KR20230080112A - Battery diagnostic device for predicting the current state of the battery - Google Patents

Abstract

The present invention relates to a battery diagnostic device for predicting a current state of a battery.
The present invention includes a battery data measuring unit that collects voltage, current and ambient temperature, battery current temperature, battery SOC, accumulated charge/discharge amount and date information at predetermined intervals; a data collection unit that collects data measured through the battery data measurement unit; a digital twin unit generating a virtual battery model identical to that of the user's battery in a virtual space through the data collected through the data collection unit; a standard data conversion unit generating battery standard data while charging and discharging the virtual battery model created in the virtual space under standard temperature, standard pressure, 1C, and constant current conditions; and a battery state and life prediction unit that predicts a battery state and a lifespan of the battery using the generated battery standard data.

Description

Battery diagnostic device for predicting the current state of the battery}

The present invention relates to a battery diagnosis device for predicting a current state of a battery, and more particularly, to a battery diagnosis device capable of predicting a current state of a battery regardless of an environment when predicting a current state of a battery.

Batteries operate under various operating conditions, and their performance characteristics (charge/discharge amount, internal resistance, aging, self-discharge rate) vary greatly depending on the operating conditions (current size during charging/discharging, temperature, etc.). It makes state prediction difficult.

In the case of a conventional battery current state prediction algorithm, output data generated under a specific condition is predicted using a result under a specific operating condition according to a cycle as input data.

Therefore, in the case of the conventional battery current state prediction algorithm, there is a disadvantage in that it is impossible to predict when operating conditions change, or it is impossible to use a model learned with data created under other conditions.

In addition, in the prior art, when charging and discharging according to users, only a part (20% to 40%) is performed, and in general, no one charges / discharges from 0 to 100%, which makes prediction difficult.

A technology that has been actively researched recently is an SOH prediction method based on deep learning. However, most studies use models that can be predicted only when specific operating conditions and charge/discharge amounts are fixed, making it difficult to use them practically.

The present invention has been devised to solve the conventional problems, building a digital twin model with data obtained under various conditions, making a virtual environment a standard condition, running a simulation, and through this, converting data obtained under various conditions into a standard condition. By standardizing and learning the standardized data to a deep learning model, a battery diagnosis device for predicting the current state of a battery can be diagnosed and predicted with higher accuracy without a separate method such as transfer learning. want to provide

The present invention combines a method of converting battery data obtained during operation into a standard state regardless of various operating conditions and states of the battery and a deep learning model to practically predict the current state and life of the battery for data under various conditions. It is intended to provide a battery diagnostic device for predicting the current state of the battery.

The object of the present invention is not limited to the object mentioned above, and other objects not mentioned will be clearly understood by those skilled in the art from the description below.

To achieve the above object, a battery diagnosis apparatus for predicting a current state of a battery according to an embodiment of the present invention includes a battery data measuring unit collecting battery state information at predetermined intervals; a data collection unit that collects data measured through the battery data measurement unit; a digital twin unit generating a virtual battery model identical to that of the user's battery in a virtual space through the data collected through the data collection unit; a standard data conversion unit generating battery standard data while charging and discharging the virtual battery model created in the virtual space under standard temperature, standard pressure, 1C, and constant current conditions; and a battery state and life prediction unit for predicting a battery state and a lifespan of the battery using the generated battery standard data.

The battery data measuring unit does not discriminate data between battery cells and packs/modules, and measures all data of the pack/module and data of each cell.

The data collection unit, the battery state information, measurement date, voltage (V), current (A), temperature (K), current charge amount (SOC), cumulative discharge amount / charge amount ( Ah) according to the recording time.

The digital twin unit preferably creates the same model as the battery to be measured in a virtual space.

The digital twin unit estimates the parameters of the physics-based model by applying an optimization algorithm based on the stored data, and through this, builds the same physics-based model of the battery reflecting the current state of the battery, which is the data collection target, in virtual space. can

The standard data conversion unit simulates the digital twin model built in virtual space according to the current battery state under standard operating conditions, and separately stores the result.

It is preferable that the standard data conversion unit stores the curve obtained by performing the simulation from SOC 0 to 100% with the digital twin model and the charge/discharge amount obtained at this time as a standard charge amount and a standard discharge amount. Here, the curve is a value calculated when SCO is the x-axis and voltage is the y-axis.

