WO2021201387A1 - Neural network-based battery capacity estimation method and device - Google Patents

Abstract

The present invention relates to a neural network-based battery capacity estimation method and device, and provides a method for training a neural network in a computing device for constant-current charge-based battery capacity estimation, the method comprising the steps of: collecting the remaining capacities of a battery from an arbitrary charge starting point to a point when a maximum voltage is reached during charging; and inputting the collected remaining capacities to the neural network to train the neural network.

Description

Method and apparatus for estimating battery capacity based on neural network

The present invention relates to a method for estimating battery capacity, and more particularly, to a method and apparatus for estimating battery capacity based on a neural network.

In recent years, as the demand for portable electronic products such as notebook computers, video cameras, and portable telephones has rapidly increased, and energy storage batteries, robots, and satellites have been developed in earnest, high-performance secondary rechargeable batteries that can be repeatedly charged and discharged Batteries are being actively researched.

Currently commercialized batteries include nickel-cadmium batteries, nickel-hydrogen batteries, nickel-zinc batteries, and lithium secondary batteries. Among them, lithium secondary batteries have almost no memory effect compared to nickel-based secondary batteries, so charging and discharging are free, Due to its advantages such as a very low self-discharge rate and high energy density, it is receiving a lot of attention.

However, in a lithium secondary battery, when charging and discharging are repeated due to the characteristics of the battery, lithium ions start to accumulate on the surface of the positive electrode or negative electrode, not between the layered structures. Due to this, the capacity of the battery is gradually reduced, and in the worst case, the separator that prevents the short circuit of the battery is pierced, which may also cause a problem in the stability of the battery.

In order to measure the changed battery capacity, there is a method of calculating the total amount of charge while gradually discharging the battery using equipment. Although this method is the most accurate, it is time consuming and requires additional equipment.

In addition, there is a method of estimating through mathematical modeling of the battery, but mathematical modeling is not easy due to the high nonlinearity of the battery, and the mathematical model must be calculated every time the battery characteristics are changed.

On the other hand, the learning-based capacity estimation algorithm using machine learning includes a “CV (Constant Voltage) charging-based battery capacity estimation algorithm” and “CC (Constant Current) charging-based battery capacity estimation algorithm”. However, each algorithm has difficulties when used in the actual field as described below.

That is, the CV charging-based battery capacity estimation algorithm requires full charging to measure the characteristics displayed in the charging period. In addition, when the cell balancing function operates on the battery packs connected in series, the method cannot be used because the charging section characteristic cannot be constantly measured due to the cell balancing function.

In addition, since the CC charging-based battery capacity estimation algorithm estimates the battery capacity by using the characteristics that exist in the low SOC (State of Charge) section, in order to use the algorithm, the battery is used up to the low SOC section just before charging. need to discharge. As such, since the CC charging-based battery capacity estimation algorithm discharges the battery to a low SOC section just before charging, the lifespan of the entire battery is reduced, and there is a problem that it takes a long time.

[Prior art literature]

(Patent Document 1) Domestic Patent Publication No. 10-2018-0121317

(Patent Document 2) Domestic Registered Patent No. 10-2032229

The present specification has been devised to solve the above problems, and in CC charging, a method and apparatus for estimating battery capacity based on a neural network capable of estimating the capacity of a battery only with data from an arbitrary charging start time to the charging end Its purpose is to provide

Another object of the present invention is to provide a method and apparatus for estimating battery capacity based on a neural network capable of estimating battery capacity without having to discharge to a low SOC section just before charging.

According to an embodiment of the present specification, for achieving this object, the neural network learning method according to the present specification is a method for learning a neural network in a computing device for constant current charging-based battery capacity estimation, any collecting the remaining capacity of the battery from the start of charging of the to the time when the maximum voltage is reached during charging; and learning the neural network by inputting the collected residual capacity into the neural network.

Preferably, the neural network is implemented through a multi-layer perceptron or a fully connected layer.

According to another embodiment of the present specification, the neural network learning method according to the present specification is a method for learning a neural network in a computing device for estimating battery capacity based on constant current charging, during charging from an arbitrary charging start time. collecting the charging voltage up to the point in time when the maximum voltage is reached; and learning the neural network by inputting a charging time, a collected charging voltage, and a battery remaining capacity at the arbitrary charging start time to the neural network.

Preferably, in the collecting of the charging voltage, the voltage curve of the charging section is divided by N (N is a natural number) at equal intervals to collect the charging voltage for each section.

