Electric vehicle battery degradation estimation based on federated learning of a model predicting battery charge level decrease per vehicle trip
The federated learning method addresses privacy and congestion issues in electric vehicle battery degradation estimation by using local models trained on individual vehicles, ensuring accurate and efficient battery health assessment.
Patent Information
- Application Number
- FR2023005010
- Authority / Receiving Office
- FR · FR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-22
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2043-05-22
AI Technical Summary
Existing methods for estimating electric vehicle battery degradation face challenges due to privacy concerns over data transmission and network congestion, particularly when centralizing data for neural network training.
A federated learning approach where a central device transmits a global model for predicting battery charge level decrease based on local models trained by individual vehicles, preserving data privacy and reducing network bandwidth usage.
This method maintains data privacy, reduces network congestion, and efficiently updates local and global models for accurate battery degradation estimation, enabling proactive maintenance.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
Title of the invention: Estimation of electric vehicle battery degradation based on federated learning of a model for predicting the decrease in battery charge level per vehicle journey Technical field
[0001] The present invention relates to the estimation of battery degradation of an electric vehicle. In particular, the present invention relates to a method and system for estimating battery degradation of an electric vehicle based on learning a model for predicting the decrease in the charge level of an electric vehicle per trip. Technological background
[0002] Electric vehicles comprising a battery are equipped with a management system for this battery which controls the state of charge of this battery and the state of health of this battery using signals coming from the battery. The state of health of the battery resulting from these management systems is a first means used in the state of the art for estimating the degradation of a battery of an electric vehicle.
[0003] Another means is to transmit data relating to electric vehicles, data relating to their battery and data relating to the driving of these electric vehicles to a central device, such as a server, so that this central device trains a neural network dedicated to estimating battery degradation of these electric vehicles.
[0004] One of the problems of centralizing data collected by electric vehicles for the purpose of training a neural network in a centralized manner is transmitting this data to the central device because this data is private and the users of these electric vehicles do not want it to be transmitted over a communication network.
[0005] Communication network congestion problems can also appear when the number of electric vehicles is significant.
[0006] Summary of the present invention
[0007] An object of the present invention is to solve at least one of the problems of the technological background described above.
[0008] Another object of the present invention is to provide an estimation of degradation of an electric vehicle battery.
[0009] Another object of the present invention is to prevent battery failures of electric vehicle.
[0010] According to a first aspect, the present invention relates to a method for degrading an electric vehicle battery of a first set of electric vehicles in communication with a central device, said central device being further in communication with electric vehicles of a second set of electric vehicles, said method comprising the following steps: - transmission, by the central device and to each electric vehicle of the first and second set of electric vehicles, of a global model for predicting the decrease in battery charge level per trip obtained from local models for predicting the decrease in battery charge level per trip trained by electric vehicle devices of the second set of electric vehicles, each local model for predicting the decrease in battery charge level per trip being obtained from data collected by an electric vehicle of the second set of electric vehicles following trips made by said electric vehicle of the second set of electric vehicles and from actual decreases in the battery charge level of said electric vehicle of the second set of electric vehicles after said trips; - following a journey made by an electric vehicle from the first set of electric vehicles, obtaining, by the device of said electric vehicle from the first set of electric vehicles, an estimate of a decrease in battery charge level per journey from the global model for predicting the decrease in battery charge level per journey; - calculation, by the device of said electric vehicle of the first set of electric vehicles, of an actual decrease in battery charge level following said journey made by said electric vehicle of the first set of electric vehicles; - obtaining, by the device of said electric vehicle of the first set of electric vehicles, an estimate of battery degradation of said electric vehicle of the first set of electric vehicles by calculating a difference between the estimate of the decrease in battery charge level per journey and the actual decrease in battery charge level.
[0011] The method makes it possible to preserve the anonymity of the data collected by an electric vehicle because it avoids the transmission of data relating to an electric vehicle and data relating to a journey made by the electric vehicle via a communication network.
[0012] The method makes it possible to preserve the bandwidth of the communication network used because only parameters of models predicting the decrease in battery charge level per journey are transmitted between the electric vehicles and a central device.
[0013] The method is advantageous because it uses the resources of each electric vehicle for training local battery charge level decrease prediction models. The central device only determines a global battery charge level decrease prediction model per journey from said local battery charge level decrease prediction models.
[0014] The method allows an update of the local battery charge level decrease prediction models by receiving parameters of the global battery charge level decrease prediction model received but also following a journey made by an electric vehicle which can then launch a learning phase locally in order to determine a new local battery charge level decrease prediction model.
[0015] The method is advantageous because it makes it possible to determine an update of the global battery charge level decrease prediction model as a function of the local battery charge level decrease prediction model(s) driven by a new electric vehicle from the second set of electric vehicles.
