Abnormality detection method for secondary battery

Abstract

To provide a secondary battery charge state estimation method with high estimation accuracy, even if the secondary battery deteriorates. Also, to provide a secondary battery capacity measurement system that can estimate the SOC with high accuracy in a short time and at low cost. If the capacity of the secondary battery can be estimated with high accuracy, it will also be possible to perform abnormality detection based on that value. To provide a new abnormality detection method for secondary batteries.SOLUTION: In the method of charging CCCV, the CC time, and CV time are used as learning parameters to construct a learning model. By using this learning model, it is possible to obtain highly accurate estimated capacity values by using the minimum number of input data, which is either the two CC time and CV time values or the three values of the charging start voltage.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] One embodiment of the present invention relates to an object, a method, or a manufacturing method. Alternatively, the present invention relates to a process, a machine, a manufacture, or a composition of matter. One embodiment of the present invention relates to a semiconductor device, a display device, a light-emitting device, a power storage device, a lighting device, an electronic device, or a manufacturing method thereof. Another embodiment of the present invention relates to a method for estimating a state of charge of a power storage device, a system for estimating a state of charge of a power storage device, and a method for detecting an abnormality. In particular, the present invention relates to a system for estimating a state of charge of a power storage device and a system for detecting an abnormality of a power storage device.

[0002] In this specification, the term "electricity storage device" refers to elements and devices in general that have an electricity storage function, including, for example, storage batteries (also called secondary batteries) such as lithium ion secondary batteries, lithium ion capacitors, nickel hydride batteries, all-solid-state batteries, and electric double layer capacitors.

[0003] Another embodiment of the present invention relates to a neural network and a state-of-charge estimation system for a power storage device using the neural network. Another embodiment of the present invention relates to a vehicle using the neural network. Another embodiment of the present invention relates to an electronic device using the neural network. Another embodiment of the present invention relates to a state-of-charge estimation system that is not limited to vehicles and can be applied to a power storage device for storing power obtained from power generation equipment such as a solar power generation panel installed on a structure or the like. [Background technology]

[0004] Methods for estimating the remaining charge of a secondary battery include the coulomb counter method and the OCV (Open Circuit Voltage) method.

[0005] In conventional methods, there is a risk that errors will accumulate and the accuracy of estimating the charging rate, i.e., the State of Charge (SOC), will decrease significantly when the battery is operated for a long period of time and repeatedly charged and discharged. In addition, the initial SOC(0) will change due to self-discharge as time passes when the battery is not in use, making it difficult to improve the accuracy of estimating the SOC. The Coulomb counter method has disadvantages such as being unable to correct errors in the initial SOC(0) and accumulating errors in the current sensor. Patent Document 1 discloses a technology that estimates the state of a secondary battery at low temperatures with high accuracy using a state estimation means based on information that associates parameters with temperature. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2016-80693 A Summary of the Invention [Problem to be solved by the invention]

[0007] During manufacturing, minor differences in the amount of active material, electrode size, and other factors during assembly can cause slight individual differences in secondary batteries, even when they are from the same production lot. Since multiple secondary batteries are used in vehicles, combining many batteries can result in individual differences that can affect the capacity of the batteries, and deterioration can lead to large differences in capacity between vehicles. Even batteries from the same lot can deteriorate to different degrees due to factors such as usage conditions (ambient temperature, frequency of charging and discharging, and storage conditions).

[0008] In addition, as the secondary battery deteriorates, the accuracy of the SOC estimation may decrease significantly. For example, when using an estimation method based on the current integral value for a long period of time, current detection errors accumulate, causing the accuracy of the SOC estimation to gradually decrease. The SOC is defined as the ratio of the remaining capacity to the maximum capacity of the secondary battery. The maximum capacity of a secondary battery can be calculated by discharging it after full charging and then integrating the current over time, but this may take a long time to fully discharge. In addition, the secondary battery must be recharged before it can be used again.

[0009] A method for estimating the state of charge of a secondary battery with high estimation accuracy even if the secondary battery has deteriorated, and a capacity measurement system for a secondary battery that estimates the SOC with high accuracy in a short time and at low cost are provided.

[0010] Furthermore, if the capacity of a secondary battery can be estimated with high accuracy, it would be possible to detect an abnormality based on that value. Another object of the present invention is to provide a new method for detecting an abnormality in a secondary battery. [Means for solving the problem]

[0011] When estimating the state of charge of a secondary battery, various parameter information of the secondary battery can be used. For example, the parameter information of the secondary battery includes the internal resistance, current value, voltage value, ambient temperature, internal temperature of the secondary battery, capacity value in a fully charged state, charging conditions, and discharging conditions of the secondary battery. The more types of data used, the higher the accuracy of the estimation. Rather, the use of many types of data may result in a large amount of noise, which may reduce the accuracy of the estimation. In addition, the use of many types of data may result in many calculations, which may take a long time to output a solution, or the solution may not converge and the calculation may not be completed.

[0012] The method for estimating the state of charge of a secondary battery disclosed in this specification finds, from many types of data, several parameters that directly or indirectly affect the deterioration of the secondary battery, and has a neural network learning device learn the fewer types of parameters as teacher data, so that the learning result of the neural network becomes the capacity of the secondary battery.

[0013] Furthermore, in a neural network learning device, increasing the number of parameters or data does not necessarily result in higher accuracy; a large amount of data can cause overlearning, resulting in reduced estimation accuracy.

[0014] It is important to select a small number of learning parameters from a large number of parameters, determine the training data, train a neural network learning device, and calculate the capacity of a secondary battery with high accuracy.

[0015] The present inventors conducted charge and discharge cycles using the CCCV charging method and measured the deterioration of the secondary battery, and found that the CV charging period (also called CV time) became longer with the deterioration of the secondary battery. The CCCV charging method is generally used for charging lithium-ion secondary batteries. CCCV charging is a charging method in which charging is first performed to a predetermined voltage by CC charging, and then charging is performed until the current flowing in CV charging decreases, specifically until the current value reaches the end current value. One charging period is divided into a CC charging period (also called CC time) and a subsequent CV charging period (CV time). During the CC charging period, a constant current is passed through the secondary battery until a predetermined voltage is reached, and during the CV charging period, charging is performed at a constant voltage until the end current value is reached.

