METHOD FOR PREDICTING AT LEAST ONE RISK OF FAILURE OF A BALANCE CONTROL SYSTEM OF AN ELECTRIC OR HYBRID MOTOR VEHICLE

Abstract

The invention relates to a method for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle comprising the following steps: a first step of collecting data (E1); a step of storing (E2) and preprocessing (E3) the data by normalizing the data and classifying the data, said data forming a model; a second step of collecting and processing new data (E4); a step of training (E5) a machine learning algorithm by comparing the data and the new data; a step of validating (E6) the model, when the data and the new data are identical; a third step of collecting data (E7) and a step of predicting (E8) said at least one risk of failures by developing a prognosis establishing the probability of occurrence of a failure when the model has been validated. Figure 1

Description

Title of the invention: METHOD FOR PREDICTING AT LEAST ONE RISK OF FAILURE OF A BALANCE CONTROL SYSTEM OF AN ELECTRIC OR HYBRID MOTOR VEHICLE

[0001] The invention relates to electric or hybrid motor vehicles, and more particularly to the battery balancing control system of these vehicles.

[0002] Known from the prior art is a patent application US20210057920 which describes a method for applying artificial intelligence to a battery. The battery comprises a plurality of sensors. The method comprises a step of receiving a learning model corresponding to a neural network. The neural network of an artificial intelligence is a structure inspired by the human brain designed to learn from collected data. The neural network implements interconnected artificial neurons whose role is to process the data. The method comprises a step of collecting data by the sensors of said battery. The model is then executed by a processor, the processor generating outgoing information from said model. The outgoing information is then transmitted to a charge controller which will emit a signal towards said battery in order to manage the recharging of said battery.Indeed, the signal can transmit a request to stop recharging the said battery, for example. However, there are still drawbacks. Indeed, the method described is not capable of detecting potential failures of the said battery. Thus, it is not possible to predict a breakdown of the vehicle due to these failures.

[0003] The objective of the present invention is to remedy these drawbacks by proposing a method for predicting a failure of the battery of an electric or hybrid motor vehicle.

[0004] To achieve this objective, the invention proposes a method for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle, said vehicle comprising a first battery and a second battery, said balancing control system allowing the second battery to be recharged by the first battery, said balancing control system comprising at least one sensor configured for the detection of a predetermined parameter, remarkable in that said method comprises the following steps: - a first step of collecting data by said at least one sensor of said at least one predetermined parameter; - a step of storing said data; - a data pre-processing step by data normalization and data classification, said data forming a model; - a second stage of collection and processing of new data; - a step of training a machine learning algorithm using the new data to assess the reliability of said model by comparing the data and the new data; - a model validation step, when the data and the new data are identical; - a third stage of data collection; - a step of predicting said at least one risk of failure by said machine learning algorithm from said model and based on said data collected during the third data collection step by developing a prognosis establishing the probability of a failure occurring when the model has been validated during the validation step.

[0005] Thanks to the invention, it is possible to predict a future breakdown by detecting a risk of failure beforehand. Thus, the breakdown can be avoided by controlling the vehicle.

[0006] Preferably, said method comprises an alternative step to the validation step during which said model is updated using the new data when the data from the first data collection step and the new data from the second data collection step are different.

[0007] Updating the model helps to promote the issuing of reliable predictions.

[0008] Advantageously, during the processing step, said machine learning algorithm is a support vector machine algorithm.

[0009] The vector machine algorithm provides data classification and regression.

[0010] Advantageously, said vehicle comprises a dashboard, said method comprising a step of issuing an alert on said dashboard when said at least one risk of failure is predicted during the prediction step.

[0011] Thus, the driver knows that he must have his vehicle checked before it breaks down.

[0012] Preferably, said first battery is a high voltage battery and said second battery is a 12V battery.

[0013] Advantageously, said at least one sensor is a sensor for measuring a value of electrical energy transmission between the first battery and the second battery, a sensor for measuring a voltage value of the first battery and / or the second battery, a sensor for measuring a value of the state of charge of the first battery and / or the second battery, a sensor for measuring a value of the recharging time of the first battery and / or the second battery.

