Method and device for predicting the end of life of a lead-acid battery

Abstract

The invention relates to a method (100) for predicting the end of life of a lead-acid battery. The method comprises, for each segment of a plurality of segments of a "constant voltage" phase, or CV phase, of a battery charging cycle: a collection (110) of several current measurements to form a "floating current" signal for the segment considered, a decomposition into empirical modes (120) of the floating current signal, a calculation (140) of an intrinsic energy for each intrinsic component obtained by the decomposition, a calculation (150) of a total intrinsic energy as a function of the intrinsic energies of the different intrinsic components, an evaluation (160) of an end-of-life detection criterion for the battery as a function of the total intrinsic energy. Figure for the abstract: Fig. 1

Description

Title of the invention: Method and device for predicting the end of life of a lead battery Field of invention

[0001] The present invention belongs to the field of lead battery management. More particularly, a method and a device are proposed for detecting the beginnings of the end of life of a lead battery. State of the art

[0002] Lead-acid batteries are widely used in industry, particularly in the equipment of railway and automotive vehicles and in uninterruptible power supply systems of data centers.

[0003] Data centers are responsible for storing, processing, and transmitting large amounts of data. They are critical infrastructure in today's digital world. Data center servers must remain operational at all times, so it is imperative that they are powered by a reliable electrical system. This is why data centers typically use uninterruptible power supply (UPS) systems.

[0004] Lead-acid batteries are a critical component of uninterruptible power supply systems for a data server. Batteries store energy and, in the event of fluctuations or outages in the power supplied by the electrical grid, they can provide backup power to ensure that the system remains operational.

[0005] Lead-acid batteries are widely used due to their low cost, long life and reliability. They are also well suited for use in relatively high temperature environments.

[0006] Lead-acid batteries, however, require regular maintenance and inspection to ensure their proper functioning. Therefore, it is important to plan for the end of a battery's life in order to replace it and thus avoid the inherent and unpredictable failures associated with this type of battery.

[0007] There are various methods for monitoring the health of a lead-acid battery. For example, it is known to monitor the battery's voltage, internal resistance, capacity, or temperature (low voltage, high internal resistance, loss of capacity, or high temperature may indicate a defective battery). It is also known to monitor the presence of corrosion in the battery, which can lead to premature failure.

[0008] Charging a lead battery is generally done in two successive phases. During a first phase called "CC" (acronym for "Constant Current"), the current flowing through the battery is maintained at a substantially constant value. During this first phase, the voltage across the battery terminals increases as the battery recharges. During a second phase called "CV" (acronym for "Constant Voltage"), the voltage across the battery terminals is maintained at a substantially constant value. During this second phase, the current generally follows a decreasing function of time. The current flowing through the battery during the CV phase is often called "floating current." It prevents the natural discharge of the lead-acid battery.

[0009] In the field of lithium batteries, patent application EP 3324197 A1 describes a method for determining the state of health of a battery cell as a function of a ratio between a charge variation and a current difference measured between two instants of a CV phase (constant voltage recharge phase) of a CC-CV cycle (charge cycle comprising a constant current recharge phase followed by a constant voltage recharge phase).

[0010] The reliability of current methods for monitoring the state of health of a lead battery is not always fully satisfactory. In particular, these methods generally do not allow the end of life of a lead battery to be predicted with sufficient precision and reliability. Statement of the invention

[0011] The present invention aims to remedy all or part of the drawbacks of the prior art, in particular those set out above.

[0012] For this purpose, and according to a first aspect, the present invention proposes a method for predicting the end of life of a lead battery. The method comprises, for each segment of a plurality of segments of a “constant voltage” phase (CV phase) of a battery charging cycle: - collecting a plurality of measurements of current flowing in the battery during said segment, said plurality of measurements forming a “floating current” signal for said segment, - a decomposition into empirical modes of the floating current signal in order to obtain a representation in the form of a sum of a residual signal and one or more intrinsic components, - a calculation of intrinsic energy for each intrinsic component, - a calculation of a total intrinsic energy as a function of the intrinsic energies intrinsic of the different intrinsic components, - an evaluation of an end-of-life detection criterion for the battery based on of the total embodied energy.