[0016] The method is advantageous because it makes it possible to control the battery degradation of an electric vehicle as soon as this vehicle is added to the first set of electric vehicles.
[0017] According to a variant, the transmission, by the central device and to each electric vehicle of the first and second set of electric vehicles, of the global model for predicting the decrease in battery charge level per journey comprises the following steps: a) transmission, by the central device, of parameters of an initial local model for predicting the decrease in battery charge level per journey to the device of each electric vehicle of the first and second sets of electric vehicles; b) training at least one local model for predicting a decrease in battery charge level per trip by the device of at least one electric vehicle of the second set of electric vehicles, each local model for predicting a decrease in battery charge level per trip is trained by an electric vehicle of the second set of electric vehicles from data collected by said electric vehicle of the second set of electric vehicles and from actual decreases in battery charge level of said electric vehicle of the second set of electric vehicles following a predetermined number of trips made by said electric vehicle of the second set of electric vehicles; c) transmitting the parameters of said at least one local di- prediction model reduction in battery charge level per trained trip to the central device, the parameters of each local model for predicting reduction in battery charge level per trained trip are output by the device of the electric vehicle of the second set of electric vehicles which trained said local model for predicting reduction in battery charge level per trip; d) determination, by the central device, of the parameters of the global model for predicting the decrease in battery charge level per trip from the parameters of said at least one local model for predicting the decrease in battery charge level per trip received; e) transmission, by the central device and to the device of each electric vehicle of the first and second sets of electric vehicles, of the parameters of the global model for predicting the decrease in battery charge level per journey obtained and steps b), c), d), and e) are iterated.
[0018] According to a variant, steps b), c), d), and e) are iterated until a stopping condition is verified.
[0019] According to a variant, the parameters of the global model for predicting the decrease in battery charge level per journey are determined from the parameters of said at least one local model for predicting the decrease in battery charge level per journey according to a method of averaging the parameters of said at least one local model for predicting the decrease in battery charge level per journey.
[0020] According to a variant, the device of an electric vehicle of the second set trains a local model for predicting the decrease in battery charge level per journey per initial battery charge level of said electric vehicle of the second set of electric vehicles and the central device obtains as many global models for predicting the decrease in battery charge level per journey as there are initial battery charge levels.
[0021] According to a variant, the parameters of a global model for predicting the decrease in battery charge level per journey for a given initial charge level are obtained from the parameters of at least one local model for predicting the decrease in battery charge level per journey corresponding to said initial charge level according to a method of averaging the parameters of said at least one local model for predicting the decrease in battery charge level per journey corresponding to said initial charge level.
[0022] According to a second aspect, the present invention relates to a device comprising a memory associated with a processor configured for the implementation of at least one step of the method according to the first aspect of the present invention.
[0023] According to a third aspect, the present invention relates to an electric vehicle, for example of the automobile type, comprising a device as described above according to the second aspect of the present invention.
[0024] According to a fourth aspect, the present invention relates to a computer program which comprises instructions adapted for executing the steps of the method according to the first aspect of the present invention, in particular when the computer program is executed by at least one processor.
[0025] Such a computer program may use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other desirable form.
[0026] According to a fifth aspect, the present invention relates to a computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the method according to the first aspect of the present invention.
[0027] On the one hand, the recording medium may be any entity or device capable of storing the program. For example, the medium may comprise a storage means, such as a ROM memory, a CD-ROM or a microelectronic circuit type ROM memory, or a magnetic recording means or a hard disk.
[0028] Furthermore, this recording medium may also be a transmissible medium such as an electrical or optical signal, such a signal being able to be conveyed via an electrical or optical cable, by conventional or hertzian radio or by self-directed laser beam or by other means. The computer program according to the present invention may in particular be downloaded from an Internet-type network.
[0029] Alternatively, the recording medium may be an integrated circuit in which the computer program is incorporated, the integrated circuit being adapted to perform or to be used in performing the method in question.
[0030] According to a sixth aspect, the present invention relates to a system for estimating battery degradation of an electric vehicle of a first set of electric vehicles comprising a central device, said central device being in communication with electric vehicles of the first set of electric vehicles and electric vehicles of a second set of electric vehicles, characterized in that said central device implements at least one step of the method according to the first aspect of the present invention and each electric vehicle of the first and second sets of electric vehicles is an electric vehicle according to the third aspect of the present invention. Brief description of the figures
[0031] Other features and advantages of the present invention will emerge from the Description of the particular and non-limiting examples of embodiment of the present invention below, with reference to the appended figures 1 to 4, in which:
[0032] [Fig.l] schematically illustrates a system (1) for estimating the degradation of an electric vehicle battery of a first set of electric vehicles, according to a particular and non-limiting exemplary embodiment of the present invention;
[0033] [Fig.2] schematically illustrates a device embedded in electric vehicles of [Fig.l], according to a particular and non-limiting exemplary embodiment of the present invention;
[0034] [Fig.3] schematically illustrates a central device, according to a particular and non-limiting exemplary embodiment of the present invention;
[0035] [Fig.4] illustrates a flowchart of the different steps of a method for estimating the degradation of an electric vehicle battery of a first set of electric vehicles, according to a particular and non-limiting exemplary embodiment of the present invention.
