligning model predictions with actual performance, accounting for real-world usage and temperature profiles, while minimizing computational and memory requirements.
Smart Images

Figure 00000032_0000 
Figure 00000032_0001 
Figure 00000032_0002
Abstract
Description
Title of the invention: Prediction of the health status of an electrical energy accumulator Technical field
[0001] The invention relates to the prediction of the state of health of an electrical energy accumulator, in particular a Lithium-ion (Li-ion) accumulator.
[0002] It finds applications, in particular, in the field of management of accumulator batteries which ensure the storage of electrical energy in chemical form. Prior art
[0003] Electrical energy storage batteries, which allow electrical energy to be stored using reversible electrochemical processes, are currently undergoing strong development. They are used in various types of applications, for example for electric or hybrid vehicles, and to support installations for generating electricity from renewable energies (in photovoltaic or wind power installations, for example). Lithium accumulators - in their various variants such as "lithium-ion", "lithium-ion-polymer", "lithium-metal-polymer" batteries, etc. - are the batteries with one of the highest stored energy density and specific energy. They are therefore the technology of choice for storing electrical energy in the aforementioned applications.
[0004] However, it is known that these accumulators exhibit a degradation in their performance over time. In particular, their storage capacity decreases and their internal resistance increases over time. These changes affect the performance of the electrical energy storage batteries incorporating them. However, in the type of applications mentioned above, the requirements for managing the energy resource, both when charging and discharging the batteries, require careful consideration of the degradation mechanisms of the accumulators. In other words, it is necessary to take into account the predicted evolution of the performance of the accumulators at different levels for the management of the electrical energy storage systems incorporating them (sizing, control strategies, maintenance, etc.).
[0005] For this purpose, it is known to use endurance models (those skilled in the art also speak of aging models), which allow the production of prognoses (also referred to as predictions, or estimations) of the state of health of an isolated accumulator, and generally of a battery of accumulators operationally arranged within an electrical energy storage system. These endurance models define parametric tables (also called maps) indexed by p-uplcts of parameters (where p is an integer strictly greater than unity). These maps give values of elements which enter into the definition of a parametric model, which models the degradation of the capacity of an accumulator, compared to its original performance (i.e. in new condition), according to determined parameters of its actual operation in the field.
[0006] An updated performance level of the accumulators can thus be determined by implementing indicators such as the SOH (State-Of-Health). Such an indicator is calculated or estimated from physical signal values stored for this purpose in the memory of a battery management system (or BMS). These physical signals can be measured continuously directly on the accumulator during operation, and / or be estimated from temporal data for monitoring the accumulators, such as voltage, current, and / or temperature, for example. The predicted evolution of the performance of the accumulators can be simulated at any time by means of a predefined endurance model, provided that the conditions (in particular temperature) and the profile of the stresses to which it has been subjected since the beginning are known.
[0007] There are several types of endurance models for predicting the state of health of an accumulator on the basis of a determined profile of the parameters of its operation during a period elapsed since a certain time reference, for example since the manufacture or commissioning of the accumulator, or since the performance of a previous diagnosis. Embodiments of the invention considered in the present description for purely illustrative purposes, use for example a type of model well known to those skilled in the art, namely the empirical model proposed in 2010 by Broussely in various scientific publications, for example the article by M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, K. Nechev, R. Staniewicz, “Aging mechanism in H ion cells and calendar life predictions”, Journal of Power Sources, 2001, volume 97, pp. 13-21.This type of model is parameterized using a database consisting of the results of battery endurance tests carried out in the laboratory during a preliminary study phase (also called an endurance test campaign or aging campaign). This database can be more or less extensive depending on the number of endurance tests carried out. In addition, the extent of the endurance test campaign is determined based on the number of battery operating parameters that can influence the battery performance and are the subject of the study.
[0008] We classically distinguish two modes of degradation of the capacity of accumulators over time, namely: - degradation known as "calendar degradation", on the one hand: this degradation occurs constantly during the life of the accumulator, whether it is subjected to electrical stress or not. The main parameters influencing the speed of this degradation are the temperature and the state of charge (or SOC, from the English "State -Of-Charge") of the accumulator; and, - degradation known as "cycling degradation", on the other hand: this degradation occurs when the accumulator is subjected to electrical stress, whether in charge or in discharge. It is linked to the quantity of electrical charges (expressed in Amperes / hour, or Ah) passed through the accumulator, in charge or in discharge.The main parameters influencing the speed of this degradation are the operating temperature, the state of charge (SOC), the demand current, and the depth of discharge (or DOD, from the English "Depth-Of-Discharge") also called SOC range (and often noted ASOC), of the accumulator during each electrical demand.
[0009] A method for empirically modeling the degradation of a battery consists of accumulating the respective effects of calendar degradation and cycling degradation, in order to obtain an estimate of the total degradation of the capacity of the accumulator. For this purpose, the aforementioned models can define parametric tables with two dimensions in calendar and four dimensions in cycling, each of these dimensions corresponding to a value of one of the above parameters. An entry in the table is defined by the value of a p-tuple of parameters (with p=2 for the table(s) in calendar, and p=4 for the table(s) in cycling).
[0010] Several publications are known relating to the general subject of defining the parameters of an aging model for electrical energy accumulators, which are identified and briefly discussed in the following.
[0011] Thus, the article by Junfu Lia, Lixin Wang, Chao Lyu, Dafang Wang, and Michael Pecht, entitled “Parameter updating method of a simplified first principles-thermal coupling model for lithium-ion batteries”, Applied Energy V256, December 2019 -113924, presents a method for determining sensitive parameters and updating them. This method is used to obtain a reliable SOC indicator. Indeed, the SOC of systems can present reliability hazards over time and depending on the use of the systems. Two recalibration methods are described: a so-called “offline” method and a so-called “online” method, but are deemed unsatisfactory. A combination of the two methods is also proposed, in the form of an algorithm called “non-standard PF algorithm” by the authors. This algorithm nevertheless appears complex to implement on the type of mapping of a model that can be used in the context of the present invention.This type of method would also be difficult to implement in embedded electronics. with low-capacity microprocessors.
[0012] The article by Biying Ren, Chenxue Xie, Xiangdong Sun, Qi Zhang, and Dan Yan, entitled "Parameter identification of a lithium-ion battery based on the improved recursive least square algorithm", IET May 2020, presents a second-order equivalent electrical circuit calibration method using an improved RLS algorithm. With this type of model as well as with the calibration method presented in this document, however, it is not a recalibration but a first calibration of the model.
[0013] The article by Huiyong Chun, Kwanwoong Yoon, Jungsoo Kim, and Soohee Han, entitled "Improving aging identifiability of lithium-ion batteries using deep reinforcement learning", IEEE TRANSACTIONS ON TRANSFORTATION ELECTRIFICATION 2022, presents a method for determining the parameters of a stoichiometric model via a deep reinforcement learning technique using neural networks. Here again, the method thus presented does not perform a recalibration of aging models of an accumulator.
[0014] Finally and more generally, most of the work reported in the relevant literature deals with the calibration of battery performance models or SOC or SOH indicators. However, as will appear from the description of detailed embodiments of the invention which will be given below, it is appropriate to distinguish the calibration of an endurance model, on the one hand, from its recalibration or dynamic recalibration, on the other hand. For the calibration, a few new accumulators are aged in the laboratory for, for example, 18 months under various conditions, in particular state of charge (SOC), demand current, and / or depth of discharge (DOD) conditions, and the data from the monitoring of this aging is recovered, making it possible to calibrate the aging model for other similar accumulators which will then be used in real operating conditions.Recalibration, on the other hand, can be implemented after a certain period of operation or non-operation of the accumulator (i.e. during the life of the accumulator). Recalibration assumes that we have a previously calibrated aging model, which we update from the monitoring of the accumulator in its final application.