Preferably, the method further comprises the step of assigning a preset weight for each type of battery to the neural network.

Preferably, the neural network is implemented through a multi-layer perceptron or a fully connected layer.

According to another embodiment of the present specification, the neural network learning apparatus according to the present specification includes a memory for storing one or more instructions: and by executing the stored one or more instructions, from any charging start time to the maximum during charging. It includes one or more processors that collect the charging voltage until the voltage is reached, and configure a neural network for estimating the capacity of the battery by receiving the charging time, the collected charging voltage, and the remaining capacity of the battery at any charging start time. .

Preferably, the one or more processors divide the voltage curve of the charging section into N equal parts (N is a natural number) at equal intervals, and collect the charging voltage for each section.

Preferably, the one or more processors assign a preset weight to the neural network for each battery type.

Preferably, the neural network is implemented through a multi-layer perceptron or a fully connected layer.

As described above, according to the present specification, the neural network-based battery capacity estimation for estimating the battery capacity by inputting and learning the remaining capacity of the battery collected from the start of any charging to the time when the maximum voltage is reached during charging into the neural network. By providing the method and apparatus, it is possible to estimate the battery capacity without the need to discharge the battery to a low SOC section just before charging.

In addition, the data server 130 according to the present invention replaces the remaining capacity when the maximum voltage is reached during charging and uses the charging time, the charging voltage, and the remaining capacity at the charging start time as a feature point for learning. By providing a method and apparatus for estimating a battery capacity, it is possible to exclude a change in a point of a residual capacity at a point in time when a maximum voltage is reached during charging due to an increase in internal resistance due to aging of the battery in estimating the battery capacity.

1 is a block diagram showing a schematic configuration of a battery capacity estimation system according to an embodiment of the present invention;

2 is a diagram showing a voltage curve and a dV/dQ curve;

3 is a block diagram showing an example of hardware capable of realizing the function of the battery capacity estimation device according to the embodiment of the present invention;

4 is a block diagram showing an example of a function of the battery capacity estimation device according to the embodiment of the present invention;

5 is a flowchart illustrating a method for estimating battery capacity based on a neural network according to a first embodiment of the present invention;

6 is a flowchart illustrating a method for estimating battery capacity based on a neural network according to a second embodiment of the present invention;

7 is a diagram showing an implementation of a fully connected layer, and

8 is a diagram for explaining the structure and implementation of a multi-perceptron.

It should be noted that the technical terms used herein are used only to describe specific embodiments, and are not intended to limit the present invention. In addition, the technical terms used in this specification should be interpreted in the meaning generally understood by those of ordinary skill in the art to which the present invention belongs, unless otherwise defined in this specification, and excessively inclusive. It should not be construed in the meaning of a human being or in an excessively reduced meaning. In addition, when the technical terms used in the present specification are incorrect technical terms that do not accurately express the spirit of the present invention, they should be understood by being replaced with technical terms that can be correctly understood by those skilled in the art. In addition, general terms used in the present invention should be interpreted as defined in advance or according to the context before and after, and should not be interpreted in an excessively reduced meaning.

Also, as used herein, the singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as "consisting of" or "comprising" should not be construed as necessarily including all of the various components or various steps described in the specification, some of which components or some steps are It should be construed that it may not include, or may further include additional components or steps.

In addition, the suffixes "module" and "part" for the components used in this specification are given or mixed in consideration of the ease of writing the specification, and do not have distinct meanings or roles by themselves.

Also, terms including an ordinal number such as first, second, etc. used herein may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

In an embodiment of the present invention, SOC is an abbreviation for State of Charge, which indicates the state of charge of the battery, and can be expressed by Equation 1 below.

[Equation 1]

Here, Q _remaining represents the remaining capacity of the battery, and Q _max represents the capacity of the battery in a fully charged state.

Meanwhile, in the embodiment of the present invention, SOC _init and SOC _cctocv indicate a charging state at a specific point in the charging process. That is, SOC _init indicates the remaining capacity at the start of charging, and SOC _cctocv indicates the remaining capacity at the time when the maximum voltage is reached during charging.

Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the accompanying drawings, but the same or similar components are assigned the same reference numerals regardless of reference numerals, and redundant description thereof will be omitted.

In addition, in the description of the present invention, if it is determined that a detailed description of a related known technology may obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, it should be noted that the accompanying drawings are only for easy understanding of the spirit of the present invention, and should not be construed as limiting the spirit of the present invention by the accompanying drawings.