The battery state and life prediction unit may include a state prediction unit for determining a standardized State of Health (SOH) based on a current battery charge amount and a battery discharge amount from a standard result, and transmitting and storing the result to a data storage unit; It predicts the battery life reduction pattern by applying the standardized SOH stored for each day and date to the deep learning model, and includes a life prediction unit that provides guidance according to battery life and safety diagnosis based on the prediction result.

The state and life predicting unit converts the standard value into a storage unit, and predicts the state of the battery using the stored standard value.

The state prediction unit may predict a battery state based on a percentage of an initial SOH (or reference SOH).

The life predicting unit may predict using various deep learning models based on standard SOH according to battery measurement cycles.

According to an embodiment of the present invention, in the case of the existing algorithm, the output of a specific condition is predicted as a result of a specific operating condition according to the CYCLE with INPUT data, and if the operating condition changes, it is impossible to predict or learn from data made under other conditions However, according to the present invention, by making various conditions into standard conditions and learning them in a deep learning model, it is possible to diagnose the current state and predict lifespan with higher accuracy without a separate method such as transfer learning. It has the effect of doing it.

According to an embodiment of the present invention, a digital twin model can be built based on data during partial charge/discharge, and the digital twin model can be operated in a virtually standard state to create a full charge/discharge curve in a virtual standard state. There is an effect.

According to an embodiment of the present invention, a digital twin model is built based on data of various operating conditions (C-rate, ambient temperature, etc.), and the digital twin model is operated in a virtually standard state to fully charge the virtual standard state. By creating a /discharge curve, when using existing algorithms, specific operating/environmental conditions are forced on the user, whereas various existing condition diagnosis and life prediction algorithms can be applied to any operating condition/state data. It has the effect of enabling

1 is a block diagram illustrating a battery diagnosis apparatus for predicting a current state of a battery according to the present invention.
2 is a configuration block diagram for explaining the detailed configuration of the digital twin unit of FIG. 1;
3 is a reference diagram for explaining an embodiment of the physically-based model of FIG. 2;
4 is a reference diagram for explaining another embodiment of the physically-based model of FIG. 2;
5 is a reference graph for explaining battery state prediction;
FIG. 6 is a block diagram illustrating detailed configurations of a battery state and life prediction unit of FIG. 1 .

Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the detailed description of the following embodiments taken in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various different forms, only these embodiments make the disclosure of the present invention complete, and common knowledge in the art to which the present invention belongs. It is provided to fully inform the holder of the scope of the invention, and the present invention is only defined by the scope of the claims. Meanwhile, terms used in this specification are for describing the embodiments and are not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase. As used herein, "comprises" and/or "comprising" means that a stated component, step, operation, and/or element is the presence of one or more other components, steps, operations, and/or elements. or do not rule out additions.

1 is a block diagram illustrating a battery diagnosis apparatus for predicting a current state of a battery according to the present invention.

1 is a battery diagnosis device for predicting a current state of a battery according to an embodiment of the present invention includes a battery data measurement unit 100, a data collection unit 200, a digital twin unit 300, and a standard data conversion unit 400 and a battery state and life prediction unit 500 .

The battery data measuring unit 100 collects voltage, current, ambient temperature, current battery temperature, battery SOC, accumulated charge/discharge amount, and date information at predetermined intervals. The battery data measurement unit 100 does not discriminate data between battery cells and packs/modules, and measures all data of the module/pack and data of each cell.

The data collection unit 200 collects data measured through the battery data measurement unit 100 . The data collection unit 200 records today's date, voltage (V), current (A), temperature (K), current charge amount (SOC), cumulative discharge/charge amount (Ah), etc. of the battery recorded through the BMS or sensor. Data is transmitted according to the recording time (usually 1s interval). That is, today's date, voltage, current, and temperature of the battery according to time are measured and stored through the battery BMS or sensor. At this time, it does not matter even if it is SOC data of a part range. As an example, on August 20, 2021, at 36 degrees, from 20% to 40%, the user collects data of high-speed charging in front of the house at 5C-rate.

The digital twin unit 300 creates a battery virtual model identical to the user's battery in a virtual space through the data collected through the data collection unit 200 .

The digital twin unit 300 creates the same model as the user's battery (company S model, used for 1 year, existing in company A's electric vehicle) in the virtual space. The digital twin unit 300 estimates the parameters of the physics-based model by applying an optimization algorithm based on the stored data, and through this, builds the same physics-based model of the battery reflecting the current state of the battery, which is a data collection target, in virtual space. Here, the optimal algorithm may be a non-linear optimization algorithm such as a genetic algorithm.

The digital twin unit 300 simulates a battery using either a partial differential equation-based electrochemical model or a circuit model.