[0015] The inventors are not aware of any technical literature dealing in detail with the recalibration (one can also say dynamic recalibration) of endurance models of the empirical or semi-empirical type during the life of an accumulator, to subsequently enable the production of health forecasts of the accumulator on the basis of updated maps. By "semi-empirical" is meant an endurance model based on the use of simplified physical equations whose parameters are identified thanks to experimental data obtained during specific tests on the battery studied. Statement of the invention
[0016] The present invention proposes a method and means for predicting the state of health of an electrical energy accumulator, in particular but not only a Li-ion accumulator, comprising a recalibration of the values of the map(s) of the endurance model of the accumulator during operation of the accumulator in a real operational situation in the field, when a discrepancy between a prognosis and a diagnosis of the state of health of the accumulator is detected. This means dynamic adaptation, i.e. also the updating of said map values, in response to the detection of such a discrepancy.
[0017] More particularly, a first aspect of the invention relates to a method for predicting the state of health of an electrical energy accumulator, in particular a Lithium-ion battery, comprising the following steps, implemented during operation of the accumulator in a real operational situation in the field: a.simulation of the degradation of the capacity of the accumulator to produce prognostic data of the state of health of the accumulator on the basis of physical data resulting from the monitoring of the operation of the accumulator in the field during a determined monitoring period, using an endurance model accounting for the predicted evolution of the performance of the accumulator as a function of the time elapsed and / or the use of the accumulator since its commissioning in the field, said model being determined by values stored in one or more parametric tables which are indexed by values of p-tuples of operating parameters of the accumulator and which are derived from characterizations of the accumulator previously carried out in the laboratory; . b. production of diagnostic data of the actual health status of the accumulator, based on physical data from monitoring the operation of the accumulator in the field during the monitoring period; and, c. recalibration of at least one of the parametric tables at the end of the monitoring period considered, by replacing only part of the values of said parametric table, the replaced values comprising the values indexed by the p-tuples of operating parameters under which the accumulator operated at least once during the monitoring period considered, when a difference between a value representative of the state of health of the accumulator estimated from the prognostic data and a value representative of the actual state of health of the accumulator based on the values of diagnosis is detected, to produce a recalibrated version of the parametric table.
[0018] This method can be implemented for the purposes of controlling and / or maintaining the electrical energy storage battery, and more generally the installation equipped with this battery, in which the accumulator in question is incorporated. It finds advantageous implementations, more broadly, in all applications requiring the use of an endurance model of one or more accumulators: energy management system (EMS), BMS, predictive maintenance, etc.
[0019] An advantage of the method is that it provides a substantial improvement in the accuracy of battery health predictions.
[0020] Another advantage is that it allows the real environment of the energy storage system to be taken into account in the endurance models of the accumulators, via the real usage and temperature profiles from actual operation data in the field.
[0021] Furthermore, the method allows the real architecture of the electrical energy storage system installed and in actual operation in the field to be taken into account, beyond the laboratory data conventionally used to model the aging of accumulators.
[0022] But the method is particularly original and advantageous in that, despite the above advantages that it provides, it makes it possible to maintain the fineness of the parameterization of the endurance model previously established in the laboratory, with a correction of the map(s) which is targeted on the values of these maps actually used due to the real operational conditions on the ground. The updating of the maps is minimal. This is particularly advantageous in the context of an implementation by electronic means embedded in the BMS, which are necessarily constrained in terms of memory capacity and computing power.
[0023] In one embodiment, the endurance model comprises at least two parametric tables, a first parametric table of which contains values reflecting the rate of degradation in calendar mode of the capacity of the accumulator, and a second parametric table of which contains values reflecting the rate of degradation in cycling mode of the capacity of the accumulator, one and / or the other of said first and second parametric tables being recalibrated in step c).
[0024] In one embodiment, one and / or the other of said first and second parametric tables are recalibrated in step c) by calculation, via the multiplication of values to be replaced by a recalibration coefficient.
[0025] In one embodiment, the recalibration coefficient is identical for all the values replaced in one of the recalibrated maps among the first and second maps and also being identical, where appropriate, for the replacement of values in one and the other of said first and second recalibrated parametric tables.
[0026] In one embodiment, the recalibration coefficient is a scalar number calculated as the ratio between the representative value of the actual health state of the accumulator based on the diagnostic values to the representative value of the health state of the accumulator estimated from the prognosis data.
[0027] In one embodiment, the endurance model further comprises two other parametric tables, a third parametric table of which contains values accounting for the form factor of the degradation in calendar mode of the capacity of the accumulator, and a fourth parametric table of which contains values accounting for the form factor of the degradation in cycling mode of the capacity of the accumulator, one and / or the other of said third and fourth parametric tables being recalibrated, in step c), by an optimization method adapted to minimize the difference between the diagnostic data and the prognostic data.
[0028] In one embodiment, recalibrating a parametric table to produce a recalibrated parametric table in step c) comprises replacing not only the indexed values by the p-tuples of operating parameters of the accumulator in which the accumulator was used at least once during the observation period, but also replacing the values of the parametric table which are close to said indexed values by said p-tuples of operating parameters of the accumulator, using a determined smoothing function.
[0029] In one embodiment, the recalibration of a parametric table to produce a recalibrated parametric table is only carried out, in step c), if the value of the recalibration coefficient differs from unity by a difference value which is greater than a determined value threshold.
[0030] In one embodiment, steps a) to c) are repeated iteratively, with a fixed or variable recurrence, at the end of successive respective monitoring periods.
[0031] In one embodiment, the recurrence of the recalibrations is variable, with an indeterminate variability which is linked to at least one given operating parameter of the accumulator.
[0032] In one embodiment, in which the diagnostic data are based, in addition, on additional physical data which come from monitoring, carried out in situ or ex situ, of other accumulators of the same type as the accumulator considered.
[0033] In embodiments, the method may further comprise, after step c), a step of comparing the diagnostic data with new diagnostic data obtained using the map(s) recalibrated in step c) in order, in the event of a persistent discrepancy between a value representative of the state of health of the accumulator estimated from the prognostic data and a value representative of the actual state of health of the accumulator based on the diagnostic values, to decide to recalibrate other parametric tables and / or to renew the recalibration of the parametric table(s) on the basis of more extensive monitoring data, comprising additional data based on the observation of other accumulators of the same type as the accumulator in question.
[0034] A second aspect of the invention relates to a device for predicting the state of health of an electrical energy accumulator, for example a system for managing a battery of electrical energy accumulators, comprising means for implementing the method according to the first aspect above.
[0035] A third aspect of the invention relates to an electrical energy storage installation comprising an electrical accumulator battery having a plurality of electrical energy accumulators, and a battery management device comprising a device for predicting the state of health of an electrical energy accumulator according to the second aspect above.