[0013] The present invention finds particularly advantageous, although in no way limiting, applications in the monitoring of a lead battery of an uninterruptible power supply system of a data server, or in the monitoring of a lead battery of a motor vehicle. However, nothing would prevent the present invention from being applied in other fields.

[0014] The proposed method clearly differs from conventional methods in that it is based on the analysis of the floating current during the CV phase of a battery charging cycle. Nothing suggests at first glance that this signal contains information relevant to monitoring the state of health of the battery.

[0015] The decomposition into empirical modes is particularly well suited to the analysis of this signal. This decomposition also has the advantage of being relatively fast and not very demanding in terms of computing capacity.

[0016] Monitoring the total intrinsic energy of the battery makes it possible to detect the beginnings of the end of the battery's life. This makes it possible to plan the replacement of the battery in an early and efficient manner in order to avoid a breakdown.

[0017] In particular embodiments, the invention may further comprise one or more of the following characteristics, taken individually or in all technically possible combinations.

[0018] In particular embodiments, the evaluation of the end-of-life detection criterion comprises a comparison of the total intrinsic energy of the segment with a predetermined energy threshold.

[0019] In particular embodiments, the evaluation of the end-of-life detection criterion comprises a verification whether the total intrinsic energy is less than or equal to the energy threshold for a predetermined number of consecutive segments.

[0020] In particular embodiments, the evaluation of the end-of-life detection criterion comprises a calculation of an average total intrinsic energy for the segment and previous segments, and a comparison of a distance between the total intrinsic energy of the segment and the average total intrinsic energy with a predetermined distance threshold.

[0021] In particular embodiments, the average total intrinsic energy is calculated by taking into account the total intrinsic energy of the segment and the total intrinsic energies of all previous segments.

[0022] In particular embodiments, the average total intrinsic energy is calculated by taking into account the total intrinsic energy of the segment and the total intrinsic energies of a subset of the preceding segments.

[0023] These different conditions can be used individually or in combination to detect the end of battery life.

[0024] In particular embodiments, the method further comprises, for each segment of the plurality of segments, an estimation of a statistical reliability of the segment, as a function of the intrinsic components of the floating current signal of the segment. The segment is then filtered if it is deemed unreliable.

[0025] In particular embodiments, the statistical reliability of the segment is estimated as a function of an entropy calculated for a sum of the intrinsic components of the floating current signal of the segment.

[0026] In particular embodiments, for each segment of the plurality of segments, the statistical reliability of a segment is estimated by comparing the total intrinsic energy of the segment with a predetermined reliability threshold, or with the total intrinsic energies calculated for all or part of the preceding segments.

[0027] This filtering step makes it possible to exclude segments presenting aberrant values ​​(segments deemed statistically unreliable).

[0028] In particular embodiments, the battery is part of an uninterruptible power supply system for a data server.

[0029] In particular embodiments, the battery is a battery of a motor vehicle.

[0030] According to a second aspect, the present invention provides a device for predicting the end of life of a lead-acid battery. The device comprises a battery management system configured to provide measurements of current flowing in the battery during a "constant voltage" phase (CV phase) of a battery charging cycle, and a computing unit connected to the battery management system. The computing unit is configured to implement a method according to any of the previously described implementation modes. Presentation of figures

[0031] The invention will be better understood on reading the following description, given by way of non-limiting example, and made with reference to Figures 1 to 4 which represent:

[0032] [Fig-1] a schematic representation of the main stages of an example of implementation implementation of the method according to the invention for predicting the end of life of a lead battery,

[0033] [Fig.2] a graph representing the evolution over time of the distance to the average for the total intrinsic energy for three different batteries,

[0034] [Fig.3] a graph illustrating the possibility of predicting the end of life of a battery by following the evolution over time of the distance from the average of the total intrinsic energy of the battery,

[0035] [Fig.4] a schematic representation of a device according to the invention for predicting the end of life of a lead battery.