[0016] In the CCCV charging method, the CC time and the CV time are used as learning parameters to construct a learning model. The construction of such a learning model is referred to as the learning stage (learning phase).

[0017] The learning parameters used in the learning model are not only the CC time and CV time data, but also various data actually obtained in charge-discharge cycle tests of a reference secondary battery.

[0018] Using this learning model, an estimated capacity value can be obtained using three items of minimum input data: CC time, CV time, and charge start voltage value. Also, obtaining an estimated capacity value using the learning results using the learning model refers to the judgment stage (judgment phase). Although both the learning stage and the judgment stage may be implemented in a vehicle, etc., the driver can obtain an estimated capacity value by obtaining the learning results in advance and installing at least the judgment stage in the vehicle. Also, when data during driving is used as a learning parameter, the driver can obtain a more accurate estimated capacity value while driving by implementing both the learning stage and the judgment stage in the vehicle.

[0019] The method for estimating the capacity of a secondary battery disclosed in this specification measures the charging start voltage value of the secondary battery, measures a first time (CC time) from the start of charging to when the terminal voltage of the secondary battery reaches a reference voltage, and measures a second time (CV time) from when the reference voltage is reached to the end of charging. A neural network unit receives the charging start voltage value, the first time, and the second time as input and calculates the capacity of the secondary battery.

[0020] In addition to the three values, if four data items are input, including the voltage value after the rest period after the end of charging and the third period until the chemical reaction inside the secondary battery stabilizes, the number of input data increases, but the highest accuracy can be achieved. For the third period, a cycle test is performed in advance on a reference secondary battery, and the time it takes to stabilize after resting after the end of charging is measured.

[0021] Another capacity estimation method for a secondary battery disclosed in this specification measures the charging start voltage value of the secondary battery, measures a first time (CC time) from the start of charging to when the terminal voltage of the secondary battery reaches a reference voltage, measures a second time (CV time) from when the reference voltage is reached to when charging is completed, and measures the voltage value after a third time from when charging is completed to when the chemical reaction inside the secondary battery stabilizes. A neural network unit to which the charging start voltage value, the first time (CC time), the second time (CV time), and the voltage value are input calculates the charging state of the secondary battery, specifically, the capacity of the secondary battery.

[0022] In addition, when a small amount of data is used, a first time (CC time) from when charging of the secondary battery starts until the terminal voltage of the secondary battery reaches a reference voltage is measured, and a second time (CV time) from when the reference voltage is reached until charging ends is measured, and the neural network unit to which the two pieces of data, the first time and the second time, are input calculates the state of charge of the secondary battery, specifically the capacity of the secondary battery. The calculation of the capacity of the secondary battery can be performed appropriately after charging of the secondary battery ends or while the secondary battery is discharging (specifically while the vehicle is running).

[0023] CC charging and CV charging will be explained below.

[0024] First, we will explain CC charging as one of the charging methods. CC charging is a charging method in which a constant current flows through the secondary battery throughout the entire charging period, and charging is stopped when a specified voltage is reached. The secondary battery is assumed to be an equivalent circuit with an internal resistance R and a secondary battery capacity C, as shown in Figure 6A. In this case, the secondary battery voltage V B is the voltage V across the internal resistance R R and the voltage V applied to the secondary battery capacity C C It is the sum of.

[0025] During CC charging, as shown in FIG. 6A, the switch is turned on and a constant current I flows through the secondary battery. During this time, the current I is constant, so V R According to Ohm's law, the voltage V across the internal resistance R is R On the other hand, the voltage V applied to the secondary battery capacity C is constant. C increases over time. Therefore, the secondary battery voltage V B increases over time.

[0026] and the secondary battery voltage V B When the voltage V reaches a certain value, for example 4.3 V, charging is stopped. When CC charging is stopped, the switch is turned off and the current I becomes 0, as shown in Figure 6B. Therefore, the voltage V across the internal resistance R R Therefore, the secondary battery voltage V B decreases.

[0027] The secondary battery voltage V during CC charging and after CC charging is stopped B An example of the charge current is shown in Fig. 6C. The secondary battery voltage V B However, it is shown that the value decreases slightly after CC charging is stopped.

[0028] Next, we will explain CCCV charging, which is a charging method different from the above. CCCV charging is a charging method in which charging is first performed up to a specified voltage by CC charging, and then charging is performed until the current flowing by CV charging decreases, specifically until the current value reaches the end current value.

[0029] During CC charging, as shown in FIG. 7A, the constant current power supply switch is on and the constant voltage power supply switch is off, and a constant current I flows through the secondary battery. During this time, the current I is constant, so V R According to Ohm's law, the voltage V across the internal resistance R is R On the other hand, the voltage V applied to the secondary battery capacity C is constant. C increases over time. Therefore, the secondary battery voltage V B increases over time.

[0030] and the secondary battery voltage V B When the secondary battery voltage V reaches a predetermined voltage, for example, 4.3 V, the charging mode is switched from CC charging to CV charging. During CV charging, as shown in FIG. 7B, the constant voltage power supply switch is turned on and the constant current power supply switch is turned off, and the secondary battery voltage V B On the other hand, the voltage V applied to the secondary battery capacity C is constant. C V increases over time. B =V R +V C Therefore, the voltage V across the internal resistance R R The voltage V across the internal resistance R decreases over time. R As becomes smaller, V R According to Ohm's law, the current I flowing through the secondary battery also becomes smaller.

[0031] When the current I flowing through the secondary battery becomes a predetermined current, for example, a current equivalent to 0.01 C, charging is stopped. When CCCV charging is stopped, as shown in FIG. 7C, all switches are turned off and the current I becomes 0. Therefore, the voltage V across the internal resistance R R becomes 0V. However, the voltage V applied to the internal resistance R due to CV charging R is small enough that even if there is no voltage drop across the internal resistance R, the secondary battery voltage V B hardly descends at all.

[0032] The secondary battery voltage V during CCCV charging and after CCCV charging is stopped B An example of the charging current is shown in Figure 8A. Even if CCCV charging is stopped, the secondary battery voltage V B It is shown that the aircraft barely descends at all.