[0036] Finally, a fourth and final aspect of the invention relates to a computer program product comprising instructions which, when the program is loaded into the memory and executed by a computer, implements all the steps of the method according to the first aspect as defined above. Brief description of the drawings
[0037] Other aspects, characteristics and advantages of the invention will become apparent from reading the description which follows. This description is purely illustrative and must be read in conjunction with the appended drawings in which the following are shown: [Fig.l] a simplified functional diagram of an electrical energy storage system in which embodiments of the invention can be implemented; [Fig.2A] a graph illustrating the difference between the prognostic data and a diagnostic data at the end of a monitoring period of one year, of an accumulator of an energy storage system such as that of [Fig.l], during operational operation of said system in the field; [Fig.2B] a graph corresponding to the graph of [Fig.2A] after recalibration, on the basis of the diagnostic data, of the mapping of the degradation rate in calendar mode and of the mapping of the degradation rate in cyclic mode of the endurance model of the accumulator considered in the electrical energy storage system, and recalculation of the prognostic data on the basis of the maps thus recalibrated; [Fig.3] a step diagram illustrating the main steps of modes of implementation of the method according to the invention; [Fig.4] a set of three graphs illustrating the classification results of calendar- and cycling-related degradations of the storage capacity of an accumulator, which were caused by the use of an accumulator with determined operating parameter tuples, which operating parameter tuples were observed by monitoring the energy storage system under actual operational conditions in the field; [Fig.5] two sets of three graphs each, which are arranged in a line and which each illustrate, for the degradation rate linked to the calendar on the top line and for the degradation rate linked to cycling on the bottom line, respectively, the mapping before recalibration, the recalibration coefficient, and the recalibrated mapping, respectively to the left, center and right of the corresponding line; and, [Fig.6] a graph illustrating the beneficial effect, during a second year of use of an accumulator in the field, of a recalibration carried out according to embodiments of the invention at the end of a first year of use of the accumulator. Description of the embodiments
[0038] In an electrical energy storage installation, a storage battery makes it possible to store electrical energy in chemical form and then release it in the form of direct current, in a controlled manner. In sectors such as wind or photovoltaic, storage batteries are used to temporarily store the energy produced in excess of the demand for electrical current by the equipment (or equipment) and / or in the current distribution network(s) supplied by the battery, and to intermittently release it during periods of higher current demand. For example, the batteries used in solar farms are batteries optimized for operation with photovoltaic panels producing electrical energy using solar energy.
[0039] With reference to [Fig.l] an electrical energy storage system 1 comprises a battery 2 of accumulators 2b 22, 23, ... 2m, for example Li-ion accumulators, which are as many elementary cells for storing electrical energy. The cells 2 i, 22, 23, ... 2m are connected together in series and / or in parallel.
[0040] The system 1 further comprises an electronic device 3, forming a battery management system (BMS), which is adapted to control the cells (accumulators) 2i , 22, 23, ... 2m of the battery 2. The device 3 comprises at least an electronic calculator 3a, for example a microprocessor or a microcontroller, and an associated electronic memory 3b, comprising for example random access memory and permanent memory. The calculator 3a and the associated memory 3b are of technologies adapted to, and of capacities sized for the use for which they are intended.
[0041] The device 3 may be configured, in general, to implement various functions of the electrical energy storage system incorporating the accumulator in question, and others. These functions may include, but are not limited to, cell balancing functions, protection functions, functions for calculating the state of charge and / or the current state of health of the battery cells, predictive maintenance functions, etc.
[0042] For this purpose, the system 1 also comprises various sensors (not shown in the figure), which are adapted in particular to produce measurements from which the device generates diagnostic data reflecting the actual state of health of each of the accumulators of the battery 2. These physical measurements relate in particular to the operating parameters of the accumulators and of the system 1 in general, such as the operating temperature for example. These measurements are carried out continuously, as part of monitoring the operation of the accumulators in the field. The data thus produced during a determined operating period are stored in the memory 3b of the device 3.
[0043] By way of non-limiting example, those skilled in the art will appreciate that the electronic device 3 may comprise one or more voltage sensors and / or one or more current sensors adapted to measure voltages and / or currents between the terminals of elementary cells and / or groups of elementary cells of the battery connected in series and / or in parallel. The device 3 may further comprise one or more temperature sensors.
[0044] In the context and for the purposes of implementing the invention, it is also possible to store in the memory 3b the diagnostic data of the state of health of the accumulators which have been generated since the start of the operational use of the accumulators, or since a previous recalibration operation, in the case of carrying out a series of recalibration operations with a fixed or variable recurrence.
[0045] Furthermore, the device 3 is suitable for implementing the steps of the method for processing the data measured by the sensors. In the context of this implementation, the computer 3a executes a computer program product, i.e., software comprising instructions suitable for implementing the prediction method which is the subject of the first aspect of the invention. More specifically, when it is loaded into the memory 3b and is executed by the computer 3a of the device 3, this software executes all the steps of the process according to the implementation methods proposed in this description.
[0046] A possible formulation of the empirical model can be written via the following aging equations (1), (2) and (3), based on the work of Grolleau et al. in 2014 (see in particular the article Grolleau Sébastien, Delaille Arnaud, Gualous Hamid, Gyan Philippe, Revel Renaud, Bernard Julien, Redondo-Iglesias Eduardo and Peter Jérémy, “Calendar Aging of Commercial Graphite / LiFePO4 Cell: Predicting Capacity Fade under Time Dependent Storage Conditions”, Journal of Power Sources [Online], 2014, volume 255, pp. 450-458. Published on 11 / 12 / 2013. DOI: 10.1016 / j.jpowsour.2013.11.098): [Math. 1] (i) dt l+Acal.Q [Math. 2] dQ / -, / \ ym = Jcyc (T, SOC, I, DOD )-(1+ Acyc.Q^) [Math. 3]
[0047] In equations (1), (2) and (3) above: • Qioss means the storage capacity of the accumulator lost, expressed in Ampere-hours (Ah), compared to its original nominal capacity (for example when the accumulator was new); • Jcat designates the rate of degradation of the accumulator linked to the calendar, expressed in Ampere-hours per second (Ah.s '); • Jcyc designates the rate of degradation of the accumulator linked to cycling, expressed in Ampere-hours lost “per” Ampere-hour transited (Ah.Ah *); • Qth designates the capacity transferred in the accumulator (charging or discharging), expressed in Ampere-hours (Ah); • t denotes the duration of the period of time considered (operation or non-operation of the accumulator), expressed in seconds (s); • T denotes the temperature of the accumulator, expressed in degrees Celsius (°C); • SOC means the state of charge of the battery, expressed as a percentage (%) of the nominal charge of a new battery; • 1 denotes the demand current of the accumulator expressed in Amperes (A); • DOD means the depth of discharge expressed as a percentage (%) of the nominal charge of a new battery; • Aca[ denotes the form factor of the degradation of the accumulator linked to the calendar, expressed in Amperes / hour (Ah *); • Acyc designates the form factor of the degradation of the accumulator linked to cycling, expressed in Amperes / hour (Ah *); and, finally, • y is a scalar whose value is 1 or -1.
[0048] Equation (1) makes it possible to determine the capacity losses of the accumulator linked to the calendar aging mode. These losses depend mainly on two factors, namely the temperature (7°) and the state of charge (SOC). Equation (2) makes it possible to determine the capacity losses linked to the cycling aging mode. These losses are caused by each stress on the accumulator. They depend mainly on the temperature (T°) and the state of charge (SOC) of the accumulator at the time of the stress, as well as the current stress level ([) and the depth of discharge (DOD) corresponding to the stress. Finally, equation (3) makes it possible to determine the total capacity losses, which are obtained according to this equation by accumulating the losses linked to the calendar mode and the losses linked to the cycling mode.