[0036] In these figures, identical references from one figure to another designate identical or similar elements. For reasons of clarity, the elements represented are not necessarily on the same scale, unless otherwise stated. Detailed description of the invention

[0037] A lead-acid battery is an electrochemical accumulator whose electrodes are lead-based and the electrolyte is a mixture of water and sulfuric acid. The battery may comprise one or more cells in series assembled in the same casing. The electrodes are generally plates or grids made of a hardened lead alloy (for example using tin, cadmium and strontium, at a rate of a few percent of the alloy).

[0038] [Fig.l] schematically represents the main steps of an example of implementation of a method 100 according to the invention for predicting the end of life of a lead battery. This may be, for example, a battery of a motor vehicle, or a battery of an uninterruptible power supply system of a data server. However, nothing would prevent the present invention from being applied in other fields.

[0039] As illustrated in [Fig.l], the method 100 comprises the following steps for each of a plurality of segments of a "constant voltage" phase, or CV phase, of a battery charge cycle: a collection 110 of several battery floating current measurements during the segment considered, a decomposition into empirical modes 120 (EMD for “Empirical Mode Decomposition” in English) of the floating current signal formed by the measurements obtained, a calculation 140 of an intrinsic energy for each intrinsic component obtained by the EMD decomposition, a calculation 150 of a total intrinsic energy as a function of the intrinsic energies of the different intrinsic components, an evaluation 160 of an end-of-life detection criterion for the battery based on the total intrinsic energy.

[0040] The different segments correspond to a time division of the CV phase considered. The floating current measurements are for example carried out by a battery management system (BMS) connected to the battery. The measurements are for example carried out with an acquisition frequency of between fifteen and sixty seconds in order to obtain between two hundred and eight hundred measurements per segment (in this case the duration of a segment is then between fifty and eight hundred minutes). However, nothing would prevent the measurements from being carried out with a different acquisition frequency, and / or with a different number of points per segment. However, it is advantageous to use a number of measurements between two hundred and eight hundred measurements per segment (using a larger number of measurements does not necessarily imply a significant improvement in the method, and this results in relatively long calculation times; using a smaller number of measurements may, however, limit the performance of the method). All of the measurements collected during the collection step 110 form a floating current signal.

[0041] In step 120, the floating current signal is decomposed according to a decomposition into empirical modes. It should be noted that nothing could suggest that this floating current signal could contain relevant information on the state of health of the cell.

[0042] Empirical mode decomposition consists of decomposing a signal in the form of a sum of functions, in a similar way to what Fourier series decomposition or wavelet decomposition does.

[0043] One of the particularities of the decomposition into empirical modes is that the basis of functions into which the signal is decomposed is not known a priori, but it is constructed adaptively according to the properties of the signal.

[0044] The empirical mode decomposition corresponds to the first part of the Hilbert-Huang transform (HHT). The empirical mode decomposition consists of decomposing a signal in the form of a sum of a residual signal and intrinsic mode functions (IMF for Intrinsic Mode Function). In the present application, these intrinsic mode functions are also called “intrinsic components”.

[0045] As previously indicated, the intrinsic components are not defined analytically. Rather, they are determined adaptively based on the properties of the signal.

[0046] An intrinsic component (IMF) resulting from an empirical mode decomposition (EMD) must satisfy the following requirements: - the number of extrema (i.e. the sum of the number of local maxima and the number of local minima) and the number of zero crossings of the intrinsic component must be equal or differ by a maximum of one; - at any point of the intrinsic component, the average value of the envelope defined by the local maxima and of the envelope defined by the local minima is zero.

[0047] An XO signal decomposed by EMD can then be written in the form: [æ48!

[0049] In this expression, corresponds to the residual signal, N is the number of intrinsic components of the EMD decomposition, and c^t) is the intrinsic component of index i. Each successive intrinsic component contains oscillations of frequency lower than that of the previous one. The residual signal corresponds to a general trend of the signal s(t).

[0050] The decomposition into empirical modes involves a succession of sifting processes. The first sifting process takes the signal s(f) directly as input. The sifting process corresponds to identifying all the local extrema of the input signal, and to connecting the local maxima, respectively the local minima, by an interpolation by cubic splines, in order to obtain an upper envelope, respectively a lower envelope. An average between the upper envelope and the lower envelope can then be calculated and subtracted from the input signal. If the intermediate signal obtained (subtraction of the input signal with the average of the upper and lower envelopes) is not an intrinsic component, the sifting process is repeated on the intermediate signal (which therefore becomes the input signal of a new sifting process) until an intrinsic component is obtained.The sieving processes are repeated until the last intrinsic component is obtained, i.e., for example, until the intermediate signal becomes monotonic or has only one local extremum. The remaining signal then corresponds to the residual signal.