[0033] Next, we will explain CC discharge, which is one of the discharging methods. CC discharge is a method in which a constant current flows from the secondary battery throughout the entire discharging period, and the secondary battery voltage V B This is a discharge method in which discharging is stopped when the voltage reaches a predetermined voltage, for example 2.5V.

[0034] The secondary battery voltage V during CC discharge B An example of the discharge current is shown in FIG. 8B. As the discharge progresses, the secondary battery voltage V B The figure shows how the light descends.

[0035] Next, we will explain the discharge rate and charge rate. The discharge rate is the relative ratio of the current during discharge to the battery capacity, and is expressed in units of C. For a battery with a rated capacity of X (Ah), the current equivalent to 1C is X (A). When a battery is discharged at a current of 2X (A), it is said to have been discharged at 2C, and when a battery is discharged at a current of X / 5 (A), it is said to have been discharged at 0.2C. The same is true for the charge rate; when a battery is charged at a current of 2X (A), it is said to have been charged at 2C, and when a battery is charged at a current of X / 5 (A), it is said to have been charged at 0.2C.

[0036] The method for estimating the state of charge of a secondary battery disclosed in this specification is basically a method for estimating the degree of deterioration of a secondary battery after charging is completed, not during actual use. For example, the capacity of a secondary battery of an electric vehicle can be estimated with high accuracy when the battery is fully charged. In this case, neural network processing is performed in a charge control device for charging the electric vehicle or in a server capable of exchanging data with the charge control device. When performing neural network processing, hardware is required that has sufficient memory to store learning data and is capable of sufficient arithmetic processing.

[0037] In addition, the software program that executes the inference program for performing neural network processing can be written in various programming languages ​​such as Python, Go, Perl, Ruby, Prolog, Visual Basic, C, C++, Swift, Java (registered trademark), and .NET. In addition, the application may be created using a framework such as Chainer (available in Python), Caffe (available in Python and C++), and TensorFlow (available in C, C++, and Python). For example, the LSTM algorithm is programmed in Python and uses a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). In addition, a chip that integrates a CPU and a GPU into one is sometimes called an APU (Accelerated Processing Unit), and this APU chip can also be used. In addition, an IC (also called an inference chip) incorporating an AI (system) may be used. An IC incorporating an AI system may also be called a circuit (microprocessor) that performs neural network calculations. Effect of the Invention

[0038] The method for estimating the state of charge of a secondary battery disclosed in this specification can estimate the capacity with high accuracy using a small number of types of data, and therefore can use a small amount of learning data and simplify the calculation process.

[0039] Since the hardware capable of executing neural network processing can be miniaturized, it can be built into a small charging control device. By using a mobile information terminal equipped with hardware capable of executing neural network processing, it is also possible to estimate the capacity of an electric vehicle based on the charging information of the electric vehicle.

[0040] Furthermore, the miniaturized hardware can be installed in electric vehicles, which allows for highly accurate capacity estimation after charging at charging spots at the vehicle's destination. [Brief description of the drawings]

[0041] [Figure 1] FIG. 1A is a graph showing estimation accuracy by a method according to one embodiment of the present invention, FIG. 1B is a table showing types of input data, and FIG. 1C is a table corresponding to FIG. 1A. [Diagram 2] FIG. 2A is a graph showing the estimation accuracy according to a method according to one embodiment of the present invention, and FIGS. 2B and 2C are tables showing the types of input data. [Diagram 3] FIG. 3 is a flow diagram showing one embodiment of the present invention. [Figure 4] FIG. 4 shows data showing the rest time and voltage change after charging of a secondary battery. [Diagram 5] 5A and 5B are diagrams showing an example of the configuration of neural network processing. [Figure 6] 6A, 6B, and 6C are diagrams for explaining a method of charging a secondary battery. [Figure 7] 7A, 7B, and 7C are diagrams for explaining a method of charging a secondary battery. [Figure 8] 8A and 8B show a charge curve and a discharge curve of a secondary battery. [Figure 9] 9A and 9B are diagrams illustrating a coin-type secondary battery. [Figure 10]10A is a perspective view, FIG. 10B is a cross-sectional perspective view, FIG. 10C is a perspective view, and FIG. 10D is a top view, all of which are diagrams for explaining a cylindrical secondary battery. [Figure 11] 11A, 11B, and 11C are perspective views illustrating examples of a secondary battery. [Figure 12] 12A, 12B, 12C, 12D, and 12E are diagrams illustrating examples of small-sized electronic devices and vehicles including a secondary battery module of one embodiment of the present invention. [Figure 13] 13A, 13B, and 13C are diagrams illustrating examples of a vehicle and a house including a secondary battery module of one embodiment of the present invention. [Figure 14] FIG. 14 is an explanatory diagram illustrating a program and an information processing method according to one embodiment of the present invention. [Figure 15] FIG. 15 is an explanatory diagram illustrating a program and an information processing method according to one embodiment of the present invention. [Figure 16] FIG. 16 is an explanatory diagram illustrating a program and an information processing method according to one embodiment of the present invention. [Figure 17] FIG. 17 is an explanatory diagram illustrating a program and an information processing method according to one embodiment of the present invention. [Figure 18] FIG. 18 is an explanatory diagram illustrating a program and an information processing method according to one aspect of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0042] Hereinafter, the embodiments of the present invention will be described in detail with reference to the drawings. However, the present invention is not limited to the following description, and it will be easily understood by those skilled in the art that the form and details of the present invention can be modified in various ways. Furthermore, the present invention is not to be interpreted as being limited to the description of the embodiments shown below.

[0043] (Embodiment 1) In this embodiment, a cycle test is performed on a reference secondary battery, a learning model is constructed based on the data, and a procedure for estimating capacity and a procedure for detecting anomalies using the model are shown in FIG.

[0044] First, a charge-discharge cycle test is performed on the reference secondary battery (S1).

[0045] Data obtained in the charge / discharge cycle test is collected. (S2) In this data collection, various data are collected. For example, CC time, CV time, temperature, discharge voltage, initial FCC (mAh), number of cycles, voltage at the start of charging, voltage 1 second after the start of charging, voltage 2 seconds after the start of charging, voltage 60 seconds after the start of charging, voltage 120 seconds after the start of charging, voltage immediately after the end of charging, voltage 1 second after pause after the end of charging, voltage 2 seconds after pause after the end of charging, voltage 10 seconds after pause after the end of charging, voltage 120 seconds after pause after the end of charging, voltage 600 seconds after pause after the end of charging, etc. are actually measured. These data (excluding the number of cycles) can be obtained in one charge / discharge. In addition, data can be obtained from the second charge / discharge and thereafter. In addition, multiple secondary batteries with almost the same characteristics may be used as the reference secondary battery.