[0049] The calibration of an empirical endurance model such as that presented above is a step which allows, by successive simulations of the model and comparison of the results of experimental measurements of the SOH, to optimize the values of the parametric tables giving Jcat, Acat, Jcyc and Acyc according each time to the p-uplets of parameters which index these maps. The simulations are carried out by means of test profiles resulting from characterizations of the accumulator carried out in the laboratory. The characterization plan of the accumulator and its results are therefore known at the time when the calibration of the model is carried out. The characterization plan of the accumulator includes several different test conditions carried out in the laboratory on several accumulators having the same reference, that is to say accumulators of the same manufacturing model, as well as the experimental measurement of the variations of SOH which result therefrom.
[0050] In one example, the characterization plan for the aging rate linked to the calendar of a given reference accumulator can be obtained on the basis of a matrix of 9 conditions, corresponding for example to the 9 possible combinations of 3 temperature values (T°) and 3 state of charge (SOC) values. These operating conditions are each applied to one of 9 respective independent cells, having the same manufacturing reference. These cells are thus operated, in a controlled manner, and are monitored during a laboratory aging campaign which lasts for example 18 months or 2 years. To then perform the optimization of the model on the basis of the data from the test profiles thus obtained, simulation software such as MATLAB™ or any other equivalent software can be used.
[0051] Thus, for example, the component of Q[oss corresponding to the capacity losses linked to the calendar mode is obtained by running the simulation software for the model defined above, which is optimized by successive comparisons with the real data of the test profiles obtained, including the experimental estimates of said component gte.„ for the values of Jcal and Acai.
[0052] In the same way, the Qloss component corresponding to the capacity losses linked to the cycling mode is obtained from the real data of the test profiles for the values of Jcyc and Acyc, by making successive comparisons between the data predicted by the model being optimized and the real data of the characterization profiles including the experimental estimation of said Qloss component, and this at the same values of the capacity Qth transited in each of the tested accumulators. It is advantageous to remove the effect of the calendar aging mode progressively accumulated over the test period, by simple subtraction, in order to model the component of the losses linked to the cycling mode with more precision.
[0053] Those skilled in the art will appreciate that the invention is not limited by the example of an empirical endurance model of the accumulator explained in the foregoing, nor by the calibration method described above, nor by the methodology followed to obtain the test profiles which are used to carry out this calibration. In particular, the invention is also compatible with the use of any other endurance model. It is also compatible with the use of a plurality of different endurance models, respectively associated, for example, with accumulators of different types which can be simultaneously used within the same electrical energy storage system, instead of a single model associated with accumulators of the same type (i.e., having the same manufacturing reference). In such a case, the method can be implemented in parallel for the different accumulators on the basis of their respective endurance models.
[0054] In an example of implementation based on the aging model considered here: • the values of the [Jcal] mapping giving the speed Jcal and / or the values of the [Acal] mapping giving the form factor Acah respectively, of the degradation of the accumulator capacity linked to the calendar mode, are indexed by the 2-tuples of parameters {T°,SOC}a, where the index a is between 1 and an integer A which is strictly greater than unity. This number A defines the maximum number of pairs of operating parameters of the accumulator which are taken into account for the implementation of the method. These are the 2-tuples of parameters {T°,SOC} to which the accumulator was subjected at least once in the context of its actual operation in the field during the monitoring period considered; and / or, • the values of the mapping [Jcyc] giving the speed Jcyc and / or the values of the mapping [Acyc] giving the form factor Acyc, respectively, of the degradation of the accumulator capacity linked to cycling, are indexed by the 4-tuples of parameters (T°,SOC,I,DOD}b, where the index b is between 1 and an integer B which is strictly greater than unity. This integer B defines the maximum number of quadruplets of operating parameters of the accumulator taken into account in the process. These are the 4-tuples of parameters {T°,SOC,I,DOD} to which the accumulator has been subjected at least once in the context of its actual operation in the field during the monitoring period considered.
[0055] Here again, it is quite obvious that the invention is not intended to be restricted to these examples, which are purely illustrative of possible but non-limiting modes of implementation of the prediction method according to the invention.
[0056] The principle underlying the method for predicting the state of health of an accumulator according to implementations of the invention lies in an optimized update (or recalibration, or even resetting) of the maps of parametric endurance models (or aging models), when a difference is detected between the state of health prognosis (SOH) obtained using these maps, on the one hand, and the state of health diagnosis of the same accumulator derived from monitoring data of the actual operation (in the field) of the accumulator produced and stored natively in the BMS which is embedded in the storage system, on the other hand.
[0057] In certain embodiments, in particular in installations equipped with batteries comprising a plurality of cells, close or distant, which are connected to each other and / or whose management is shared, it is also possible to use more extensive diagnostic data, further based on additional physical data which come from monitoring other accumulators of the same type (i.e. having the same manufacturing reference) carried out in situ or ex situ. These additional physical data can be recovered by the computer 31 implementing the method, by any appropriate means of communication on which it does not appear useful to expand here: wired data network, for example Ethernet network and / or wireless data network, using for example long-distance communication technologies such as 3G-LTE, 4G or 5G and / or short-distance wireless communication technologies such as NFC, Bluetooth, Wi-Fi, etc..
[0058] According to the principle of the proposed method, the updating of the maps is improved compared to the prior art in that not all the values of the maps are modified when the maps are recalibrated according to the implementations of the invention. On the contrary, only the values of the maps giving J are modified. cai, Jcyc, Aca[ and / or Acyc for determined p-uplets of operating parameters of the accumulator (p-uplets by which these maps are indexed). These are the values which were actually used by being read in the corresponding maps via the p-uplets of operating parameters operated at least once during the considered use of the accumulator, to produce the prognostic data which, in the example considered here, are represented by curve 21 of [Fig.2A].This means that, for each mapping, only the values of the mapping in the operating zones of the accumulator defined by the p-uplets of operating parameters which have actually been observed at least once in the field are modified, i.e. also the values of the mapping indexed by all the p-uplets of parameters in which the accumulator has been used at least once during the period of operating time considered, and only these values.
[0059] The graph in [Fig.2A] gives the value of a state of health (SOH) indicator, namely the residual capacity of the accumulator at time t, expressed as a percentage of the nominal capacity of the accumulator at the beginning of its life. More particularly, the graph in [Fig.2A] illustrates the difference between the prognostic data represented by curve 21 and a diagnostic datum at the end of a monitoring period of one year (i.e. 365 days) represented by point 22, for an accumulator of an energy storage system such as that in [Fig. 1]. Curve 21 therefore represents a decreasing function as a function of the time t elapsed. The diagnostic datum 22 is generated from the diagnostic data accumulated during the continuous monitoring of the accumulator in the actual operational situation, in the field, of the electrical energy storage system incorporating it.
[0060] As can be seen, point 22 is above curve 21 at t = 265 days. This shows that the model defined by equations (1), (2) and (3) and the maps [Jcal], [Acal], [Jcyc] and [Acyc] turns out to be too pessimistic, in the sense that the state of health (SOH) of the accumulator predicted by this model and reflected by curve 21 of [Fig.2A] is below the value actually diagnosed after one year (i.e., at t = 365 days) and reflected by point 22.
[0061] More particularly, we see that the value of the deviation is a little greater than 10% of SOH, since the diagnostic value 22 is equal to 82%, while the 365-day prediction given by curve 21 corresponds to approximately a value between 71% and 72%. This deviation is sufficiently significant to justify a recalibration of the maps. Indeed, the tolerance on the precision of the predictions is only of the order of 2 to 3% error.