[0051] A stopping criterion may be defined for the sieving process. For example, the stopping criterion is satisfied if the standard deviation between the results of two successive sieving processes is less than or equal to a predetermined stopping threshold. The stopping threshold may typically be between 0.2 and 0.3.

[0052] The paper "The empirical mode decomposition and the Hilbert spectrum for non-linear and non-stationary time series analysis", Norden E. Huang et al., Proc. R. Soc. Lond. A (1998) 454, pp. 903-995, describes empirical mode decomposition in detail, particularly in sections 4 and 5.

[0053] Empirical mode decomposition algorithms are available in programming libraries, for example in MATLAB or Python language.

[0054] In step 140 an intrinsic energy is calculated for each intrinsic component obtained by the EMD decomposition. The energy E, of an intrinsic component ci corresponds for example to the integral of the square of the amplitude of the intrinsic component c> over the duration of the segment considered:

[0055] £, = J |c[t)\2dt

[0056] In step 150, a total intrinsic energy is calculated based on the intrinsic energies of the different intrinsic components of the EMD decomposition. The total intrinsic energy of the segment considered is for example equal to the sum of the energies of the intrinsic components obtained by the EMD decomposition:

[0057] E = ^Ei

[0058] However, nothing would prevent, in a variant, the calculation of the total intrinsic energy by summing the energies of a subset of the intrinsic components obtained by the EMD decomposition (for example by considering only a predefined maximum number of the first intrinsic components obtained by the EMD decomposition).

[0059] In step 160, a battery end-of-life detection criterion is evaluated based on the total intrinsic energy of the segment considered.

[0060] Different conditions can be evaluated, individually or in combination, to detect the end of battery life from the total intrinsic energy value of the segment considered.

[0061] The evaluation 160 of the end-of-life detection criterion may in particular comprise a comparison of the total intrinsic energy of the segment with a predetermined energy threshold. For example, the end of life of the battery (and therefore the need to replace the battery) may be detected when the total intrinsic energy becomes less than or equal to the energy threshold. Optionally, the end-of-life detection criterion may also comprise a check whether the total intrinsic energy is less than or equal to the energy threshold for a predetermined number of consecutive segments. For example, the end of life of the battery is detected if the total intrinsic energy remains less than or equal to the energy threshold for at least five consecutive segments.

[0062] According to yet another example, and particularly advantageously, the evaluation 160 of the end-of-life detection criterion may comprise a calculation of an average total intrinsic energy for the current segment and the previous segments, and a comparison of a distance between the total intrinsic energy of the current segment and the average total intrinsic energy with a predetermined distance threshold.

[0063] The average total intrinsic energy, noted Emoy, for the segment and the preceding segments can be written in the form:

[0064] p _ iy* F _ i yk yNi F - k ~ k

[0065] In this expression, k corresponds to the number of the current segment (which also corresponds to the total number of segments considered by the prediction method); Ej corresponds to the total intrinsic energy of the segment of index j (with j varying between 1 and k); N j is the number of intrinsic components of the decom-