[0046] At least three pieces of data, CC time, CV time, and charge start voltage, are collected. In this embodiment, a cycle test is performed using multiple commercially available lithium-ion secondary batteries (NCR18650B) to acquire data. The nominal capacity of the lithium-ion secondary batteries is 3350mAh, and the average voltage is 3.6V. The cycle test involves charging at 4.2V, 0.5C (CV cutoff 0.02C), discharging to an arbitrary voltage after a 10-minute rest period, and then repeating this operation of resting for 10 minutes.

[0047] Data on the rest time after the end of charging is also collected in advance. For this data, the data is taken with the time of rest (left alone) after the lithium ion secondary battery is fully charged as the horizontal axis and the voltage as the vertical axis, and the time when the voltage change becomes small is selected. Fig. 4 shows actual measured values ​​in a graph of the voltage change over time when the battery is rested after the end of full charging. In Fig. 4, the voltage change becomes small from about 110 seconds to about 130 seconds after the start of the rest. The point at which the voltage change becomes small coincides with the point at which the chemical reaction inside the secondary battery becomes stable. In addition, in the secondary battery that has actually been subjected to 1000 cycle tests, the internal resistance increases due to deterioration and the voltage drop becomes large, but the voltage over time after the end of full charging shows almost the same tendency. In this embodiment, the voltage value 120 seconds (2 minutes) after the rest is used as an important parameter. In addition, since the time until the chemical reaction inside the secondary battery becomes stable differs depending on the type of secondary battery, it may be determined from the data obtained by performing a cycle test using the secondary battery for which the capacity is to be estimated.

[0048] Next, the obtained data is trained to build a learning model (S3).

[0049] In this embodiment, optimal weights and biases are set for each node connecting neurons, and a learning model is created. Chainer is used as the framework, and fully connected neural network processing based on the official mnist source is used. There are three intermediate layers and 200 hidden layers. Adam is used as the optimizer that performs optimization. At least three pieces of data, CC time, CV time, and charging start voltage, are used as learning data, and the dischargeable capacity is used as the correct label. All data is linearly interpolated and normalized before learning.

[0050] 5A and 5B illustrate an example of the operation of neural network processing.

[0051] As shown in FIG. 5A, the neural network processing NN can be configured with an input layer IL, an output layer OL, and an intermediate layer (hidden layer) HL. The input layer IL, the output layer OL, and the intermediate layer HL each have one or more neurons (units). The intermediate layer HL may be one layer or two or more layers. A neural network processing having two or more intermediate layers HL may be called a DNN (deep neural network), and learning using the deep neural network processing may be called deep learning.

[0052] Input data is input to each neuron in the input layer IL, an output signal of a neuron in the previous or next layer is input to each neuron in the hidden layer HL, and an output signal of a neuron in the previous layer is input to each neuron in the output layer OL. Each neuron may be connected to all neurons in the previous or next layer (full connection), or may be connected to only a portion of the neurons.

[0053] FIG. 5B shows an example of a computation by a neuron. Here, neuron N and two neurons in the previous layer that output signals to neuron N are shown. Neuron N receives the output x1 of a neuron in the previous layer and the output x2 of a neuron in the previous layer. Neuron N then calculates the sum (x1w1+x2w2) of the multiplication result (x1w1) of the output x1 and the weight w1 and the multiplication result (x2w2) of the output x2 and the weight w2, and then adds a bias b as necessary to obtain a value a=x1w1+x2w2+b. The value a is then transformed by the activation function h, and neuron N outputs an output signal y=h(a).

[0054] Thus, the computation by a neuron includes the computation of adding the products of input data and weights, that is, the sum-of-products computation. This sum-of-products computation can be performed by a sum-of-products computation circuit having a current source circuit, an offset absorption circuit, and a cell array. Moreover, the conversion of signals by the activation function h can be performed by a hierarchical output circuit. In other words, computations of the intermediate layer or the output layer can be performed by the computation circuit.

[0055] The cell array of the product-sum calculation circuit is composed of a plurality of memory cells arranged in a matrix.

[0056] The memory cell has a function of storing first data. The first data is data corresponding to the weight between neurons in neural network processing. The memory cell also has a function of multiplying the first data by second data input from outside the cell array. That is, the memory cell has a function as a memory circuit and a function as a multiplication circuit.

[0057] When the first data is analog data, the memory cell functions as an analog memory, and when the first data is multi-valued data, the memory cell functions as a multi-valued memory.

[0058] The results of the multiplications by the memory cells in the same column are then added together. This results in a multiply-and-accumulate operation of the first data and the second data. The result of the operation by the cell array is then output to the hierarchical output circuit as the third data.

[0059] The hierarchical output circuit has a function of converting the third data output from the cell array according to a predetermined activation function. The analog signal or multi-value digital signal output from the hierarchical output circuit corresponds to the output data of the intermediate layer or output layer in the neural network processing.

[0060] Examples of the activation function that can be used include a sigmoid function, a tanh function, a softmax function, a ReLU function, and a threshold function. The signal converted by the hierarchical output circuit is analog data or multi-valued digital data (data D analog )

[0061] In this way, one calculation circuit can realize the calculation of either the intermediate layer or the output layer of the neural network processing.

[0062] The analog data or multi-value digital data output from the first arithmetic circuit is supplied to the second arithmetic circuit as second data. The second arithmetic circuit then performs an arithmetic operation using the first data stored in the memory cell and the second data input from the first arithmetic circuit. This allows for the arithmetic operation of a neural network consisting of multiple layers.

[0063] To obtain the capacity of the secondary battery, the charging data is input and an estimated value is obtained using the learning model (S4).

[0064] In addition, by using a learning model with the three pieces of data, CC time, CV time, and charging start voltage, as learning data and inputting each piece of data as input 1, the average error can be set to 6.088 mAh.