1 is a block diagram showing a schematic configuration of a battery capacity estimation system according to an embodiment of the present invention.

Referring to FIG. 1 , the battery capacity estimation system according to the present invention may include a battery charging device 110 , an AP 120 , and a data server 130 . Here, the battery charging device 110 and the data server 130 may be implemented as one device.

The battery charging device 110 is a device for charging a battery using a constant current (hereinafter, referred to as 'CC').

The battery charging device 110 according to the present invention does not separately discharge the battery, and charges the battery regardless of the current SOC of the battery. _{In addition, while charging the battery, the battery charging device 110 collects data such as the charging voltage (V cc} ) and the remaining capacity of the battery from an arbitrary charging start time to a time when the maximum voltage is reached during charging, and collects the collected data. The data is transmitted to the battery capacity estimating device, that is, the data server 130 via the AP 120 . In this case, the battery charging device 110 may divide the voltage curve of the charging section into N equal intervals (here, N is a natural number) and collect _{V cc for each section.}

The data server 130 estimates the capacity of the battery by analyzing the remaining capacity of the battery collected by the battery charging device 110 through artificial intelligence (AI).

Meanwhile, the capacity of the battery may be expressed by Equation 2 below.

[Equation 2]

Here, I _cc represents the charging current and t _cc represents the charging time.

That is, the data server 130 according to the present invention estimates the capacity of the battery by comparing _{I cc} × t _cc and SOC _cctocv - SOC _{init .} To this end, the data server 130 may be implemented as a neural network composed of one or more computers, and may estimate the capacity of the battery by inputting the collected data into the neural network.

Specifically, the data server 130 according to the present invention inputs and learns the remaining capacity of the battery collected from the start of any charging to the time when the maximum voltage is reached during charging into the neural network to learn the battery until the low SOC section just before charging. The capacity of the battery can be estimated without the need to discharge the battery.

On the other hand, the non-linear characteristics of the lithium-ion battery are expressed in the voltage curve, and it is possible to replace the _{SOC cctocv.} That is, FIG. 2(a) shows a voltage curve, and FIG. 2(b) shows a dV/dQ curve. As shown in FIG. 2 , it can be seen that nonlinear features 210 according to SOC are observed in the voltage curve and the dV/dQ curve. Reference numeral 220 of the dV/dQ curve denotes distortion generated during the filtering process, and is not a nonlinear feature point.

Therefore, the data server 130 according to the present invention _{can replace SOC cctocv} and use t _cc , V _cc , and SOC _init as key points for learning. That is, the data server 130 may estimate the capacity of the battery by inputting _{t cc} , V _cc , and SOC _{init to the neural network and learning.} In addition, the data server 130, when the battery charging device 110 divides the voltage curve of the charging section into N equal intervals, for each section collected by the battery charging device 110 together with _{t cc} and SOC _{init .} By inputting V _cc to the neural network and learning it, the capacity of the battery can be estimated.

To this end, the data server 130 may estimate the capacity of the battery through machine learning using a multi-layer perceptron algorithm, which is a kind of artificial neural network.

An artificial neural network is an algorithm that models the regularity between input data and output data through learning using a given training dataset as an input. The artificial neural network that has completed the learning process may predict the output data with respect to the untrained input data through regularity modeled through the training data set.

The algorithm proposed in this embodiment is an algorithm for estimating the capacity of a battery with input data measured during charging in a real situation by using an artificial neural network trained with a learning data set constructed through a prior experiment.

With the development of deep learning technology, it has become possible to model complex regularities that are not simply expressed from raw inputs. Even in the corresponding algorithm of this embodiment, it became possible to model even the regularity that is not easily seen in the voltage curve. Using this, the data server 130 according to the present invention can find regularity through learning only with information present _{in V cc} , t _cc , and SOC _int , and estimate the _{battery capacity (C max ).}

The data server 130 according to an embodiment of the present invention divides the charging period into N (eg, 100) equal parts _{to input V cc} to find regularity by sufficiently reflecting information present in the voltage curve. Of course, as long as the information present in the voltage curve can be expressed well, it _{does not matter if V cc} is less or more than 100 equal parts.

As such, the data server 130 according to the present invention replaces _{SOC cctocv and uses t cc} , V _cc , and SOC _init as key points for learning. _Therefore, it is possible to exclude the change of the SOC cctocv point.