[0036] Description of the examples of embodiment
[0037] A method and a device for estimating the degradation of an electric vehicle battery of a first set of electric vehicles in communication with a central device will now be described in the following with joint reference to FIGS. 1 to 4. The same elements are identified with the same reference signs throughout the description which follows.
[0038] [Fig.l] schematically illustrates a system 1 for estimating the degradation of an electric vehicle battery of a first set of electric vehicles of a first set 10 of electric vehicles 101, 102, ..., 10i, according to a particular and non-limiting exemplary embodiment of the present invention.
[0039] The system 1 comprises a first set 10 of electric vehicles and a second set 11 of electric vehicles 1 Ij.
[0040] The first set 10 of electric vehicles 10i may be a subset of a set of electric vehicles of the same make, of the same model and using the same batteries.
[0041] For example, the first set 10 of electric vehicles 10i is composed of the oldest electric vehicles (more than 6 months old) of the set of electric vehicles of the same make, of the same model and using the same batteries.
[0042] The other electric vehicles in the set of electric vehicles of the same make, same model and using the same batteries, i.e. the most recent electric vehicles (less than 6 months old) form the second set 11 of electric vehicles 1 Ij.
[0043] The first 10 and second 11 sets of electric vehicles are dynamic that is, new electric vehicles can be added to the second set 11 and electric vehicles 1 Ij older than 6 months disappear from the second set 11 and are added to the first set 10 of electric vehicles 10i.
[0044] It is subsequently assumed that most of the batteries of the electric vehicles of the second set 11 are functioning correctly and that if one of them is defective it is then considered as an aberrant data.
[0045] Each electric vehicle 10i and 10j comprises a device 2 implementing part of the steps of the method 2 for estimating the degradation of the electric vehicle battery of the first set 10.
[0046] The system 1 also comprises a central device 14.
[0047] The central device 14 is remote from the device 2 on board each vehicle. electrical 10i and 1 Ij. The central device 14 communicates with a device 2 by communication means such as an interface element of a block 303 or a communication interface 304 of the central device 14 of [Fig. 3] and an interface element of a block 203 or a communication interface 204 of the device 2 of [Fig. 2],
[0048] [Fig.l] illustrates the example of devices 2 that communicate with the central device 14 via a relay antenna 12 of a communication network infrastructure 13. The different elements of the communication network infrastructure used to route a message between the antenna 12 and the central device 14 are illustrated here by a cloud. The communications between the central device 14 and a device 2 are bidirectional, that is to say that the devices 14 and 2 as well as the elements of the communication network infrastructure 13 are adapted so that the central device 14 can transmit information to a device 2 (downstream communication) and receive information transmitted by a device 2 (upstream communication).
[0049] [Fig.2] schematically illustrates a device 2 embedded in vehicles 10i and 1 Ij of [Fig.l], according to a particular and non-limiting exemplary embodiment of the present invention. Examples of such a device 2 include, but are not limited to, various electronic devices such as a smartphone, a tablet, a laptop, electronic equipment embedded in one of the electric vehicles 10i or 10j. The elements of the device 2, individually or in combination, may be integrated into a single integrated circuit, into several integrated circuits, and / or into discrete components. The device 2 may be produced in the form of electronic circuits or software (or computer) modules or even a combination of electronic circuits and software modules. The device 2 comprises one (or more) processor(s) 201 configured for the execution of the instructions of the software(s) embedded in the device 2. The processor 201 may include integrated memory, an input / output interface, and various circuits known to those skilled in the art. The device 2 further comprises at least one memory 202, by a volatile and / or non-volatile memory and / or comprises a memory storage device which may comprise volatile and / or non-volatile memory, such as EEPROM, ROM, PROM, RAM, DRAM, SRAM, flash, magnetic or optical disk.
[0050] The computer code comprising the instructions to be loaded and executed by the processor is for example stored in the memory or the memory storage device 202.