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] EMD position of the segment of index j; E is the energy of the intrinsic component of index i (with i varying between 1 and N The distance A Ek between the total intrinsic energy of the current segment of index £ and the average total intrinsic energy Emi)y can then be written: A Ek — - Emoy — E^ - Emoy It should be noted that it would also be possible, for the calculation of the average, to consider only a subset of the previous segments instead of considering all the previous segments (in other words, it is possible to use a sliding average over a certain number of previous segments, instead of using an average over all the previous segments). The value of the distance threshold is considered to be related to the corrosion processes of the battery. Corrosion processes involve a multitude of events. The number and energies of these events are related to the internal resistance of the battery. When the energy decreases, this means that the cell is reaching the end of its life: the resistance increases. It is then possible to set a distance threshold from which the end of battery life is considered to be detected when the following two conditions are met: - the total intrinsic energy Ek of the current segment is less than the average total intrinsic energy EmOy; and - the distance A Ek between the total intrinsic energy Ek of the current segment and the average total intrinsic energy Emoy becomes greater than the threshold in absolute value. We can also set a negative distance threshold and focus on the relative value of the distance (the distance A Ek is negative when the total intrinsic energy Ek of the current segment is less than the average total intrinsic energy Emoy). In this case, the end of battery life is detected when the distance A Ek between the total intrinsic energy Ek of the segment and the average total intrinsic energy Emoy is less than the threshold. As an example, the graph in Figure 2 shows the evolution over time of the distance from the mean for the total embodied energy for three different batteries in an uninterruptible power supply system for a data server. In this graph, the y-axis represents the distance from the mean for the total embodied energy. The x-axis represents the segment number. Each segment has a duration of 800 minutes. In this example, the distance threshold is set to -0.045. The distance threshold can be determined empirically in the laboratory, and it can be specific to a particular type of battery. In the example illustrated in Figure 2, the batteries are undergoing accelerated aging. can observe that after about 350 segments (i.e. about 195 days), the difference A Ek between the total intrinsic energy Ek of the current segment and the average total intrinsic energy Emoy becomes lower than the distance threshold. In the example considered and illustrated in [Fig.2], the end-of-life detection criterion of a battery is satisfied as soon as this condition remains satisfied for at least three consecutive segments.

[0072] [Fig.3] illustrates a second application example for a car battery (lead battery used for starting the car). Curve 30 of the graph in [Fig.3] illustrates the evolution over time of the distance to the average of the total intrinsic energy of the battery.

[0073] Here again the method according to the invention is particularly well adapted because apart from the very short start-up period, the battery is always kept at full charge by the alternator. The car's on-board computer can be configured to measure A Ek continuously and in real time, and to display an alert message when A Ek satisfies the end-of-life detection criterion of the battery. The alert message indicates that a battery change is to be expected. This makes it possible to avoid inherent and unpredictable breakdowns linked to this type of battery.

[0074] The threshold value used for the evaluation 160 of the end-of-life detection criterion of the battery may be determined in the laboratory by carrying out tests on the battery. The threshold value may in particular be determined to leave a predefined period between the moment of detection and the actual end of life, in order to allow easy replacement of the battery (for example two months of normal use). The determination of the threshold may for example comprise the following steps: - aging of the battery until the end of its actual life (the instant of end of actual life of the battery is represented by tg in [Fig.3], this corresponds to the instant when the battery becomes completely unusable); - a determination of a number of segments equivalent to a duration of two months (in figure 3, time 1a precedes time Ig by two months, the number of segments equivalent to two months corresponds to the number of segments included in the duration (f _ ; and - a determination of the threshold value starting from the actual end of life and shifting by the number of segments equivalent to two months (the threshold value to be used corresponds to the value taken by the curve 30 at time 1a)-

[0075] As illustrated in [Fig.l], the method 100 according to the invention may also include an optional step 130 of estimating the statistical reliability of the segment considered, and filtering the segment if it is deemed unreliable (the segment and the associated measurements are then ignored, i.e. they are not taken into account in assessment 160 of the end-of-life detection criterion).

[0076] This filtering of a segment deemed unreliable makes it possible to avoid taking into account aberrant values ​​in the analysis of the state of health of the battery.

[0077] The statistical reliability of a segment is estimated based on the intrinsic components of the floating current signal obtained for the segment.

[0078] According to a first example, the statistical reliability of the segment is estimated as a function of an entropy calculated for a sum of the intrinsic components of the floating current signal (for example for the sum of all the intrinsic components obtained by the EMD decomposition, or for the sum of a predefined maximum number of the first intrinsic components obtained by the EMD decomposition). Different methods of calculating entropy can be envisaged, such as for example a Shannon entropy calculation, or a Kolmogorov entropy calculation. For example, the segments for which the calculated entropy is too low (less than a predetermined entropy threshold) are filtered. A Shannon entropy threshold of between 0.25 and 0.5 can in particular be envisaged.