[0065] In addition, by using the five pieces of data, namely CC time, CV time, charging start voltage, voltage 1 second after the end of charging, and voltage 2 seconds after the end of charging, as learning data for a learning model and inputting each of these pieces of data as input 2, the average error can be set to 6.382 mAh.

[0066] In addition, by using a learning model with four pieces of data, namely CC time, CV time, charging start voltage, and voltage 120 seconds after the end of charging, as learning data and inputting each of these pieces of data as input 3, the average error can be set to 5.844 mAh.

[0067] In addition, by using a learning model with six pieces of data, namely CC time, CV time, charging start voltage, voltage 1 second after charging is completed, voltage 2 seconds after charging is completed, and the ratio of CC time to CV time (CCCV time ratio), as learning data and inputting each piece of data as input 4, the average error can be set to 6.66 mAh.

[0068] A bar graph comparing these results is shown in FIG. 1A, a table of inputs is shown in FIG. 1B, and a summary of average errors is shown in FIG. 1C.

[0069] These results show that by using at least the three pieces of data, namely CC time, CV time, and charging start voltage, the error in the estimated capacity value can be reduced to about 7 mAh. Among these, the most accurate capacity estimation can be achieved by using a learning model that uses the four pieces of data, namely CC time, CV time, charging start voltage, and voltage 120 seconds after the end of charging, as learning data.

[0070] Steps S1 to S4 can be said to be the procedure for constructing a learning model and estimating capacity.

[0071] The learning data is only normal data. Therefore, if any abnormality occurs in the secondary battery, the estimated value changes and the estimation error increases. This can be used to detect abnormalities.

[0072] Subsequently, a charge / discharge cycle is performed in which the secondary battery is used and charged, and after charging is completed, the capacity is estimated using the learning model.

[0073] Assume that step 5 (S5) occurs during a certain charging cycle when an abnormality occurs in the secondary battery.

[0074] The estimation error after the abnormality occurs is calculated, and the large estimation error is output (S6).

[0075] If the estimation error in S6 exceeds the estimation error threshold value that can be regarded as an abnormality, it is determined that an abnormality has occurred (S7).

[0076] In order to distinguish between the occurrence of noise and the occurrence of an abnormality, a threshold value for the estimation error is determined in advance.

[0077] If an abnormality occurs, it can be detected by going through steps S5, S6, and S7.

[0078] From the above explanation, the procedure for capacity estimation is shown using the flow in Fig. 3, and the result in Fig. 1 shows that excellent and highly accurate capacity estimation can be performed. In addition, the procedure for abnormality detection is shown using the flow in Fig. 3, and it shows that abnormality detection is performed based on highly accurate capacity estimation.

[0079] The estimation error refers to the difference between the value estimated using the learning model and the dischargeable capacity, and the average error refers to the average of the estimation errors for each battery cell used. In this embodiment, the test was performed on 10 battery cells, so the average error is the sum of the estimation errors for each of the 10 cells divided by 10.

[0080] (Embodiment 2) The present embodiment will be described below in comparison with a comparative example different from the first embodiment with reference to FIG.

[0081] FIG. 2A shows the results of finding the estimation error by using the same learning model as in the first embodiment and varying the input data in various ways.

[0082] It should be noted that the input 3 shown in FIGS. 2A and 2B is the same as the input 3 shown in FIG. 1A, and the results are shown under the same conditions.

[0083] 2A and 2B, input 5 is a result using CC time and CV time, and is one of the present inventions. In input 5, the average value is 5.9, and the minimum value is 3.2 compared to input 3, and the estimation accuracy is lower than that of input 3.

[0084] Furthermore, inputs 6, 7, 8, and 9 shown in FIG. 2C are comparative examples, and the estimation errors of the comparative examples are all 10 (mAh) or more. The data for input 6 uses the charging start voltage and the voltage 120 seconds after the end of charging and pause. The data for input 7 uses the voltage 1 second after the end of charging, the voltage 2 seconds after the end of charging, and the CCCV time ratio. The data for input 8 uses the voltage 1 second after the end of charging and the voltage 2 seconds after the end of charging. The data for input 9 uses the CCCV time ratio.

[0085] The estimated capacity was calculated for each of 10 battery cells, and the average was shown as the error capacity (mAh).

[0086] From the results of Figure 2A, it is possible to output the most accurate estimated capacity compared to other conditions by using at least the CC time and CV time in the learning model, and further using the charging start voltage and the voltage 120 seconds after the end of charging and rest. If you want to reduce the number of data, it is preferable to use a learning model that uses two pieces of data, the CC time and the CV time.

[0087] (Embodiment 3) An example of a coin-type secondary battery will be described below. Fig. 9A is an external view of a coin-type (single-layer flat) secondary battery, and Fig. 9B is a cross-sectional view thereof.

[0088] In the coin-type secondary battery 300, a positive electrode can 301, which also serves as a positive electrode terminal, and a negative electrode can 302, which also serves as a negative electrode terminal, are insulated and sealed with a gasket 303 made of polypropylene or the like. The positive electrode 304 is formed of a positive electrode current collector 305 and a positive electrode active material layer 306 provided in contact with the positive electrode current collector. The negative electrode 307 is formed of a negative electrode current collector 308 and a negative electrode active material layer 309 provided in contact with the negative electrode current collector.

[0089] In addition, the positive electrode 304 and the negative electrode 307 used in the coin-type secondary battery 300 each need only have an active material layer formed on one side.

[0090] Metals such as nickel, aluminum, and titanium that are corrosion-resistant to the electrolyte, or alloys of these metals and other metals (e.g., stainless steel, etc.) can be used for the positive electrode can 301 and the negative electrode can 302. In order to prevent corrosion by the electrolyte, it is preferable to coat them with nickel, aluminum, etc. The positive electrode can 301 is electrically connected to the positive electrode 304, and the negative electrode can 302 is electrically connected to the negative electrode 307.

[0091] These negative electrode 307, positive electrode 304, and separator 310 are impregnated with an electrolyte, and as shown in FIG. 9B, positive electrode can 301 is placed at the bottom, and positive electrode can 304, separator 310, negative electrode 307, and negative electrode can 302 are stacked in this order, and positive electrode can 301 and negative electrode can 302 are crimped together via gasket 303, to produce coin-type secondary battery 300.