[0062] The graph in [Fig.2B] corresponds to the graph in [Fig.2A] but after recalibration, on the basis of the diagnostic data 22, of the mapping of the speed of gradation in calendar mode [Jcal] and the mapping of the degradation rate in cyclic mode [Jcyc] of the endurance model of the accumulator of the electrical energy storage system which is considered here. On this graph, curve 23 represents the predictions of the state of health (SOH) as a function of the elapsed time, that is to say the prognostic data which can be obtained on the basis of the recalibrated maps according to implementations of the method of the invention. It can be seen that, at the end of the observation period considered, i.e. at t = 365 days, curve 23 representing the predicted values (prognosis) now coincides with the value actually observed (diagnosis). This reflects the effectiveness of the method. Indeed, this means that the prognoses provided by the model defined by the recalibrated maps are well aligned with the diagnosis available on the date considered.
[0063] With reference to the step diagram in [Fig.3], we will now describe the main steps of the methods of implementing the method, which make it possible to obtain such a result. It is in fact the implementation of these steps which made it possible to obtain the results presented above with reference to [Fig.2B].
[0064] In step 31 of the method, the computer 3a recovers the diagnostic data obtained by monitoring the actual operational functioning of the accumulator during the period considered (and where appropriate additional diagnostic data, as explained above, obtained for other accumulators of the same type, which are operated in situ or ex situ).
[0065] The computer 3a also recovers the values of the actual operational operating parameters in the field of the accumulator, namely T°, SOC, 1 and DOD, which are part of the composition of the parameter p-uplets indexing the parametric tables (maps) from which the prognostic values of the degradation of the capacity of the accumulator can be calculated. All these data have been stored over time in the memory 3b, during the monitoring period of the accumulator concerned. It is therefore sufficient to read them from the memory 3b. If necessary, the parameters of the additional diagnostic data can be recovered in the same way as said additional data themselves.
[0066] In step 32, parameter p-uplets indexing the [Jcal] and [Jcyc] maps are selected, based on the actual operating conditions of the accumulator during the monitoring period considered. We thus list: • a number A of pairs of parameters {T°,SOC}a, where the index a is between 1 and A, indexing the mapping [Jcal] of the degradation rate in calendar mode; as well as • a number B of quadruplets of parameters {T°,SOC,l,DOD}h, where the index b is between 1 and B, indexing the mapping [Jcyc] of the degradation rate in cyclic mode. Furthermore, the calculator 3a performs, from the prognosis data, the calculation of the calendar contribution and the cycling contribution to the predicted value AQprognosis of the degradation of the capacity of the accumulator over the period considered, using equations (1) and (2), respectively, of the empirical endurance model.
[0067] These calculations are performed from the elements (terms and factors) of these equations and whose values are read in the maps [Jcal] and [Acal] and in the maps [Jcyc] and [Acyc], respectively. For this purpose, access to the [Jcal] and [Acal] maps and to the [Jcyc] and [Acyc] maps is achieved via the 2-tuples {T°,SOC}a for a between 1 and A and via the 4-tuples {T°,SOC,l,DOD}h for b between 1 and B, respectively, which have been previously selected.It is recalled that these numbers A and B correspond to the total number of pairs and the total number of quadruplets, respectively, of values of the operating parameters of the accumulator which were actually observed during the monitoring period considered. These p-uplets of parameters and the corresponding values of the losses Qloss given by equations (1) and (2) from the values indexed by these p-uplets in the maps [Jcal] and [Acal] and in the maps [Jcyc] and [Acyc], are represented by diagrams (A) and (C) as well as by diagram (B), respectively, of [Fig.4], which will be described in detail later.
[0068] Of course, the predicted value AQpronostic of the degradation of the capacity of the accumulator over the period considered is then obtained by summing the two contributions (calendar and cycling), in accordance with equation (3) of the empirical endurance model considered here. Although taking into account the two contributions simultaneously is not an absolute obligation, it is nevertheless preferred because it gives better results, which better reflect the reality of the phenomena of degradation of the capacity of the accumulator over time and according to its use.
[0069] Furthermore, still in step 32, the value AQdiagnostic of the loss of capacity of the accumulator observed at the instant t at which the diagnosis of the state of the accumulator is carried out is evaluated.
[0070] Step 33 comprises obtaining a correction coefficient a, also called a recalibration coefficient because it is used to recalibrate the maps to reduce the gap between the prognostic data and the diagnostic data. Those skilled in the art will appreciate that the recalibration coefficient a is a scalar. Its use as a multiplicative factor for recalibrating the maps (see steps 34 described below) is therefore advantageous, given that it does not require very significant computing power (compared to matrix multiplication, for example).
[0071] In an embodiment as illustrated in [Fig.3], step 33 is performed by calculation. More particularly, the recalibration coefficient a is a scalar ratio obtained by dividing the AQdiagnostic value by the AQpronostic value. It is recalled that the AQdiagnostic value designates the loss of capacity of the accumulator observed at time t at which the diagnosis of the state of the accumulator is carried out. The AQprognostic value, already mentioned above, designates the overall loss of capacity of the accumulator which is predicted at time t considered, on the basis of the cartographic model, i.e. the accumulation (i.e., the sum) of the capacity loss values given by equation (3) on the basis of the values read in the parametric tables (cartographies) for the p-uplcts of parameters under which the accumulator has been used up to time t.
[0072] This implementation of step 33 by calculation gives good results and, advantageously, it is relatively fast. It can nevertheless be replaced, as an optional alternative, by a step of obtaining the coefficient a by optimization, which constitutes one of the possible variants to the method of recalibrating the maps according to the proposed method. The determination by an optimization method is more precise, but slower and requires more computing capacity. The principle of determining the values of the parameters by optimization consists in fact of carrying out successive simulations of the model, and of comparing the results of the prognoses with the diagnostic value AQdiagnostic in order to minimize the difference between the value AQdiagnostic obtained by simulation and this diagnostic value AQdiagnostic obtained in real conditions. For this, an optimization method can be used such as, for example, the least squares method.An alternative method based on the implementation of a neural network can also be used. This variant gives even better results than the computational method, but requires greater processing power and a higher memory capacity at the level of the electronic device 30.
[0073] It should be noted that the recalibration coefficient a is the same for all p-tuples of parameters, both for the calendar mode and for the cycling mode. Indeed, the diagnosis does not give any indication of the distribution of the degradations individually observed for the cycling mode and for the calendar mode, respectively. It is therefore not possible to dissociate their effects through respective correction / recalibration coefficients.
[0074] Step 34 of the method comprises the application of the correction factor a to the values of the starting maps, to be recalibrated, which are each identified in the figure and in what follows by the index “i”, in order to obtain the values of the recalibrated maps, each identified in the figure and in what follows by the index “i+J”. As already mentioned above, this updating of the maps consists of multiplying by the scalar a the relevant values of the starting maps (values identified by the index “i”), in order to obtain the values of the recalibrated map (values identified by the index “i+1”).
[0075] Advantageously, the recalibration coefficient a is identical for all the values that are replaced within each mapping that is recalibrated. This has the advantage of simplicity. And it has been observed that this gives good results in practice.
[0076] In the implementation mode as illustrated in [Fig.3], we simply update: • the A values {Jcal}ia, for a between 1 and A, of the [Jcal] mapping giving the degradation rate linked to the calendar mode and corresponding to the number A of pairs of parameters {T°,SOC}a under which the accumulator operated at least once during the monitoring period considered; and / or • the B values {Jcyc}ith, for b between 1 and B, of the [Jcyc] mapping giving the degradation speed linked to the cycling mode and corresponding to the number B of quadruplets of parameters {T°,SOC, 1,D0D}h under which the accumulator operated at least once during the monitoring period considered.