In addition, the data server 130 considers changes in characteristics such as the shape of the charging voltage or the charging voltage curve according to the series/parallel connection type of the battery pack or the internal chemical composition of individual batteries, and sets a weight preset for each type of battery in the neural network. may be given.

[Explanation of code]

232: storage unit 234: battery data collection unit

236: battery capacity estimation unit

Hereinafter, the battery capacity estimation apparatus 130 according to an embodiment of the present invention will be described in detail.

Referring to FIG. 3 , hardware capable of realizing the function of the battery capacity estimation apparatus 130 will be described. 3 is a block diagram showing an example of hardware capable of realizing the function of the battery capacity estimation device according to the embodiment of the present invention.

The function of the battery capacity estimating device 130 can be realized using, for example, the hardware resources shown in FIG. 3 . That is, the function of the battery capacity estimation device 130 is realized by controlling the hardware shown in FIG. 3 using a computer program.

As shown in FIG. 3 , this hardware mainly includes a CPU 302 , a Read Only Memory (ROM) 304 , a RAM 306 , a host bus 308 , and a bridge 310 . In addition, the hardware includes an external bus 312 , an interface 314 , an input unit 316 , an output unit 318 , a storage unit 320 , a drive 322 , a connection port 324 , and a communication unit 326 . has

The CPU 302 functions, for example, as an arithmetic processing unit or a control unit, based on various programs recorded in the ROM 304 , the RAM 306 , the storage unit 320 , or the removable recording medium 328 . Controls all or part of the operation of each component. The ROM 304 is an example of a storage device that stores a program read by the CPU 302, data used for calculation, and the like. In the RAM 306, for example, a program read by the CPU 302, various parameters that change when the program is executed, and the like are temporarily or permanently stored.

These elements are connected to each other via, for example, a host bus 308 capable of high-speed data transfer. On the other hand, the host bus 308 is connected to an external bus 312 having a relatively low data transfer rate, for example, via a bridge 310 . As the input unit 316, for example, a mouse, a keyboard, a touch panel, a touch pad, a button, a switch, a lever, and the like are used. Further, as the input unit 316, a remote controller capable of transmitting a control signal using infrared or other radio waves may be used.

As the output unit 318 , for example, a display device such as a cathode ray tube (CRT), a liquid crystal display (LCD), a plasma display panel (PDP), or an electro-luminescence display (ELD) may be used. In addition, as the output unit 318, an audio output device such as a speaker or headphones, or a printer may be used.

The storage unit 320 is a device for storing various data. As the storage unit 320, for example, a magnetic storage device such as an HDD is used. Further, as the storage unit 320, a semiconductor storage device such as an SSD (Solid State Drive) or a RAM disk, an optical storage device, a magneto-optical storage device, or the like may be used.

The drive 322 is a device that reads information recorded on the removable recording medium 328 which is a removable recording medium or writes information into the removable recording medium 328 . As the removable recording medium 328, for example, a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is used. Also, in the removable recording medium 328 , a program for regulating the operation of the battery capacity estimating apparatus 130 may be stored.

Connection port 324 is, for example, a USB (Universal Serial Bus) port, IEEE 1394 port, SCSI (Small Computer System Interface), RS-232C port, or an optical audio terminal, such as an external connection device 330 for connecting it's a port As the externally connected device 330, for example, a printer or the like is used.

The communication unit 326 is a communication device for connecting to the network 332 . As the communication unit 326, for example, a communication circuit for wired or wireless LAN, a communication circuit for WUSB (Wireless USB), a communication circuit for a cellular phone network, or the like can be used. The network 332 is, for example, a network connected by wire or wireless.

The hardware of the battery capacity estimation apparatus 130 has been described above. In addition, the above-mentioned hardware is an example, and a deformation|transformation which omits some elements, a deformation|transformation which adds new elements, etc. are possible.

Next, the function of the battery capacity estimation apparatus 130 will be described with reference to FIG. 4 . 4 is a block diagram showing an example of a function of the battery capacity estimation device according to the embodiment of the present invention.

As shown in FIG. 4 , the battery capacity estimation device 130 may include a storage unit 232 , a battery data collection unit 234 , and a battery capacity estimation unit 236 .

In addition, the function of the memory|storage part 232 is implement|achieved using the RAM 306, the memory|storage part 320, etc. which were mentioned above. The function of the battery data collection unit 234 can be realized by using the above-described communication unit 326 or the like. The function of the battery capacity estimation unit 236 can be realized using the above-described CPU 302 or the like.