[0092] [Cylindrical secondary battery] Next, an example of a cylindrical secondary battery will be described with reference to Figures 10A to 10D. As shown in Figure 10A, a cylindrical secondary battery 600 has a positive electrode cap (battery lid) 601 on the top surface, and a battery can (external can) 602 on the side and bottom surfaces. The positive electrode cap and battery can (external can) 602 are insulated by a gasket (insulating packing) 610.

[0093] FIG. 10B is a schematic diagram showing a cross section of a cylindrical secondary battery. Inside a hollow cylindrical battery can 602, a battery element is provided in which a strip-shaped positive electrode 604 and a negative electrode 606 are wound with a separator 605 sandwiched therebetween. Although not shown, the battery element is wound around a center pin. One end of the battery can 602 is closed and the other end is open. For the battery can 602, metals such as nickel, aluminum, and titanium that are resistant to corrosion by the electrolyte, or alloys of these metals and other metals (e.g., stainless steel, etc.) can be used. In addition, it is preferable to coat the battery can 602 with nickel, aluminum, or the like to prevent corrosion by the electrolyte. Inside the battery can 602, the battery element in which the positive electrode, negative electrode, and separator are wound is sandwiched between a pair of opposing insulating plates 608 and 609. In addition, a nonaqueous electrolyte (not shown) is injected into the inside of the battery can 602 in which the battery element is provided. The non-aqueous electrolyte may be the same as that used in a coin-type secondary battery.

[0094] Since the positive and negative electrodes used in a cylindrical storage battery are wound, it is preferable to form an active material on both sides of the current collector. A positive electrode terminal (positive electrode current collector lead) 603 is connected to the positive electrode 604, and a negative electrode terminal (negative electrode current collector lead) 607 is connected to the negative electrode 606. Both the positive electrode terminal 603 and the negative electrode terminal 607 can be made of a metal material such as aluminum. The positive electrode terminal 603 is resistance-welded to a safety valve mechanism 612, and the negative electrode terminal 607 is resistance-welded to the bottom of the battery can 602. The safety valve mechanism 612 is electrically connected to the positive electrode cap 601 via a PTC element (Positive Temperature Coefficient) 611. The safety valve mechanism 612 cuts off the electrical connection between the positive electrode cap 601 and the positive electrode 604 when the rise in the internal pressure of the battery exceeds a predetermined threshold value. The PTC element 611 is a thermosensitive resistor whose resistance increases when the temperature rises, and the increase in resistance limits the amount of current to prevent abnormal heat generation. Barium titanate (BaTiO3)-based semiconductor ceramics can be used for the PTC element.

[0095] 10C, a plurality of secondary batteries 600 may be sandwiched between conductive plates 613 and 614 to form a module 615. The plurality of secondary batteries 600 may be connected in parallel, in series, or in parallel and then further in series. By forming a module 615 having a plurality of secondary batteries 600, a large amount of power can be extracted.

[0096] FIG. 10D is a top view of the module 615. To clarify the drawing, the conductive plate 613 is shown by a dotted line. As shown in FIG. 10D, the module 615 may have a conductor 616 that electrically connects the multiple secondary batteries 600. A conductive plate can be provided by overlapping the conductor 616. A temperature control device may also be provided between the multiple secondary batteries 600. When the secondary battery 600 is overheated, it can be cooled by the temperature control device, and when the secondary battery 600 is too cold, it can be heated by the temperature control device. Therefore, the performance of the module 615 is less susceptible to the influence of the outside air temperature. It is preferable that the heat medium of the temperature control device is insulating and non-flammable.

[0097] [Example of secondary battery structure] A laminated secondary battery 980 will be described with reference to Figures 11A, 11B, and 11C. The laminated secondary battery 980 has a wound body 993 shown in Figure 11A. The wound body 993 has a negative electrode 994, a positive electrode 995, and a separator 996. The wound body 993 is formed by stacking the negative electrode 994 and the positive electrode 995 with the separator 996 sandwiched therebetween, and winding the laminated sheet.

[0098] As shown in Fig. 11B, a film 981 serving as an exterior body and a film 982 having a recess are bonded together by thermocompression or the like to form a space, and the above-mentioned wound body 993 is stored in the space, thereby producing a secondary battery 980 as shown in Fig. 11C. The wound body 993 has lead electrodes 997 and 998, and is impregnated with an electrolyte solution between the film 981 and the film 982 having a recess.

[0099] For example, a metal material such as aluminum or a resin material can be used for the film 981 and the film 982 having the recesses. If a resin material is used as the material for the film 981 and the film 982 having the recesses, the film 981 and the film 982 having the recesses can be deformed when an external force is applied, and a flexible storage battery can be produced.

[0100] Although an example in which two films are used for sealing is shown in FIG. 11B and FIG. 11C, a space may be formed by folding one film, and the above-mentioned wound body 993 may be housed in the space.

[0101] The type of the secondary battery shown in FIGS. 9A to 11C in this embodiment is not particularly limited.

[0102] 9A to 11C, a system for estimating capacity or detecting abnormality may be constructed by measuring the time required for the chemical reaction inside the secondary battery to stabilize in advance and following the flow shown in Fig. 3. This embodiment can be freely combined with the first or second embodiment.

[0103] Hardware such as a GPU may be installed in an electronic device or a vehicle to install the learning model shown in the first embodiment. By installing the hardware, a system that estimates the capacity of a secondary battery with high accuracy can be provided. In addition, after charging the secondary battery, a system that performs two-way communication with a server capable of neural network processing using the learning model may be constructed.

[0104] (Embodiment 4) In this embodiment, an example of constructing a system for estimating the capacity of a secondary battery of an electronic device with high accuracy as described in the previous embodiment will be described with reference to Figures 12A to 12E and 13A to 13C. Note that the secondary battery module has at least a secondary battery and a protection circuit.

[0105] First, an example in which a secondary battery module is mounted in a small electronic device will be described with reference to FIGS. 12A to 12C.

[0106] 12A shows an example of a mobile phone. The mobile phone 2100 includes a display unit 2102 built into a housing 2101, as well as operation buttons 2103, an external connection port 2104, a speaker 2105, and a microphone 2106. The mobile phone 2100 also includes a secondary battery module 2107.

[0107] The mobile phone 2100 is capable of executing a variety of applications, such as mobile phone calls, e-mail, document browsing and creation, music playback, Internet communications, and computer games.