[0079] According to a second example, the statistical reliability of a segment is estimated by comparing the total intrinsic energy calculated for the segment with a predetermined energy threshold. For example, segments that have an aberrant total intrinsic energy value (greater than the energy threshold) are filtered.

[0080] According to yet another example, the statistical reliability of a segment is estimated by comparing the total intrinsic energy calculated for the segment with the total intrinsic energies calculated for all or part of the previous segments (for example, the total intrinsic energies can be compared with each other, or the total intrinsic energy of the current segment can be compared with an average value of the total intrinsic energies of previous segments). Different statistical tests can be envisaged for this purpose (Pierce test, Pierson test, etc.).

[0081] [Fig.4] schematically represents a device 10 for predicting the end of life of a lead battery 21. The device 10 comprises in particular a memory 11, a battery management system 13 and a computing unit 12 connected to the memory 11 and to the battery management system 13.

[0082] The battery management system 13 is configured to provide current measurements made at the battery 21 during at least one segment of a CV phase of the battery.

[0083] The calculation unit 12 is configured to implement the method 100 according to any one of the implementation modes described above.

Claims

Claims

1. Method (100) for predicting the end of life of a lead-acid battery (21), the method (100) comprising, for each segment of a plurality of segments of a "constant voltage" phase, or CV phase, of a battery charging cycle: - a collection (110) of a plurality of measurements of current flowing in the battery (21) during said segment, said plurality of measurements forming a "floating current" signal for said segment, - a decomposition into empirical modes (120) of the floating current signal in order to obtain a representation thereof in the form of a sum of a residual signal and one or more intrinsic components, - a calculation (140) of an intrinsic energy for each intrinsic component, - a calculation (150) of a total intrinsic energy as a function of the intrinsic energies of the different intrinsic components,- an evaluation (160) of an end-of-life detection criterion for the battery (21) as a function of the total intrinsic energy.,

2. The method (100) of claim 1 wherein the evaluation (160) of the end-of-life detection criterion comprises a comparison of the total intrinsic energy of the segment with a predetermined energy threshold.

3. The method (100) of claim 2 wherein evaluating (160) the end-of-life detection criterion comprises checking whether the total intrinsic energy is less than or equal to the energy threshold for a predetermined number of consecutive segments.

4. Method (100) according to any one of claims 1 to 3 in which the evaluation (160) of the end-of-life detection criterion comprises: - a calculation of an average total intrinsic energy for the segment and previous segments, - a comparison of a distance between the total intrinsic energy of the segment and the average total intrinsic energy with a predetermined distance threshold.

5. The method (100) of claim 4 wherein the average total intrinsic energy is calculated by taking into account the total intrinsic energy of the segment and the total intrinsic energies of all previous segments.

6. The method (100) of claim 4 wherein the average total intrinsic energy is calculated by taking into account the total intrinsic energy of the segment and the total intrinsic energies of a subset of the preceding segments.

7. Method (100) according to any one of claims 1 to 6, further comprising, for each segment of the plurality of segments, an estimation (130) of a statistical reliability of the segment, as a function of the intrinsic components of the floating current signal of the segment, and a filtering of the segment if it is judged to be unreliable.

8. The method (100) of claim 7, wherein the statistical reliability of the segment is estimated based on an entropy calculated for a sum of the intrinsic components of the floating current signal of the segment.

9. A method (100) according to any one of claims 7 to 8 wherein, for each segment of the plurality of segments, the statistical reliability of a segment is estimated by comparing the total intrinsic energy of the segment with a predetermined reliability threshold, or with the total intrinsic energies calculated for all or part of the preceding segments.

10. A method (100) according to any one of claims 1 to 9, wherein the battery (21) is part of an uninterruptible power supply system for a data server.

11. A method (100) according to any one of claims 1 to 9, wherein the battery (21) is a battery of a motor vehicle.

12. Device (10) for predicting the end of life of a lead-acid battery (21), said device (10) comprising: - a battery management system (13) configured to provide measurements of current flowing in the battery (21) during a “constant voltage” phase, or CV phase, of a charging cycle of the battery (21), - a calculation unit (12) connected to the management system (13) battery, said computing unit (12) being configured to implement a method according to any one of claims 1 to 11.

Images