[0014] These are relevant parameters for detecting whether the first battery is able to correctly recharge said second battery.

[0015] Advantageously, said method comprises a step of automatically generating a report presenting said at least one predicted failure risk.

[0016] Generating a report simplifies the diagnosis of a vehicle following a breakdown.

[0017] The invention also relates to a computer program comprising instructions which, when the program is executed by a computer, cause the latter to implement the steps of the method for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle described above.

[0018] Furthermore, the invention relates to a system for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle comprising a computer implementing said computer program previously described.

[0019] Furthermore, the invention relates to an electric or hybrid motor vehicle comprising a system for predicting at least one risk of failure of a previously described balancing control system.

[0020] The invention will be further detailed by the description of a non-limiting embodiment, and on the basis of the appended figure illustrating the invention, in which [Fig.l] schematically illustrates, in the form of a flowchart, a method for predicting at least one risk of failure of a balancing control system for an electric or hybrid motor vehicle according to an embodiment of the invention.

[0021] [Fig.l] schematically illustrates, in the form of a flowchart, a method for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle. The vehicle comprises an electric motor, a first battery and a second battery. Preferably, the first battery is a high-voltage battery allowing the transmission of electrical energy to the electric motor of said vehicle, and the second battery is a 12V battery. The 12V battery has the role of ensuring the operation of the electrical equipment of the vehicle even when said vehicle is not in operating condition. In addition, the 12V battery provides a safety function in the event that the high-voltage battery suffers a failure. The balancing control system is a device designed to monitor and manage the first battery and the second battery.The balancing control system ensures that the second battery is recharged by the first battery. Advantageously, the balancing control system. comprises a database. The database is a structured set of data so that each data item can be accessed, managed and updated if necessary. The balancing control system comprises at least one sensor configured to measure a predetermined parameter. Advantageously, the sensor is a sensor for measuring a value of electrical energy transmission between the first battery and the second battery, a sensor for measuring a voltage value of the first battery and / or the second battery, a sensor for measuring a value of the state of charge of the first battery and / or the second battery, a sensor for measuring a value of the recharging time of the first battery and / or the second battery. During a first data collection step E1, the data from said at least one sensor are collected.If we take the sensors mentioned above, the data therefore correspond to a value of electrical energy transmission between the first battery and the second battery, a voltage value of the first battery, a voltage value of the second battery, a value of the state of charge of the first battery, a value of the state of charge of the second battery, a value of the recharging time of the first battery and / or a value of the recharging time of the second battery. During a storage step E2, the data is stored, for example in the database, in order to be accessible for said balancing control system. During a preprocessing step E3, the model undergoes a normalization of this data. Data normalization is a process used to adjust the collected data relative to each other.Normalization standardizes the mean and standard deviation of a data distribution to improve the performance of models and analyses based on these data. Preferably, outliers are removed. An outlier is a data item that differs from other data collected during the same observation, for the same subject, under the same conditions. Outliers are, in most cases, the result of a measurement error or a technical problem. The presence of outliers distorts the results, which is why it is preferable to remove them. At the end of the processing, the data forms a model. In a second step of collecting new data E4, new data from the at least one sensor are collected. The new data from the second step of collecting new data E4 are also preprocessed by data normalization, and preferably by removing outliers.In a training step E5, the new data previously collected in the second new data collection step E4 are compared to the data collected in the first data collection step El by a machine learning algorithm. The algorithm learns by comparing the model trained by the data from the first collection step El and the new data. data from the second collection step E4, to make predictions about the future evolution of the data from said at least one sensor. The reliability of said model is then evaluated by comparing the data from the first collection step El and the new data from the second collection step E4. If the data from the first collection step El and the new data from the second collection step E4 are identical, the model is validated during a validation step E6. In other words, the model is considered reliable. If the data from the first collection step El and the new data from the second collection step E4 diverge during the training step, then the model is not validated because it is not considered reliable.In this case, the model is updated during an alternative step to said validation step E6 using the new data from the second collection step E4. Preferably, the machine learning algorithm used in the context of the invention are support vector machine algorithms. Support vector machines are machine learning algorithms allowing the classification and regression of data. Data regression is a statistical method allowing a value to be approximated from the other values which are correlated with it. It is also referred to as curve fitting when the data are represented graphically. During a third data collection step E7, data are collected again in order to compare them to the model. During a detection step E8, at least one risk of failure is predicted.For example, it is determined that it is likely that a failure of said balancing control system will occur at the level of energy transmission between the first accumulator and the second accumulator, the voltage of the first accumulator, the voltage of the second accumulator, the state of charge of the first accumulator, the state of charge of the second accumulator, the recharging time of the first accumulator and / or the recharging time of the second accumulator. Advantageously, the vehicle comprises a dashboard and the method comprises a step of issuing an alert on said dashboard when said at least one risk of failure is detected. This makes it possible to warn the driver of the presence of said risk of failure. Thus, the driver can have the vehicle checked by a specialist before it causes a breakdown.Advantageously, the method comprising a step of automatically generating a report presenting said at least one risk of failure is predicted. The method emits a probability of the occurrence of a failure, in other words the method emits a risk of failure. The invention also relates to a computer program comprising instructions which, when the program is executed by a computer, cause the latter to implement the steps of the method described above. In addition, the invention relates to a . system for predicting at least one risk of failure by a balancing control system for an electric or hybrid motor vehicle implementing said method. Finally, the invention relates to an electric or hybrid motor vehicle comprising such a system for predicting at least one risk of failure by a balancing control system.