[0051] According to a particular and non-limiting embodiment, the device 2 comprises a block 203 of interface elements for communicating with external devices, for example the remote device 14 or a relay antenna 12. The interface elements of the block 203 comprise one or more of the following interfaces:
[0052] RF radio frequency interface, for example of the Bluetooth® or Wi-Fi® type;
[0053] USB interface (from the English “Universal Serial Bus” or “Universal Serial Bus”) in French);
[0054] HDMI interface (from the English “High Definition Multimedia Interface” or “High Definition Multimedia Interface” in French); - RFID interface (from the English Radio Frequency Identification).
[0055] According to another particular embodiment, the device 2 comprises a communication interface 204 which makes it possible to establish communication with other devices. The communication interface 204 corresponds for example to a transmitter configured to transmit and receive information and / or data via a communication channel 205. The communication interface 204 comprises for example a modem and / or a network card and the communication channel can for example be implemented in a wired and / or wireless medium.
[0056] Data is for example exchanged with the central device 14 using a Wi-Fi® network such as according to IEEE 802.11 via an access point. Communication between a device 2 on board an electric vehicle and this access point can then be established according to a wireless communication protocol such as Wifi® (according to IEEE 802.11), for example in the 2.4 or 5 GHz frequency bands, or Bluetooth (according to IEEE 802.15.1), in the 2.4 GHz frequency band. The wireless access point can also communicate with another access point via a wired link (or "backbone"), for example MoCA (Multimedia over Coax Alliance), Ethernet, PLC (Powerline Communication), POF (Plastic Optical Fiber) plastic) or ITU G.hn (corresponding to the standard for next-generation home network technologies from ITU, from the English "International Telecommunication Union" or in French "Union internationale des télécommunications").
[0057] Data can also be exchanged with the central device 14 using a cellular network such as a 4G network (or LTE Advanced according to 3GPP release 10 -version 10) or 5G via the relay antennas 107 and / or 110.
[0058] Data can also be exchanged with the central device 14 by reading a “radio tag” (RFID tag) then affixed to an element of the electric vehicles 10i or 10j or to equipment on board these electric vehicles 10i or 10j.
[0059] The communication interface 204 can also behave as an interface of an ad hoc wireless network (also called WANET (from the English “Wireless Ad Hoc Network”) or MANET (from the English “Mobile Ad Hoc Network”)) corresponding to a decentralized wireless network. Unlike a centralized network which relies on an existing infrastructure comprising for example routers or access points connected to each other by a wired or wireless infrastructure, the ad hoc wireless network is made up of nodes which each participate in the routing of data by retransmitting the data from one node to another, from the transmitter to the recipient, according to the connectivity of the network and a routing algorithm. The ad hoc wireless network advantageously corresponds to a vehicular ad hoc network (or VANET, from the English "Vehicular Ad hoc NETwork") or to an intelligent vehicular ad hoc network (or InVANET, from the English "Intelligent Vehicular Ad hoc NETwork").In such a network, two or more electric vehicles each carrying a node, i.e. a communication interface 204 behaving like a WANET or MANET interface, can communicate with each other in the context of an electric vehicle-to-electric vehicle (V2V) communication (from the English "vehicle-to-vehicle"); or an electric vehicle-to-infrastructure (V2I) communication (from the English "vehicle-to-infrastructure"). Each electric vehicle can also communicate with one or more pedestrians equipped with mobile devices (for example a smartphone)) in the context of an electric vehicle-to-pedestrian (V2P) communication (from the English "vehicle-to-pedestrian").
[0060] According to a variant, a node of the ad hoc wireless network is connected to the communication network infrastructure 13 via a wired and / or wireless connection. The node can thus act as a relay between this communication network infrastructure and an electric vehicle 10i or 1 Ij. A relay antenna 12 can also implement an interface of an ad hoc wireless network. Such an antenna therefore ensures the link between the ad hoc network and the communication network infrastructure 13 of which it is part.
[0061] The devices 2 embedded in the electric vehicles 10i and 11j can thus communicate with the central device 14 via at least one node of a wireless ad hoc network using communication technologies such as ITS G5 (from the English “Intelligent Transportation System G5” or in French “Système de transport intelligent G5”) in Europe or DSRC (from the English “Dedicated Short Range Communications” or in French “Communications réservés à courte gamme”) in the United States of America which are both based on the IEEE 802.1 Ip standard or even the technology based on cellular networks called C-V2X (from the English “Cellular - Vehicle to Everything” or in French “Cellulaire - Véhicule électrique vers tout”) which is based on 4G based on LTE (from the English “Long Term Evolution” or in French “Evolution à long terme”) and soon 5G.
[0062] According to an additional particular embodiment, the device 2 can provide output signals to one or more external devices, such as a display screen 206, one or more speakers 207 and / or other peripherals 208 via output interfaces 209, 210 and 211 respectively. According to a variant, one or other of the external devices is integrated into the device 2. The display screen 206 corresponds for example to a screen, touch-sensitive or not.