[0077] Indeed, it has been found that it is these two elements of equation (3) of the empirical endurance model which contribute most to a discrepancy between the prognostic data and the diagnostic data. It is therefore sufficient, in certain applications, to limit oneself to updating the two maps [Jcal] and [Jcyc]. This limits the complexity of the implementation of the method, thereby also reducing the computational cost to which the electronic device 30 is exposed.
[0078] Nothing prevents, however, the provision of implementation modes in which one would recalibrate, in addition, one and / or the other of the two maps [Acal] and [Acyc] which give the values of the form factor of the degradations of the capacity of the accumulator. In these embodiments, one would thus also update, in addition: • the A values {Acal}ia for a between 1 and A, of the [Acal] mapping giving the form factor of the degradation linked to the calendar mode; and / or, • the B values {Acyc}ih for b between 1 and B, of the [Acyc] mapping giving the form factor of the degradation linked to the cycling mode. The recalibration coefficient used, if any, for the value replacements made in each of the [Acal] and [Acyc] maps is noted It differs from the recalibration coefficient a for the [Jcal] and [Jcyc] maps. The recalibration coefficient can however be obtained, if any, in the same way as the recalibration coefficient a, namely by applying the method which has been described in the above with reference to the step diagram of [Fig.3], but from the {Acal} values, and the / Acyc] values, instead of the {Jcal} values, and the {Jcyc} values,, respec- tively.
[0079] The fact of recalibrating only one, several, or even all of the different parametric tables [Jcal], [Acal], [Jcyc] and [Acyc] of the endurance model considered, makes it possible to graduate the consideration of the degradation specifically caused by each of the elements of equations (1), (2) and (3) which determine the empirical model of the degradation of the capacity of the accumulator. This makes it possible to establish, if necessary, an “order of priority” and / or a “level of importance” (or weight) respective to the recalibration of the different parametric tables each entering into the composition of the elements of the endurance model.
[0080] Where appropriate, the [Acal] mapping and the [Acyc] mapping are recalibrated, in step 34, preferably by an optimization method. Such a method is suitable for minimizing the difference between the diagnostic data and the prognostic data. It has indeed been found that an optimization method has better efficiency for updating form factor values of the degradation of the accumulator capacity. This may be, for example, the least squares method, or any other comparable method that a person skilled in the art may consider on the basis of his general knowledge in the technical field in question.
[0081] In embodiments, the respective advantages of a relatively fast computational method (i.e., method for obtaining a recalibration coefficient by calculations, such as for example the method described above for the ratio a) and a relatively more precise optimization method can be combined. Thus, for example, an optimization method making it possible to correct the [Acal] and [Acyc] maps of the form factor of the degradation in calendar mode and in cyclic mode, respectively, can be implemented if the prior correction of the [Jcal] and [Jcyc] maps of the degradation rate in calendar mode and in cyclic mode, respectively, via the recalibration coefficient a obtained by a computational method is not sufficient.In other words, in these embodiments, a correction is first made on the [Jcal] and [Jcyc] maps directly via the recalibration coefficient a, then a correction is made in a second step by optimization on the [Acal] and [Acyc] maps. This makes it possible to initially make a rough but rapid correction on the speeds of degradation of the accumulator capacity in each of the calendar and cyclic modes, then, if necessary, to refine the correction thus obtained by a slower optimization method on the form factors of said degradations.
[0082] Furthermore, in another possible variant, the correction factor can also be applied in step 34 to the values stored in the maps corresponding to unused parameters, but which are close to values corresponding to pa parameters used and which have been updated. This makes it possible in particular to provide a solution to any discontinuities that the mapping could present after recalibration, for values corresponding to one or more p-tuples of isolated parameters. Indeed, such discontinuities can then give rise to simulation results marked by singularities which can produce aberrant effects, for example jumps in the health status forecasts which would not be representative of a real evolution of the health status of the accumulator.
[0083] A person skilled in the art, based on his general knowledge in the field of digital data processing and in particular data in parametric tables, will be able to identify and implement a neighborhood relationship in such a table, linking values to be modified around a determined value of the table. He will also be able to identify and use a function for smoothing the values around an isolated value that is modified in such a table, in order to reduce the discontinuities that this modification can generate. As an example, we can cite the normal distribution law as a function of temperature, in which the temperature of the p-uplet is considered and in which the standard deviation can be set at 5°C for example.
[0084] Notwithstanding the above implementation variant, and regardless of the updated map(s), it will be noted that the recalibration of at least one of the parametric tables at the end of the monitoring period considered is carried out by replacing only part of the values of said parametric table. The replaced values include the values {Jcal,}^ and {Jcyc,}ih indexed by the 2-tuples of operating parameters {T°,SOC}a and by the 4-tuples of operating parameters {T°,SOC,I,DOD}b, respectively, under which the accumulator has operated at least once during the monitoring period considered. The updating of the map(s) is minimal, which is advantageous in terms of memory capacity and computing power required.The partial nature of the update of the parametric table(s) also makes it possible to maintain the precision of the parameterization of the endurance model previously established in the laboratory, with a correction of the map(s) which is targeted on, or even limited to, the values of these maps actually used in the context of real operational conditions in the field.
[0085] It is recalled that the replacements of values in the mapping(s) are carried out to produce a recalibrated version of the corresponding parametric table(s) when a difference between a value representative of the health state AQ prognosis of the accumulator estimated from the prognosis data and a value representative of the actual health state AQdiagnostic of the accumulator based on the diagnostic values is detected. In a possible implementation mode, the recalibration of a mapping to produce a recalibrated mapping is carried out, at step 34, that if in addition the difference which is detected between the value AQpronostic representative of the state of health predicted using the endurance model and the value representative of the real state of health of the accumulator AQdiagnostic, is considered sufficient to justify a recalibration of the maps. Advantageously, this condition can be assessed by comparing the value of the recalibration coefficient a with a determined value threshold ath The recalibration coefficient a being, in embodiments, a scalar whose “neutral” value is equal to unity (i.e. 1), this can be achieved, for example, by calculation in the following manner: • if a distance between a and unity, defined for example as the absolute value of their difference, is strictly greater than ath (i.e. if ll-a I > ath), the recalibration is carried out; whereas, • if the distance between a and the unit, as defined for example as indicated above, is less than or equal to ath (i.e. if ll-al = < ath), the recalibration is not carried out.
[0086] In the case described, the recalibration coefficient a is a scalar number calculated as the ratio between the representative value of the actual health state AQdiagnostic of the accumulator based on the diagnostic values on the representative value of the health state AQprognosis of the accumulator estimated from the prognostic data. The invention is not limited to the use of the distance indicated above. Any other distance can be used, depending on the topology of the set of values of the recalibration coefficient. For example, if the recalibration coefficient a is a vector, the distance of the norm 1 (or distance of the absolute values), the distance of the norm 2 (or Euclidean distance), or the distance of the norm 3 (or uniform distance) can be used, for example, defined each time on the set of values of a.In an example of implementation, moreover, the ath threshold can be set in consideration of the margin of uncertainty (or tolerance) on the precision of the predictions which can be admitted for the application concerned. Indeed, it can be considered that it is not useful to recalibrate the mappings of the elements of the empirical model if the difference observed between the prognosis given by this model and the diagnosis made on the basis of the real data is less than this tolerance. In practice, the tolerance or margin of error admitted being of the order of 2 to 3%, the ath threshold can be set at 3% in relative value (that is to say that ath = 0.03 in the example given in the paragraph above), for example, or even at 4 or 5%.