Claims

Claims

1. Method for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle, said vehicle comprising a first battery and a second battery, said balancing control system allowing the second battery to be recharged by the first battery, said balancing control system comprising at least one sensor configured for the detection of a predetermined parameter, characterized in that said method comprises the following steps: - a first step of collecting data (El) by said at least one sensor of said at least one predetermined parameter; - a step of storing (E2) said data; - a step of preprocessing (E3) the data by normalizing the data and classifying the data, said data forming a model; - a second step of collecting and processing new data (E4);- a step (E5) of training a machine learning algorithm using the new data to evaluate the reliability of said model by comparing the data and the new data; - a step (E6) of validating the model, when the data and the new data are identical; - a third step (E7) of collecting new data; - a step (E8) of predicting said at least one risk of failure by said machine learning algorithm from said model and according to said data collected during the third step (E7) of collecting new data by developing a probability of occurrence of a failure when the model has been validated during the validation step (E6).;

2. Method according to claim 1 characterized in that said method comprises an alternative step to the validation step (E6) during which said model is updated using the new data when the data from the first data collection step (El) and the new data from the second data collection step (E4) are different.

3. Method according to claim 1 or 2 characterized in that, during the training step (E5), said learning algorithm au- tomatic is a support vector machine algorithm.

4. Method according to any one of claims 1 to 3 characterized in that said vehicle comprises a dashboard, said method comprising a step of issuing an alert on said dashboard when said at least one risk of failure is predicted during the prediction step (E8).

5. Method according to any one of claims 1 to 4 characterized in that said first battery is a high voltage battery and said second battery is a 12V battery.

6. Method according to any one of claims 1 to 5 characterized in that said at least one sensor is a sensor for measuring a value of electrical energy transmission between the first battery and the second battery, a sensor for measuring a voltage value of the first battery and / or the second battery, a sensor for measuring a value of the state of charge of the first battery and / or the second battery, a sensor for measuring a value of the recharging time of the first battery and / or the second battery.

7. Method according to any one of claims 1 to 6 characterized in that said method comprises a step of automatically generating a report presenting said at least one predicted risk of failure.

8. Computer program comprising instructions which, when the program is executed by a computer, cause the latter to implement the steps of the method for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle according to any one of claims 1 to 7.

9. System for predicting at least one risk of failure of a balancing control system of an electric or hybrid motor vehicle comprising a computer implementing the computer program according to claim 8.

10. Electric or hybrid motor vehicle comprising a system for predicting at least one risk of failure of a balancing control system according to claim 9.

Images