[0063] [Fig.3] schematically illustrates a central device 14, according to a particular and non-limiting exemplary embodiment of the present invention. Examples of such a central device 14 include, but are not limited to, various electronic devices such as one or more computers or servers, one or more calculators. The elements of the central device 14, individually or in combination, may be integrated into a single integrated circuit, into several integrated circuits, and / or into discrete components. The central device 14 may be implemented in the form of electronic circuits or software (or computer) modules or even a combination of electronic circuits and software modules. The central device 14 comprises one (or more) processor(s) 301 configured to execute the instructions of the software(s) embedded in the central device 14.The processor 301 may include integrated memory, an input / output interface, and various circuits known to those skilled in the art. The central device 14 further comprises at least one memory 302, by a volatile and / or non-volatile memory and / or comprises a memory storage device which may comprise volatile and / or non-volatile memory, such as EEPROM, ROM, PROM, RAM, DRAM, SRAM, flash, magnetic or optical disk.
[0064] The computer code comprising the instructions to be loaded and executed by the processor is for example stored in the memory or the memory storage device 302.
[0065] According to a particular and non-limiting embodiment, the central device 14 comprises a block 303 of interface elements for communicating with external devices, for example a remote device 2 or a relay antenna 12. The interface elements of the block 303 comprise one or more of the following interfaces:
[0066] RF radio frequency interface, for example of the Bluetooth® or Wi-Fi® type;
[0067] USB interface (from the English “Universal Serial Bus” or “Universal Serial Bus”) in French);
[0068] HDMI interface (from the English “High Definition Multimedia Interface” or “High Definition Multimedia Interface” in French); - RFID interface (from the English Radio Frequency Identification).
[0069] According to another particular embodiment, the central device 14 comprises a communication interface 304 which makes it possible to establish communication with other devices. The communication interface 304 corresponds for example to a transmitter configured to transmit and receive information and / or data via a communication channel 305. The communication interface 304 comprises for example a modem and / or a network card and the communication channel can for example be implemented in a wired and / or wireless medium.
[0070] Data is for example exchanged with a device 2 embedded in an electric vehicle 10i or 11j using a Wi-Fi® network such as according to IEEE 802.11 via an access point. Communication between the central device 14 and an access point can then be established according to a wireless communication protocol such as Wifi® (according to IEEE 802.11), for example in the 2.4 or 5 GHz frequency bands, or Bluetooth (according to IEEE 802.15.1), in the 2.4 GHz frequency band. A wireless access point can also communicate with another access point via a wired link (or "backbone"), for example MoCA (Multimedia over Coax Alliance), Ethernet, PLC (Powerline Communication), POF (Plastic Optical Fiber) or ITU G.hn (corresponding to the standard for next-generation home network technologies from ITU, from the English "International Telecommunication Union" or in French "Union internationale des télécommunications").
[0071] Data is for example exchanged with a device 2 embedded in an electric vehicle 10i or 11j using a cellular network such as a 4G network (or LTE Advanced according to 3GPP release 10 - version 10) or 5G via the relay antennas 107 and / or 110.
[0072] Data can also be exchanged with a device 2 embedded in an electric vehicle 10i or 11j by reading a “radio-tag” (from the English RFID tag) then affixed to an element of the electric vehicles 10i or 11j or to a equipment on board these electric vehicles 10i or 1 Ij.
[0073] According to an additional particular embodiment, the central device 14 can provide output signals to one or more external devices, such as a display screen 306, one or more speakers 307 and / or other peripherals 308 via output interfaces 309, 310 and 311 respectively. According to a variant, one or other of the external devices is integrated into the central device 14. The display screen 306 corresponds for example to a screen, touch-sensitive or not.
[0074] [Fig.4] illustrates a flowchart of the different steps of a method for estimating the degradation of a battery of an electric vehicle of a first set of electric vehicles, according to a particular and non-limiting exemplary embodiment of the present invention.
[0075] The method is for example implemented partially by the central device 14 and by the device 2 embedded in each electric vehicle 1 Ij and in each electric vehicle 10i. The electric vehicles 10i and 11 j are in communication with the central device 14 as discussed previously.
[0076] In a first step 31, the central device 14 obtains a global model M for predicting the decrease in battery charge level per journey from local models Mj for predicting the decrease in battery charge level per journey trained by the devices 2 of electric vehicles 1 Ij from data collected by said electric vehicles 1 Ij following journeys made by said electric vehicles 1 Ij and from actual decreases in battery charge level of said electric vehicles 1 Ij after said journeys.
[0077] In a second step 32, the central device 14 transmits the parameters of the global model M for predicting the decrease in battery charge level per journey to a device 2 of each electric vehicle 10i of the first set of electric vehicles.