[0087] This implementation has the advantage of saving on the calculations required to obtain the desired thresholding effect, if necessary.
[0088] Those skilled in the art will appreciate that other modes of implementation may be considered to obtain such a thresholding effect, for example by comparing the difference between the AQprognostic value representative of the state of health which has been prognosticated in using the endurance model, on the one hand, and the AQdiagnostic value representative of the state of health of the accumulator which was actually observed, on the other hand.
[0089] In another possible implementation, steps 31 to 34 are repeated iteratively, with a fixed or variable recurrence, at the end of successive respective accumulator monitoring periods. In other words, the method is implemented as an iterative process, each iteration making it possible to recalibrate, if necessary, all or part of the maps. The recurrence of any recalibrations may be fixed, for example half-yearly, annual, biannual, etc. It may also be variable, for example with a predetermined variability according to a calendar setting recalibration dates, or with an indeterminate variability which may for example be linked to one (or more) given operating parameters of the accumulator, such as the average, maximum or minimum temperature (possibly during determined periods of time), or such as the level of stress on the accumulator during the observation period.In the latter case, the level of stress on the accumulator can be determined on the basis of the capacity transferred into the accumulator (Qth), the number of times a threshold value determined by the stress current (Z) and / or by the depth of discharge (DOD), etc. is exceeded.
[0090] Step 35, which is optional, can be implemented to verify the efficiency of the recalibration performed. This step consists of comparing new diagnostic data (identified in [Fig. 3] by the index i+1 in step 35) which have been recalculated using the maps as they have possibly been recalibrated in step 34, with the diagnostic data used to decide to carry out this recalibration in view of the initial prognostic data (identified in [Fig. 3] by the index i in step 31), on the one hand, and to recalibrate the maps, on the other hand.This verification may reveal an insufficiency or an imperfection of the recalibration in the event of a persistent discrepancy between the prognostic data and the diagnostic data, and lead to a decision to extend the recalibration to other maps (when, for example, not all the [Jcal], [Acal], [Jcyc] and [Acyc] maps have been recalibrated) and / or to renew the calibration on the basis of more extensive monitoring data (in particular with additional data based on the observation of other accumulators of the same type, as mentioned above as a possible option).
[0091] [Fig.4] shows a set of three graphs illustrating the classification results of calendar and cycling-related degradations of the storage capacity of an accumulator, which were caused by the use of an accumulator with determined p-tuples of operating parameters, which p-tuples of operating parameters were observed by monitoring the energy storage system under actual operational conditions in the field.
[0092] More particularly, graph [A] of [Fig.4] gives the elementary values of the degradation in calendar mode that were caused by the use of the accumulator with pairs (2-uplets) of operating parameters {T°,SOC}a, for a included in 1 and A. This graph shows the distribution of elementary degradations in calendar mode, in % of the total degradation predicted by the endurance model, according to the pairs of parameters of the mapping [Jcal] of the degradation rate in calendar mode. Graph (C) in [Fig.4] is a zoom of graph (A) on the first pairs of parameters {T°,SOC}a. In the same way, graph (B) of [Fig.4] shows the distribution of elementary degradations in cycling mode, in % of the total degradation predicted by the endurance model, as a function of the parameter quadruplets (T°, SOC, I, DOD}b, for b included in 1 and B of the mapping [Jcyc] of the degradation rate in cycling mode.
[0093] [Fig.5] shows two sets of three graphs, each arranged in respective lines. Graphs (A) and (C) of the top line illustrate, for the degradation rate linked to the calendar mode: the mapping [Jcal]i before recalibration, and the recalibrated mapping [Jcal],+;, respectively to the left and right of the line. The mapping presented in the center of the top line by graph (B) represents the ratio of the recalibrated mapping [Jcal],+; to the mapping [Jcal], before recalibration. The ratio of two matrices being a matrix, we therefore obtain a three-dimensional mapping noted “[Jcal],+; / [Jcal]” in the title of graph (B), for the ratio values {Jcal}i+i / {Jcal}i. This mapping accounts for the effect of the coefficient a in the entire parameter range of the mapping [Jcal], which makes it possible to highlight the areas affected by the recalibration.Similarly, graphs (D) and (F) in the bottom row illustrate, for the degradation rate related to the cycling mode: the [Jcyc] mapping, before recalibration and the recalibrated [Jcyc],+; mapping, respectively to the left and right of the line. The mapping presented in the center of the bottom row by graph (E) represents the ratio of the recalibrated [Jcyc],+; mapping to the [Jcyc] mapping, before recalibration. This mapping, noted “[Jcyc],+; . / [Jcyc], » in the title of graph (E) reports the effect of the recalibration coefficient a in the entire parameter range of the [Jcyc] mapping. It should be noted that graphs (D), (E) and (F) only illustrate the mapping values for the T° and SOC parameters corresponding to the temperature and the state of charge, respectively, and not the parameter quadruplets {T°, SOC, 1, DOD}. This makes these graphs easier to read. Furthermore, this is justified by the fact that, in practice, the degradation rate linked to the cyclic mode has proven to be little dependent on the regime (Z parameter) and the depth of discharge (DOD parameter).
[0094] Considering graphs (B) and (E), we first notice that they are identical. This illustrates the identity of the recalibration coefficient applied for the update of each modified value in the [Jcal] mapping and in the mapping [Jcyc], which has already been mentioned above.
[0095] We then note that the values of the recalibration coefficient a are all less than unity, being more particularly between 0.9 and 1. This reflects the effect of the recalibration of the maps which, in the example considered here, consists of reducing the value of the elements of the aging equations (1) and (2) of the endurance model used, knowing that the value of the degradation of the capacity of the accumulator evolves in the same direction as these elements. However, as mentioned above with reference to the graph in [Fig.2A], in the example considered here, the model defined by (in particular) the [Jcal] and [Jcyc] maps of the degradation rate turns out to be pessimistic in the sense that the state of health (SOH) of the accumulator predicted by this model and reflected by curve 21 in [Fig.2A] is below the value actually diagnosed after one year (i.e., at t = 365 days) and reflected by point 22 in [Fig.2A].Therefore, it is logical that the recalibration carried out according to the invention has the effect of reducing the values of the elements of equations (1) and (2) of the empirical model, so that the predictions concerning the state of health of the accumulator obtained with the recalibrated maps [Jcal],+; and | Jcyc],+ / are raised compared to those previously obtained with the previous maps [Jcal], and [Jcyc],. .
[0096] Finally and above all, we observe that the [Jcal]i and [Jcyc]i maps have not been entirely modified: in this example, in fact, only the values of the tables in the SOC and temperature parameter zones, which were selected in step 32 of the diagram in [Fig.3], have been corrected.
[0097] The graph in [Fig.6] illustrates the beneficial effect, during a second year of use of an accumulator in the field, of a recalibration carried out according to modes of implementation of the invention at the end of a first year of use of the accumulator.
[0098] On the graph, the following curves are represented: • curve 61 (in thin / weak continuous line) and curve 62 (in broken line) illustrate the evolution of the SOH over a period of 1 year and over a period of 2 years, respectively, based on forecasts over these respective periods, and established in both cases on the basis of the maps without recalibration (i.e., without the implementation of the invention). These curves of course overlap between t = 0 and t = 365 days; • curve 63 (small circle curves) and curve 64 (star curves) illustrate the evolution of the SOH over a period of 1 year and over a period of 2 years, respectively, according to the diagnoses observed during these respective periods. These curves obviously overlap between t = 0 and t = 365 days; • Curve 65 (in thick / strong continuous line) illustrates the evolution of the SOH over a period of 2 years established on the basis of the maps (prognosis) with a recalibration carried out at t = 365 days using the diagnostic data accumulated over the previous period of 1 year (i.e., with the implementation of the invention).