As shown in FIG. 2 , the digital twin unit 300 includes a physically based model 310 and an optimization algorithm 320 .

Here, the physics-based model 310 simulates the battery being measured, and as shown in FIG. 3, the electrochemical model (P2D) by Reference: A. Jokar et al., 2016, Journal of Power Sources and As shown in FIG. 4, a circuit model (ECM, Equivalent Circuit Model) by Reference: X. Lai et al., 2018, Electrochimica Acta can be used.

The optimization algorithm 320 estimates the parameter (X) of the physics-based model and builds the same physics-based model of the battery reflecting the current state of the battery, which is a data collection target, in a virtual space.

To this end, the fitness function when the optimization algorithm 320 is defined as the parameter (x) of the battery model, the measured voltage data (V), and the measured temperature data (T) is the optimization algorithm (model parameter estimation) and the battery. It includes a physically based model 310 that simulates.

It can be calculated by [Equation 1].

[Equation 1]

) 이고, Z는 배터리 수치해석 모델에서 생성한 데이터로 구성된 벡터로서 실제 센서에서 얻을 수 있는 데이터와 대응되는 전압, 온도 등에 해당하며, N_데이터는 각 파라미터에 해당하는 데이터의 개수이다. Here, the parameter (X) is a model parameter vector to be estimated (

), Z is a vector composed of data generated from the battery numerical analysis model, and corresponds to voltage and temperature corresponding to data obtained from actual sensors, and N _data is the number of data corresponding to each parameter.

So, among possible parameter values, find a parameter that minimizes this fitness function (difference between actual data and model output), and set this parameter as a parameter that can represent the current battery state.

And, the battery model composed of these parameters is a digital twin model using [Equation 2] below.

[Equation 2]

As an optimization algorithm that can be used at this time, any nonlinear optimization algorithm such as a genetic algorithm may be used. Any of the commonly used models can be used, and the example below shows two representative electrochemical models (P2D) and circuit models (ECM, Equivalent Circuit Model). That is, [Equation 2] means that, as described above, the parameter that minimizes the fitness funciton among possible parameter values is set as an optimal parameter, that is, a parameter that can represent the current battery state.

The standard data conversion unit 400 generates standard battery data while charging and discharging the battery virtual model created in the virtual space under standard temperature, standard pressure, 1C, and constant current conditions. Here, in the case of standard operating conditions, the user can set it. In general, 1C-rate, standard temperature pressure (SATP: temperature 25 °C (298.15 K), atmospheric pressure), which indicates that a fully charged 1Ah battery is discharged at 1A for 1 hour. 1 bar (100000 Pa)). And save the result with the date. The standard data conversion unit 400 simulates the digital twin model built in virtual space according to the current battery state under standard operating conditions, and separately stores the result.

으로 바꿔서 저장한다. 여기서, 커브는 SCO를 x축으로 하고, 전압을 y축으로 할 때 산출되는 값이다. The standard data conversion unit 400 performs a simulation that performs model simulation from SOC 0 to 100% or cut-off of the battery for user convenience with a digital twin model, and converts the curve and the charge/discharge amount obtained at this time into standard charge amount and standard discharge amount. Save. In other words, the data at that point in time are converted to standard conditions.

Change to and save. Here, the curve is a value calculated when SCO is the x-axis and voltage is the y-axis.

The battery state and life prediction unit 500 predicts the state of the battery and the life of the battery using the generated battery standard data. That is, as shown in FIG. 5 , the battery state and life prediction unit 500 may predict predicted data at an arbitrary point in time using standardized standardized battery state data.

As shown in FIG. 6 , the battery state and life prediction unit 500 includes a state prediction unit 510 , a data storage unit 520 and a life prediction unit 530 .

The state prediction unit 510 determines the state of health (SOH) in a standardized state based on the current charge amount and discharge amount of the battery from the standard result, and transmits the result to the data storage unit 520 for storage. The life prediction unit 510 predicts a battery life reduction pattern based on a deep learning model based on the standardized SOH stored for each day, and provides guidance according to battery life and safety diagnosis based on the prediction result.

At this time, the state prediction unit 510 may predict the battery state based on what percentage of the initial SOH (or reference SOH).

In addition, the data storage unit 520 of the state and life prediction unit 500 does not store general SOH, but reduces it to a standard value and stores it. The standard value stored in this way is used to predict the state of the battery in the future.

The life prediction unit 530 predicts using various deep learning models based on the standard SOH according to the cycle. As one example, there is a long short-term memory (LSTM) model for time-series forecasting.