[0108] The operation button 2103 can have various functions, such as time setting, power on / off operation, wireless communication on / off operation, silent mode activation / cancellation, power saving mode activation / cancellation, etc. For example, the function of the operation button 2103 can be freely set by an operating system built into the mobile phone 2100.

[0109] The mobile phone 2100 is also capable of performing short-range wireless communication according to a communication standard. For example, the mobile phone 2100 can communicate with a wireless headset to enable hands-free calling.

[0110] The mobile phone 2100 also includes an external connection port 2104, and can directly exchange data with other information terminals via a connector. Charging can also be performed via the external connection port 2104. Note that charging may also be performed by wireless power supply without using the external connection port 2104.

[0111] It is preferable that the mobile phone 2100 has a sensor. For example, it is preferable that a human body sensor such as a fingerprint sensor, a pulse sensor, a body temperature sensor, a touch sensor, a pressure sensor, an acceleration sensor, or the like is mounted as the sensor.

[0112] After charging with a charging device, the mobile phone 2100 can estimate the capacity with high accuracy using a learning model constructed in the charging device or a server capable of two-way communication with the charging device. In addition, the estimated capacity can also be used to detect anomalies.

[0113] FIG. 12B is a perspective view of a device also called a tobacco-receiving smoking device (electronic cigarette). In FIG. 12B, an electronic cigarette 2200 has a heating element 2201 and a secondary battery module 2204 that supplies power to the heating element 2201. When a stick 2202 is inserted into it, the stick 2202 is heated by the heating element 2201. To enhance safety, a protection circuit that prevents overcharging and overdischarging of the secondary battery module 2204 may be electrically connected to the secondary battery module 2204. The secondary battery module 2204 shown in FIG. 12B has an external terminal so that it can be connected to a charging device. Since the secondary battery module 2204 is the tip part when held, it is desirable that the total length is short and the weight is light.

[0114] After charging with a charging device, the secondary battery module 2204 can estimate the capacity with high accuracy using a learning model constructed in the charging device or a server capable of two-way communication with the charging device. In addition, the estimated capacity can also be used to detect anomalies.

[0115] By using a learning model to detect abnormalities after charging, the safety of the secondary battery module can be increased, thereby providing a small, lightweight electronic cigarette 2200 that can be used safely for long periods of time.

[0116] 12C shows unmanned aerial vehicle 2300 having multiple rotors 2302. Unmanned aerial vehicle 2300 has a secondary battery module 2301, a camera 2303, and an antenna (not shown). Unmanned aerial vehicle 2300 can be remotely controlled via the antenna.

[0117] After charging with a charging device, the secondary battery module 2301 can estimate the capacity with high accuracy using a learning model constructed in the charging device or a server capable of two-way communication with the charging device. In addition, the estimated capacity can also be used to detect anomalies.

[0118] By using a learning model to detect abnormalities after charging, the safety of the secondary battery module can be increased, allowing it to be used safely for long periods of time, making it suitable as a secondary battery module to be installed on unmanned aerial vehicle 2300.

[0119] Next, an example in which the capacity estimation system or anomaly detection system for a secondary battery according to one embodiment of the present invention is mounted on a vehicle will be described with reference to FIGS. 12D, 12E, and 13A to 13C.

[0120] 12D shows an electric two-wheeler 2400 using a secondary battery module. The electric two-wheeler 2400 includes a secondary battery module 2401, a display unit 2402, and a handlebar 2403. The secondary battery module 2401 can supply electricity to a motor that serves as a power source. The display unit 2402 can display the remaining charge of the secondary battery module 2401, the speed of the electric two-wheeler 2400, the horizontal state, and the like.

[0121] 12E is an example of an electric bicycle using a secondary battery module. Electric bicycle 2500 includes a battery pack 2502. Battery pack 2502 includes a secondary battery module.

[0122] The battery pack 2502 can supply electricity to a motor that assists the rider. The battery pack 2502 can be removed from the electric bicycle 2500 and carried around. The battery pack 2502 and the electric bicycle 2500 may have a display unit that can display the remaining battery power, etc.

[0123] After charging the battery pack 2502 with a charging device, the battery pack 2502 can estimate the capacity with high accuracy using a learning model constructed in the charging device or in a server capable of two-way communication with the charging device. In addition, the estimated capacity can also be used to detect anomalies.

[0124] By using a learning model to detect anomalies after charging, the safety of the battery pack 2502 can be increased, allowing it to be used safely for long periods of time, making it suitable as an anomaly detection system to be installed on an electric bicycle 2500.

[0125] As shown in FIG. 13A, a secondary battery module 2602 having a plurality of secondary batteries 2601 may be mounted on a hybrid electric vehicle (HEV), an electric vehicle (EV), a plug-in hybrid electric vehicle (PHEV), or other electronic devices.

[0126] FIG. 13B shows an example of a vehicle equipped with a secondary battery module 2602. The vehicle 2603 is an electric vehicle that uses an electric motor as a power source for traveling. Or, it is a hybrid vehicle that can appropriately select and use an electric motor and an engine as a power source for traveling. The vehicle 2603 that uses an electric motor has multiple ECUs (Electronic Control Units), and the ECUs perform engine control and the like. The ECUs include microcomputers. The ECUs are connected to a CAN (Controller Area Network) provided in the electric vehicle. CAN is one of the serial communication standards used as a vehicle-based LAN.

[0127] The ECU uses a CPU and a GPU. A chip that integrates a CPU and a GPU into one is sometimes called an APU (Accelerated Processing Unit), and this APU chip can also be used. An IC (also called an inference chip) incorporating an AI system can also be used. After charging with a charging device, the secondary battery module 2602 can estimate the capacity with high accuracy using a learning model constructed in the charging device or the ECU, or a server capable of two-way communication with the charging device. Furthermore, the estimated capacity can also be used to detect abnormalities.

[0128] By using a learning model to detect abnormalities after charging, the safety of the secondary battery module can be increased, allowing it to be used safely for long periods of time, and it is suitable as a capacity estimation system or anomaly detection system to be installed in a vehicle 2603.