[0078] In a third step 33, following a journey made by an electric vehicle 10i, the device 2 of the electric vehicle 10i obtains an estimate of a decrease in battery charge level per journey from the global model M for predicting a decrease in battery charge level per journey.
[0079] In a fourth step 34, the device 2 of the electric vehicle 10i calculates an actual decrease in battery charge level following said journey made by the electric vehicle 10i.
[0080] According to a variant, said actual decrease in the charge level of a battery of an electric vehicle 10i is calculated by calculating the difference between an initial charge level of the battery of the electric vehicle 10i before the start of said journey and a charge level of this battery at the end of said journey.
[0081] In a fifth step 35, the device 2 of the electric vehicle 10i obtains a estimated battery degradation of the electric vehicle lOi by calculating a difference between the estimated decrease in battery charge level per trip and the actual decrease in battery charge level.
[0082] According to a variant, if the calculated degradation of the battery is greater than a threshold value, in a sixth step 35, an actuator of the electric vehicle 10i is activated to indicate that a preventive maintenance phase of the battery of the electric vehicle 10i is to be planned.
[0083] According to a variant, an indicator light or an area of an on-board human-machine interface of the electric vehicle 10i is actuated.
[0084] The electric vehicle battery degradation prediction model M is trained in a federated manner in the sense defined by the article by Brendan McMahan et al. entitled “Communication-Efficient Learning of Deep Networks from Decentralized Data (In Aarti Singh and Jerry Zhu, editors, Proceedings of the 20th International Conference on Artificial Intelligence and Statistics, volume 54 of Proceedings of Machine Learning Research, pages 1273-1282. PMLR, 20-22 Apr 2017. pages 1, 2, 3).
[0085] According to a particular and non-limiting exemplary embodiment, the obtaining (step 31), by the central device 14, of the global model M for predicting the reduction in battery charge level per journey comprises steps 310 to 314.
[0086] In step 310, the central device 14 transmits parameters of an initial local model MO for predicting the decrease in battery charge level per journey to the device 2 of each electric vehicle 1 Ij of the second set 11 of electric vehicles and of each electric vehicle 10i of the first set 10 of electric vehicles.
[0087] In step 311, at least one local model Mj for predicting the decrease in battery charge level per journey is trained by the device 2 of at least one electric vehicle 11 j (learning of said at least one local model Mj).
[0088] Each local model Mj for predicting the decrease in battery charge level per journey is trained by an electric vehicle 1 Ij from data collected by said electric vehicle 1 Ij and from actual decreases in battery charge level of said electric vehicle 1 Ij following a predetermined number of journeys made by said electric vehicle 1 Ij.
[0089] According to a variant, a reduction in a battery charge level after a journey made by an electric vehicle 1 Ij is represented by a percentage of battery charge consumed for this journey.
[0090] A percentage of battery consumed for a journey of an electric vehicle 1 Ij is calculated by the difference between an initial battery charge level of the electric vehicle 1 Ij before the start of the journey and the charge level of this battery after electric vehicle 1 Ij made this journey.
[0091] According to a variant, the data collected by an electric vehicle 1 Ij are at least one of the following data:
[0092] a distance of a journey made by the electric vehicle 1 Ij;
[0093] an average speed along a journey made by the electric vehicle 1 Ij;
[0094] a load (weight) of the electric vehicle 1 Ij along a route;
[0095] data representative of activation of an on-board air conditioning system of the electric vehicle 11 d during a journey;
[0096] data representative of activation of headlights of the electric vehicle 1 Ij along a route;
[0097] an average outside temperature data along a route of the electric vehicle 1 Ij.
[0098] The data collected by an electric vehicle 1 Ij are normalized by maximum values.
[0099] In step 312, the parameters of said at least one local model Mj for predicting a decrease in battery charge level per trained journey (311) are transmitted to the central device 14. The parameters of each local model Mj for predicting a decrease in battery charge level per trained journey are transmitted by the device 2 of the electric vehicle 1 Ij which trained this local model Mj.
[0100] Thus, the parameters of each local model Mj trained by the device 2 of an electric vehicle 11 j are sent to the central device 14.
[0101] In step 313, the central device 14 determines parameters of the global model M for predicting the decrease in battery charge level per journey from the parameters of said at least one local model Mj received.
[0102] According to a variant, the parameters of the global model M are obtained from the parameters of said at least one local model Mj for predicting the decrease in battery charge level per journey according to a method of averaging the parameters of said at least one local model Mj.
[0103] According to a variant, the method of averaging the parameters of the models Mj is similar to that described in section 2 of the aforementioned article by Brendan McMahan et al.