The storage unit 232 stores data including the charging voltage and the remaining capacity of the battery.

The battery data collection unit 234 collects data such as the charge voltage and the remaining capacity of the battery from an arbitrary charging start time to the time when the maximum voltage is reached during charging while charging the battery and stores the data in the storage unit 232 . . In this case, the battery data collection unit 234 may divide the voltage curve of the charging section into N equal parts (N is a natural number) at equal intervals to collect _{V cc for each section.}

The battery capacity estimating unit 236 estimates the battery capacity by inputting and learning the remaining capacity of the battery collected from an arbitrary charging start time to a time when the maximum voltage is reached during charging into the neural network.

In addition, the battery capacity estimator 236 may _{replace SOC cctocv} and use t _cc , V _cc , and SOC _init as key points for learning. That is, the battery capacity estimator 236 may estimate the capacity of the battery by inputting _{t cc} , V _cc , and SOC _{init to the neural network and learning.} At this time, the battery capacity estimating unit 236 is collected by the battery data collection unit 234 together with _{t cc} and SOC _init when the battery data collection unit 234 divides the voltage curve of the charging section into N equal intervals. The capacity of the battery can be estimated by inputting _{V cc} for each section to the neural network and learning it.

Additionally, the battery capacity estimator 236 may give the neural network a preset weight for each battery type.

5 is a flowchart illustrating a method for estimating battery capacity based on a neural network according to the first embodiment of the present invention.

Referring to FIG. 5 , while charging the battery, the battery capacity estimating apparatus 130 collects the remaining capacity of the battery from an arbitrary charging start time to a time when the maximum voltage is reached during charging ( S510 ).

The battery capacity estimating apparatus 130 estimates the battery capacity by inputting the collected residual capacity of the battery from the start of any charging to the time when the maximum voltage is reached during charging into the neural network and learning ( S520 ). In this case, the battery capacity estimating apparatus 130 may give the neural network a preset weight for each battery type.

6 is a flowchart illustrating a method for estimating battery capacity based on a neural network according to a second embodiment of the present invention.

Referring to FIG. 6 , while charging the battery, the battery capacity estimating device 130 _{collects a charging voltage (V cc} ) from an arbitrary charging start time to a time point reaching the maximum voltage during charging ( S610 ). In this case, the battery capacity estimating apparatus 130 may divide the voltage curve of the charging section into N equal parts (N is a natural number) at equal intervals to collect _{V cc for each section.}

The battery capacity estimating apparatus 130 estimates _{the capacity of the battery by inputting t cc} , V _cc , and SOC _init to the neural network to learn ( S620 ). At this time, when the voltage curve of the charging section is divided into N equal intervals, the battery capacity estimating device 130 _{inputs V cc} for each section collected together with _{t cc} and SOC _init to the neural network and learns, thereby the capacity of the battery can be estimated. Also, the battery capacity estimating apparatus 130 may give the neural network a preset weight for each battery type.

The above-described method may be implemented through various means. For example, embodiments of the present invention may be implemented by hardware, firmware, software, or a combination thereof.

In the case of implementation by hardware, the method according to embodiments of the present invention may include one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), and Programmable Logic Devices (PLDs). , FPGAs (Field Programmable Gate Arrays), processors, controllers, microcontrollers and microprocessors, and the like.

In the case of implementation by firmware or software, the method according to the embodiments of the present invention may be implemented in the form of a module, procedure, or function that performs the functions or operations described above. The software code may be stored in the memory unit and driven by the processor. The memory unit may be located inside or outside the processor, and may transmit and receive data to and from the processor by various known means.

Meanwhile, the artificial neural network may be configured by stacking several layers, such as a fully connected layer, a convolution layer, and a dense layer. At this time, each layer has several nodes therein, and is divided according to an input/output configuration method between nodes (inputs) of the previous layer and nodes (outputs) in each layer.

The fully connected layer mainly used in this embodiment may be represented by the structure shown in FIG. 7 .

Referring to FIG. 7 , reference numeral 710 denotes a node, reference numeral 720 denotes a previous layer, and reference numeral 730 denotes a current layer. That is, as shown in FIG. 7 , a layer in which all possible connections exist between nodes of the previous layer 720 and nodes of the current layer 730 is referred to as a fully connected layer.