[0129] The secondary battery can not only drive an electric motor (not shown), but also supply power to light-emitting devices such as headlights, room lights, etc. The secondary battery can also supply power to display devices and semiconductor devices such as a speedometer, a tachometer, and a navigation system that the vehicle 2603 has.

[0130] The vehicle 2603 can charge the secondary battery of the secondary battery module 2602 by receiving power supply from an external charging facility by a plug-in method, a non-contact power supply method, or the like.

[0131] FIG. 13C shows a state in which a vehicle 2603 is being charged via a cable from a ground-mounted charging device 2604. When charging, the charging method and connector standards may be appropriately performed using a predetermined method such as CHAdeMO (registered trademark) or Combo. For example, by using plug-in technology, a secondary battery module 2602 mounted on the vehicle 2603 can be charged by an external power supply. Charging can be performed by converting AC power to DC power via a conversion device such as an ACDC converter. The charging device 2604 may be provided in a house as shown in FIG. 13C, or may be a charging station provided in a commercial facility.

[0132] After charging with the ground-mounted charging device 2604, the capacity can be estimated with high accuracy using a learning model constructed in the charging device 2604 or in a server capable of bidirectional communication with the charging device 2604. Also, as shown in the first embodiment, an anomaly detection system can be constructed.

[0133] Although not shown, a power receiving device may be mounted on a vehicle, and charging may be performed by supplying power contactlessly from a ground power transmitting device. In the case of this contactless power supply method, by incorporating a power transmitting device in a road or an exterior wall, charging may be performed not only while the vehicle is stopped but also while the vehicle is moving. This contactless power supply method may also be used to transmit and receive power between vehicles. Furthermore, a solar cell may be provided on the exterior of the vehicle, and a secondary battery may be charged while the vehicle is stopped or moving. For such contactless power supply, an electromagnetic induction method or a magnetic field resonance method may be used.

[0134] 13C includes a power storage system 2612 having a secondary battery module, and a solar panel 2610. The power storage system 2612 is electrically connected to the solar panel 2610 via wiring 2611 or the like. The power storage system 2612 may be electrically connected to a ground-mounted charging device 2604. The power obtained by the solar panel 2610 can be charged to the power storage system 2612. The power stored in the power storage system 2612 can be charged to a secondary battery module 2602 of a vehicle 2603 via the charging device 2604.

[0135] The power stored in the power storage system 2612 can also be supplied to other electronic devices in the house. Therefore, even when power cannot be supplied from the commercial power source due to a power outage or the like, the power storage system 2612 can be used as an uninterruptible power source to enable the use of electronic devices.

[0136] This embodiment mode can be implemented in appropriate combination with other embodiment modes. EXAMPLES

[0137] In this embodiment, an example of a program that has actually been created is shown in FIG. 14, FIG. 15, FIG. 16, FIG. 17, and FIG.

[0138] A CPU or GPU capable of constructing a learning model uses data in memory to access a program (Python in this embodiment) stored on an SSD (or hard disk), reads the program, and the program stored on the SSD (or hard disk) is loaded into memory and deployed as a process in memory.

[0139] The programs for constructing a learning model and for estimating and outputting capacity are shown in Figures 14, 15, 16, 17, and 18. Note that data is referenced in the programs, but since the data is huge, only the file names are shown here and the contents are omitted.

[0140] For example, by transferring the model and parameters obtained by learning using a neural network to an on-board ECU, specifically a microcomputer or microprocessor (hereinafter also referred to as a microcomputer), it is possible to predict the deterioration of an actual vehicle battery. The data for learning is acquired in advance using a secondary battery manufactured in the same manufacturing equipment as the target secondary battery.

[0141] Even if there is a lot of complex data related to the charging and discharging of a secondary battery, two or more parameters, in this embodiment, the CC time and the CV time, are the keys to capacity estimation, and a high level of accuracy in capacity estimation can be obtained by using these parameters and the model obtained by learning using a neural network.

[0142] In addition, when the process of estimating the capacity of a secondary battery is executed by software, the program constituting the software may be installed from a computer in which the program is incorporated in the hardware, or from a network or a recording medium. A program recorded in a recording medium such as a computer-readable CD-ROM (Compact Disk Read Only Memory) is installed, and the program for estimating the capacity of a secondary battery is executed. The process executed by the program is not limited to being executed in a sequence, and does not have to be chronological, and may be executed in parallel, for example. [Explanation of symbols]

[0143] 300: secondary battery, 301: positive electrode can, 302: negative electrode can, 303: gasket, 304: positive electrode, 305: positive electrode current collector, 306: positive electrode active material layer, 307: negative electrode, 308: negative electrode current collector, 309: negative electrode active material layer, 310: separator, 600: secondary battery, 601: positive electrode cap, 602: battery can, 603: positive electrode terminal, 604: positive electrode, 605: separator, 606: negative electrode, 607: negative terminal, 608: insulating plate, 609: insulating plate, 610: gasket, 611: PTC element, 612: safety valve mechanism, 613: conductive plate, 614: conductive plate, 615: module, 616: conductor, 980: secondary battery, 981: film, 982: film, 993: wound body, 994: negative electrode, 995: positive electrode, 996: separator, 997: lead electrode, 998: lead electrode, 2100: mobile phone, 2101: housing, 2102: display unit, 2103: operation button, 2104: external connection port, 2105: speaker, 2106: microphone, 2107: secondary battery module, 2200: electronic cigarette, 2201: heating element, 2202: stick, 2204: secondary battery module, 2300: unmanned aerial vehicle, 2301: secondary battery module, 2302: rotor, 2303: camera, 2400: electric two-wheeler, 2401: secondary battery module, 2402: display unit, 2403: handle, 2500: electric bicycle, 2502: battery pack, 2601: secondary battery, 2602: secondary battery module, 2603: vehicle, 2604: charging device, 2610: solar panel, 2611: wiring, 2612: power storage system

Claims

1. Obtaining first data obtained by a charge-discharge cycle test; A first step of creating a learning model using the first data, The first data includes a CC time, a CV time, and a charge start voltage, A second step of obtaining an estimated value of the capacity after the end of charging using the learning model; The method for detecting an abnormality in a secondary battery further comprises a third step of determining that an abnormality has occurred if the estimated value exceeds a threshold value.

2. In claim 1, The first data further includes a voltage one second after the end of charging and a voltage two seconds after the end of charging.

Images