[0104] In step 314, the central device 14 transmits to the device 2 of each electric vehicle 1 Ij, the parameters of the global model M obtained and steps 311, 312, 313 and 314 are iterated.
[0105] According to a variant, steps 311, 312, 313 and 314 are iterated until a stopping condition is verified.
[0106] According to a variant, a stopping condition is verified when a maximum number of iterations is reached.
[0107] According to a variant, a stopping condition is verified when a variation of the pa parameters of the global model M between two successive iterations is less than a threshold value.
[0108] According to a variant, the device 2 of an electric vehicle 1 Ij trains a local model Mj,n for predicting the decrease in battery charge level per journey per initial battery charge level of said electric vehicle 1 Ij and the central device 14 obtains as many global models Mn for predicting the decrease in battery charge level per journey as there are initial battery charge levels. The index n belongs to a set of integer values 1 to N where N is a number of initial battery charge levels.
[0109] According to a variant, the parameters of a global model Mn for a given initial load level are obtained from the parameters of at least one local model Mj,n corresponding to said initial load level according to a method of averaging the parameters of said at least one local model Mj,n.
[0110] Several local models Mj,n for predicting the decrease in battery charge level per journey can thus be trained by the device 2 of an electric vehicle 1 Ij, each corresponding to an initial battery charge level of the electric vehicle 11 j (step 311). The parameters of each of the models Mj,n are transmitted to the central device 14 (step 312) and the central device 14 determines N global models Mn for predicting the decrease in battery charge level per journey from the received local models Mj,n (step 313). The parameters of the global models Mn are transmitted to each electric vehicle 10i and to each electric vehicle 1 Ij (step 314) and steps 311, 312, 313 and 314 are iterated.Following a journey made by an electric vehicle lOi, the device (2) of the electric vehicle lOi selects one of the global models Mn for predicting the decrease in battery charge level per journey as a function of the initial battery charge level at the start of the journey made. The device (2) of the electric vehicle lOi obtains (step 32) an estimate of the decrease in battery charge level per journey from the selected global model Mn. The device 2 of the electric vehicle lOi then calculates (step 33) an actual decrease in battery charge level following said journey made by the electric vehicle lOi. The device 2 of the electric vehicle lOi also obtains (34) an estimate of battery degradation of the electric vehicle lOi by calculating a difference between the estimate of the decrease in battery charge level per journey and the actual decrease in battery charge level.
[0111] According to a variant, 10 models Mj,n (n=1 to 10) are trained: A first model Ml,j is trained for an initial battery charge percentage between 1 and 10%, a second model M2,j is trained for an initial battery charge percentage between 11 and 20%, etc.
[0112] According to a variant, a local model Mj or Mj,n for predicting a decrease in battery charge level per trip and a global model M or Mn for predicting the decrease in battery charge level per trip is implemented by a neural network, also called a deep regression network. The parameters of a local model Mj or Mj,n and the parameters of a global model M or Mn are the internal parameters of the neural network.
[0113] According to a variant, the neural network can for example comprise three layers of neurons. The first layer comprises as many neurons as input data. The second layer is a hidden and fully connected layer comprising 64 neurons and the third layer comprises a single neuron corresponding to an output data. The layers of neurons use a ReLu activation function (from the English Rectified Linear unit, in French rectified linear unit) which allows, basically, to replace negative results with zero.
[0114] The training of a local model Mj or Mj,n consists of determining internal parameters of the neural network so as to minimize a cost function defined as a function of values of reductions in the battery charge level provided by the neural network when data collected by an electric vehicle 1 Ij following journeys made by the electric vehicle 1 Ij, and as a function of values of reductions in the battery charge level of the electric vehicle 1 Ij as they were actually recorded following these journeys.
[0115] Of course, the present invention is not limited to the exemplary embodiments described above but extends to a method for estimating the degradation of an electric vehicle battery of a first set of electric vehicles in communication with a central device which would include secondary steps without thereby departing from the scope of the present invention. The same would apply to a device configured for implementing such a method.
[0116] The present invention also relates to an electric vehicle, for example an automobile or more generally an autonomous electric vehicle with a land motor, comprising the device 2 of [Fig.2].