Meanwhile, in the embodiment of the present invention, the artificial neural network is limited to being implemented through multiple perceptrons or fully connected layers, but is not limited thereto, and may be implemented through structures other than multiple perceptrons or fully connected layers.

8 is a diagram for explaining the structure and implementation of a multi-perceptron.

Referring to FIG. 8 , in the multi-perceptron, one node 810 calculates the nodes and weights of the previous layer as a weighted sum and then applies a kind of threshold value. Outputs the result value as 1, otherwise 0. Here, the process (or function) for applying the threshold value is called an activation function. According to scientists, a neuron does not immediately emit an output when receiving an input, but emits an output only when the input is accumulated and grows to a certain level. This function is called an activation function. Also, a bias b indicating a direct bias may be added to the activation function.

One node 810 may be expressed by Equation 3 below.

[Equation 3]

Also, one layer 820 may include n nodes of an input layer and m nodes of a hidden layer.

That is, one layer 820 may be expressed by Equation 4 below.

[Equation 4]

Consequently, multiple perceptrons may be composed of an input layer, a plurality of hidden layers, and an output layer.

The embodiments disclosed herein have been described above with reference to the accompanying drawings. As such, the embodiments shown in each drawing should not be construed as being limited, and may be combined with each other by those skilled in the art having read the contents of the present specification, and when combined, it may be construed that some components may be omitted.

Here, the terms or words used in the present specification and claims should not be construed as being limited to conventional or dictionary meanings, but should be interpreted as meanings and concepts consistent with the technical idea disclosed in the present specification.

Therefore, the embodiments described in the present specification and the configurations shown in the drawings are only the embodiments disclosed in the present specification, and do not represent all the technical ideas disclosed in the present specification, so various equivalents that can replace them at the time of the present application It should be understood that there may be water and variations.

When the learning method of the present invention is applied, the remaining capacity of the battery can be estimated only from data from an arbitrary charging start time to the charging end. Specifically, when the learning method of the present invention is utilized, the remaining battery capacity can be estimated at a faster rate than in the prior art.

In addition, the computing device for estimating the battery capacity based on constant current charging using the learning method of the present invention can be utilized in portable electronic products such as laptops, video cameras, and portable phones, thereby contributing to battery life management and prevention of accidents due to battery short. have.

Claims (10)

1. In a method for learning a neural network in a computing device for estimating battery capacity based on constant current charging,

   collecting the remaining capacity of the battery from the start of any charging to the time when the maximum voltage is reached during charging; and

   learning the neural network by inputting the collected residual capacity into the neural network;

   A neural network learning method comprising a.
2. According to claim 1,

   The neural network learning method, characterized in that implemented through a multi-layer perceptron (Multi-layer Perceptron) or a fully connected layer (Fully Connected Layer).
3. In a method for learning a neural network in a computing device for estimating battery capacity based on constant current charging,

   collecting the charging voltage from the start of any charging to the time when the maximum voltage is reached during charging; and

   learning the neural network by inputting a charging time, a collected charging voltage, and a battery remaining capacity at the arbitrary charging start time to the neural network;

   A neural network learning method comprising a.
4. 4. The method of claim 3, wherein in the collecting the charging voltage,

   A neural network learning method, characterized in that by dividing the voltage curve of the charging section into equal intervals by N (N is a natural number), and collecting the charging voltage for each section.
5. 4. The method of claim 3,

   assigning a preset weight to the neural network for each battery type;

   Neural network learning method, characterized in that it further comprises.
6. 4. The method of claim 3,

   The neural network learning method, characterized in that implemented through a multi-layer perceptron (Multi-layer Perceptron) or a fully connected layer (Fully Connected Layer).
7. a memory storing one or more instructions; and

   By executing the stored one or more instructions,

   A neural network that collects the charging voltage from the start of any charging to the time when the maximum voltage is reached during charging, and estimates the capacity of the battery by receiving the charging time, the collected charging voltage, and the remaining capacity of the battery at any charging start time. One or more processors constituting the;

   A neural network learning device comprising a.
8. 8. The method of claim 7,

   The one or more processors divide the voltage curve of the charging section into N equal parts (N is a natural number) at equal intervals, and collect the charging voltage for each section.
9. 8. The method of claim 7,

   The one or more processors are configured to assign a preset weight to the neural network for each battery type.
10. 8. The method of claim 7,

    The neural network learning apparatus, characterized in that implemented through a multi-layer perceptron (Multi-layer Perceptron) or a fully connected layer (Fully Connected Layer).

Images