[0099] As the person skilled in the art understands from these curves, observing that curve 65 coincides with curve 63 between t = 0 and t = 365 days and also with curve 64 beyond t = 365 days, the recalibration carried out at 1 year makes it possible to improve the prognostic values of the state of health of the accumulator produced over the entire following period, in this case the period between 1 year and 2 years in the illustrated example, compared to the values which would be predicted without the implementation of the invention and the recalibration carried out, in the example considered here, concerning the [Jcal] and [Jcyc] maps at the end of an observation period of 1 year. This example of course has a general scope of value, being given here only as a simple illustration of one of the advantages of the invention.
[0100] Advantageously, the method can be carried out, according to certain modes of implementation, on the basis of and / or for several identical accumulators or groups of accumulators (identical accumulators being accumulators having the same manufacturing reference, or accumulators which are in all respects comparable and technically substitutable for each other), which operate according to different field use profiles. Indeed, an endurance model and the associated maps being specific to an accumulator reference, this makes it possible to benefit for the implementation of the method from an aggregation of diagnostic data, and thus to obtain better recalibration.Recalibration is better in that it is based on diagnostic data that are more numerous, and therefore statistically more significant on average, and / or in that it covers battery usage ranges that are wider, for example temperature ranges corresponding to the superposition of respective temperature ranges that are potentially different from each other.
[0101] The present invention has been described and illustrated in this detailed description and in the figures of the accompanying drawings, in possible embodiments. The present invention is not limited, however, to the embodiments presented. Other variations and embodiments may be deduced and implemented by those skilled in the art upon reading this description and the accompanying drawings.
[0102] In the claims, the term "comprise" or "comprise" does not exclude other elements or other steps. A single processor or several other units may be used to implement the invention. The various features presented and / or claimed may be advantageously combined. Their presence in the description or in different dependent claims, do not exclude this possibility. Reference signs should not be understood as limiting the scope of the invention.
Claims
Claims
1. Method for predicting the state of health of an electrical energy accumulator, in particular a lithium-ion battery, comprising the following steps, implemented during operation of the accumulator in a real operational situation in the field: a. simulation (32) of the degradation of the capacity of the accumulator to produce prognostic data of the state of health of the accumulator on the basis of physical data resulting from the monitoring (31) of the operation of the accumulator in the field during a determined monitoring period, using an endurance model accounting for the predicted evolution of the performance of the accumulator as a function of the time elapsed and / or the use of the accumulator since its commissioning in the field, said model being determined by values stored in one or more parametric tables ([Jcal], [Acal], [Jcyc], [Acyc]) which are indexed by values of p-uplets of operating parameters of the accumulator and which are derived from characterizations of the accumulator previously carried out in the laboratory; b. production (31) of diagnostic data of the actual state of health of the accumulator, based on the physical data resulting from the monitoring (31) of the operation of the accumulator in the field during the monitoring period; and, c. recalibration (33,34) of at least one of the parametric tables at the end of the monitoring period considered, by replacing only part of the values of said parametric table, the replaced values comprising the values; {Jcyc,}^ indexed by the p-tuples of operating parameters under which the accumulator operated at least once during the monitoring period considered, when a deviation between a representative value of the health state (AQpronostic) of the accumulator estimated from the prognosis data and a representative value of the actual health state (AQdiagnostic) of the accumulator based on the diagnostic values is detected, to produce a recalibrated version of the parametric table.
2. Method according to claim 1, in which the endurance model comprises at least two parametric tables ([Jcal]; [Jcyc]), of which a first parametric table ([Jcal]) contains values reflecting the rate of degradation in calendar mode of the capacity of the accumulator, and of which a second parametric table ([Jcyc]) contains values reflecting the rate of degradation in cycling mode of the capacity of the accumulator, one and / or the other of said first and second parametric tables being recalibrated in step c).
3. Method according to claim 2, wherein one and / or the other of said first and second parametric tables are recalibrated in step c) by calculation, via the multiplication of values to be replaced ({Jcal,}^; {Jcyc,}^ by a recalibration coefficient (a).
4. A method according to claim 3, wherein the recalibration coefficient is identical for all values replaced in one of the recalibrated maps among the first and second maps and also being identical, where appropriate, for the replacement of values in one and the other of said first and second recalibrated parametric tables.
5. A method according to any one of claims 2 to 4, wherein the recalibration coefficient (a) is a scalar number calculated as the ratio between the representative value of the actual health state (AQdiagnostic) of the accumulator based on the diagnostic values and the representative value of the health state (AQprognosis) of the accumulator estimated from the prognosis data.
6. Method according to any one of claims 2 to 5, in which the endurance model further comprises two other parametric tables ([Acal]; [Acyc]), of which a third parametric table ([Acal]) contains values accounting for the form factor of the degradation in calendar mode of the capacity of the accumulator, and of which a fourth parametric table ([Acyc]) contains values accounting for the form factor of the degradation in cycling mode of the capacity of the accumulator, one and / or the other of said third and fourth parametric tables being recalibrated, in step c), by an optimization method adapted to minimize the difference between the diagnostic data and prognostic data.
7. A method according to any one of claims 1 to 6, wherein recalibrating a parametric table to produce a recalibrated parametric table in step c) comprises replacing not only the values ; {Jcyc,}^ indexed by the p-uplets of operating parameters of the accumulator in which the accumulator has been used at least once during the observation period, but also replacing the values of the parametric table which are close to said indexed values by said p-uplets of operating parameters of the accumulator, using a determined smoothing function.
8. A method according to any one of claims 1 to 7, wherein the recalibration of a parametric table to produce a recalibrated parametric table is carried out, in step c), only if the value of the recalibration coefficient (a) differs from unity by a difference value which is greater than a determined value threshold (ath).
9. A method according to any one of claims 1 to 8, wherein steps a) to c) are repeated iteratively, with a fixed or variable recurrence, at the end of successive respective monitoring periods.
10. Method according to claim 9, in which the recurrence of the recalibrations is variable, with an indeterminate variability which is linked to at least one given operating parameter of the accumulator.
11. Method according to any one of claims 1 to 10, in which the diagnostic data are based, in addition, on additional physical data which come from monitoring, carried out in situ or ex situ, of other accumulators of the same type as the accumulator considered.
12. Method according to any one of claims 1 to 11, further comprising, after step c), a step (35) of comparing the diagnostic data with new diagnostic data obtained using the map(s) recalibrated in step c) in order, in the event of a persistent discrepancy between a value representative of the state of health (AQ prognostic) of the accumulator estimated from the prognostic data and a value representative of the actual state of health (AQdiagnostic) of the accumulator based on the diagnostic values, to decide to recalibrate other parametric tables and / or to renew the recalibration of the parametric table(s) on the basis of more extensive monitoring data, comprising additional data based on the observation of other accumulators of the same type as the accumulator considered.
13. Device for predicting the state of health of an electrical energy accumulator, comprising means for implementing all the steps of a method according to any one of claims 1 to 12.
14. An electrical energy storage installation comprising a battery (2) of electrical accumulators having a plurality of electrical energy accumulators (2;-2m) and a battery management device (3) comprising a device for predicting the state of health of an electrical energy accumulator according to claim 13.
15. A computer program product comprising instructions which, when the program is loaded into memory and executed by a computer of a computer, implements all the steps of the method according to any one of claims 1 to 12.