In the case of existing algorithms, INPUT data predicts output under specific conditions as a result of specific operating conditions according to CYCLE, and when operating conditions change, prediction is impossible or a model trained with data made under other conditions cannot be used. , According to the present invention, when various conditions are made into standard conditions and learned in a deep learning model, there is an effect of diagnosing the current state and predicting lifespan with higher accuracy without a method such as a separate transfer learning.

According to an embodiment of the present invention, a digital twin model can be built based on data during partial charge/discharge, and the digital twin model can be operated in a virtually standard state to create a full charge/discharge curve in a virtual standard state. There is an effect.

According to an embodiment of the present invention, a digital twin model is built based on data of various operating conditions (C-rate, ambient temperature, etc.), and the digital twin model is operated in a virtually standard state to fully charge the virtual standard state. By creating a /discharge curve, when using existing algorithms, specific operating/environmental conditions are forced on the user, whereas various existing condition diagnosis and life prediction algorithms can be applied to any operating condition/state data. It has the effect of enabling

In the above, the configuration of the present invention has been described in detail with reference to the accompanying drawings, but this is only an example, and various modifications and changes within the scope of the technical idea of the present invention to those skilled in the art to which the present invention belongs Of course this is possible. Therefore, the scope of protection of the present invention should not be limited to the above-described embodiments, but should be defined by the description of the claims below.

Claims (11)

a battery data measuring unit that collects battery state information at predetermined intervals;
a data collection unit that collects data measured through the battery data measurement unit;
a digital twin unit generating a virtual battery model identical to that of the user's battery in a virtual space through the data collected through the data collection unit;
a standard data conversion unit generating battery standard data while charging and discharging the virtual battery model created in the virtual space under standard temperature, standard pressure, 1C, and constant current conditions; and
A battery diagnosis device for predicting a current state of a battery comprising a battery state and life prediction unit for predicting a state of a battery and a lifespan of the battery using the generated battery standard data.
According to claim 1,
The battery data measurement unit,
A battery diagnostic device for predicting the current state of a battery, characterized in that it does not distinguish data between battery cells and packs/modules and measures both the entire data of the packs/modules and the data of each cell.
According to claim 1,
The data collection unit,
One or more of the battery's measurement date, voltage (V), current (A), temperature (K), current charge (SOC), and cumulative discharge/charge (Ah) information collected through the BMS or sensor is recorded according to the recording time. A battery diagnostic device for predicting a current state of a battery, characterized in that for transmitting.
According to claim 1,
The digital twin unit,
A battery diagnostic device for predicting the current state of a battery, characterized in that it creates the same model as the battery to be measured in a virtual space.
According to claim 4,
The digital twin unit,
The digital twin unit estimates the parameters of the physics-based model by applying an optimization algorithm based on the stored data, and through this, builds the same physics-based model of the battery reflecting the current state of the battery, which is the target of data collection, in virtual space. Battery diagnostic device for current state prediction.
According to claim 5,
The standard data conversion unit,
A battery diagnostic device for predicting the current state of a battery, characterized in that it simulates a digital twin model built in virtual space to suit the current battery state under standard operating conditions and separately stores the result.
According to claim 6,
The standard data conversion unit,
Battery diagnostic device for predicting the current state of the battery, characterized in that for storing the curve obtained by performing the simulation from SOC 0 to 100% with the digital twin model and the charge / discharge amount obtained at this time, the standard charge amount and the standard discharge amount.
According to claim 1,
The battery state and life prediction unit,
A state prediction unit that determines the state of health (SOH) in a standardized state based on the current battery charge amount and battery discharge amount from the standard result and transmits and stores the result to the data storage unit;
Battery diagnosis for predicting the current state of the battery, including a life prediction unit that predicts the life reduction pattern of the battery by applying the standardized SOH stored for each date to the deep learning model, and provides guidance according to the battery life and safety diagnosis based on the prediction result Device.
According to claim 8,
The state and life prediction unit,
A battery diagnostic device for predicting a current state of a battery, characterized in that the value is reduced to a standard value, stored in a storage unit, and a state of the battery is predicted using the stored standard value.
According to claim 9,
The state prediction unit,
A battery diagnostic device for predicting the current state of a battery, characterized in that it predicts the battery state based on the percentage of the initial SOH (or reference SOH).
According to claim 8,
The life prediction unit,
A battery diagnostic device for predicting the current state of a battery, characterized in that it predicts using various deep learning models based on the standard SOH according to the battery measurement cycle.

Images