Claims
Claims
1. Method for estimating battery degradation of an electric vehicle of a first set of electric vehicles in communication with a central device (14), said central device (14) being further in communication with electric vehicles of a second set of electric vehicles, said method comprising the following steps: - transmission (31), by the central device (14) and intended for each electric vehicle of the first and second sets of electric vehicles, of a global model (M) for predicting the decrease in battery charge level per journey obtained from local models (Mj, Mj,n) for predicting the decrease in battery charge level per journey trained by devices (2) of electric vehicles of the second set of electric vehicles, each local model (Mj, Mj,n) prediction of decrease in battery charge level per trip being obtained from data collected by an electric vehicle of the second set of electric vehicles following trips made by said electric vehicle of the second set of electric vehicles and from actual decreases in battery charge level of said electric vehicle of the second set of electric vehicles after said trips;, - following a journey made by an electric vehicle of the first set of electric vehicles, obtaining (32), by the device (2) of said electric vehicle of the first set of electric vehicles, an estimate of a decrease in battery charge level per journey from the global model (M) for predicting a decrease in battery charge level per journey; - calculation (33), by the device (2) of said electric vehicle of the first set of electric vehicles, of an actual decrease in battery charge level following said journey made by said electric vehicle of the first set of electric vehicles; - obtaining (34), by the device (2) of said electric vehicle of the first set of electric vehicles, an estimate of battery degradation of said electric vehicle of the first set of electric vehicles by calculating a difference between the estimate of the decrease in battery charge level per journey and the actual decrease in battery charge level.
2. A method according to claim 1, wherein the emission (31), by the central device (14) and intended for each electric vehicle of the first and second set of electric vehicles, the global model (M) for predicting the decrease in battery charge level per journey comprises the following steps: a) transmission (310), by the central device (14), of parameters of an initial local model for predicting the decrease in battery charge level per journey to the device of each electric vehicle of the first and second set of electric vehicles; b) learning (311) at least one local model (Mj, Mj,n) for predicting a decrease in battery charge level per journey by the device (2) of at least one electric vehicle of the second set of electric vehicles, each local model (Mj, Mj,n) for predicting a decrease in battery charge level per journey is trained by an electric vehicle of the second set of electric vehicles from data collected by said electric vehicle of the second set of electric vehicles and from actual decreases in battery charge level of said electric vehicle of the second set of electric vehicles following a predetermined number of journeys made by said electric vehicle of the second set of electric vehicles; c) transmission (312) of the parameters of said at least one local model (Mj, Mj,n) for predicting a decrease in battery charge level per trained journey to the central device (14), the parameters of each local model (Mj) for predicting a decrease in battery charge level per trained journey are transmitted by the device (2) of the electric vehicle of the second set of electric vehicles which trained said local model (Mj) for predicting a decrease in battery charge level per journey; d) determination (313) by the central device (14), of the parameters of the global model (M) for predicting the decrease in battery charge level per journey from the parameters of said at least one local model (Mj, Mj,n) for predicting the decrease in battery charge level per journey received; e) transmission (314), by the central device (14) and to the device (2) of each electric vehicle of the first and second sets of electric vehicles, the parameters of the global model (M) for predicting the decrease in battery charge level by path obtained and steps b), c), d), and e) are iterated.
3. Method according to claim 2, for which the parameters of the global model (M) for predicting the decrease in battery charge level per journey are determined (312) from the parameters of said at least one local model (Mj, Mj,n) for predicting the decrease in battery charge level per journey according to a method of averaging the parameters of said at least one local model (Mj, Mj,n) for predicting the decrease in battery charge level per journey.
4. Method according to one of claims 1 to 3, for which the device (2) of an electric vehicle of the second set trains a local model (Mj,n) for predicting the decrease in battery charge level per journey per initial battery charge level of said electric vehicle of the second set of electric vehicles and the central device (14) obtains as many global models (Mn) for predicting the decrease in battery charge level per journey as there are initial battery charge levels.
5. Method according to claim 4, for which the parameters of a global model (Mn) for predicting the decrease in battery charge level per journey for a given initial charge level are obtained from the parameters of at least one local model for predicting the decrease in battery charge level per journey corresponding to said initial charge level according to a method of averaging the parameters of said at least one local model for predicting the decrease in battery charge level per journey corresponding to said initial charge level.
6. Computer program comprising instructions for implementing the method according to any one of the preceding claims, when these instructions are executed by a processor.
7. A computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the method according to one of claims 1 to 5.
8. Device (2) comprising a memory (201) associated with at least one processor (200) configured for the implementation of at least one step of the method according to any one of claims 1 to 5.
9. Electric vehicle (10i, 11j) comprising the device (2) according to claim 8.
10. System (1) for estimating the degradation of an electric vehicle battery of a first set of electric vehicles comprising a central device (14), said central device (14) being in communication with electric vehicles of the first set of electric vehicles and electric vehicles of a second set of electric vehicles, characterized in that said central device (14) implements at least one step of the method according to one of claims 1 to 5 and each electric vehicle of the first and second sets of electric vehicles is an electric vehicle according to claim 9.
Citation Information
Patent Citations
Determining the aging state of an energy storage device
DE102020124182A1
Method and device for determining the state of charge of a battery in a battery-powered machine
DE102020212293A1
Method and system for predicting battery degredation
GB2586655A
