JP7448201B2 - Secondary battery condition estimation device using artificial intelligence technology - Google Patents

Description

The present invention relates to a secondary battery state estimating device, and more particularly to a secondary battery state estimating device using artificial intelligence technology.

A secondary battery that stores supplied power and supplies the stored power to an electrical load may undergo internal chemical changes, or even if no chemical change occurs, the movement of ions due to charging and discharging may occur. changes are occurring. Various state changes including chemical changes associated with charging and discharging are different in each part of the secondary battery, and it takes a long time for the entire secondary battery to reach a stable state. Therefore, it is difficult to accurately measure the state of a secondary battery that is repeatedly charged and discharged in a short period of time, and it is necessary to wait for a long time until the secondary battery becomes stable. However, momentary changes in the secondary battery during charging and discharging operations are necessary for performing various controls or ensuring safety. For this reason, in the past, the condition was measured moment by moment with a sensor or the like, assuming that a certain amount of error was included.

As a method for reducing the above error, control using artificial intelligence technology (hereinafter referred to as AI technology) equipped with a neural network (hereinafter referred to as AI control) has been proposed. For example, this technology changes various parameters of the secondary battery in advance, accurately measures the characteristics of the secondary battery at that time taking into consideration the state stabilization time of the secondary battery, and then calculates the characteristics based on the parameters and measurement results. It is used to construct a neural network. By constructing a neural network in this manner, the characteristics of the secondary battery in a stable state can be accurately estimated by inputting the parameters of the secondary battery that change from time to time.

However, building a neural network requires a long time and a lot of effort. On the other hand, regarding secondary batteries, material manufacturers and secondary battery manufacturers are very actively engaged in technological development, and improvements are being made from time to time. It is extremely difficult to build neural networks without delay in response to these improved products. The prior art related to the AI technology using the above-mentioned neural network is shown below.

Patent No. 4038788 Patent No. 6579287

In response to the ever-improving secondary batteries, it is desirable to be able to provide a measuring device that can estimate the state of the secondary battery with high precision using AI technology. There is a need for a secondary battery state estimation device that can quickly respond to improvements in the secondary battery being measured. An object of the present invention is to provide a secondary battery state estimating device using AI technology that can relatively easily respond to changes in secondary battery products or technology that are rapidly improving.

A first invention for solving the above problem includes a detection circuit for detecting primary information including terminal voltage and current value of a secondary battery to be measured, and a first neural for estimating a deterioration state of the secondary battery. network, a first correction circuit that corrects the output of the first neural network, and an input device that inputs the name of the product to be measured or the capacity of the secondary battery, and the first neural network inputs the layer, an intermediate layer, and an output layer, the primary information detected by the detection circuit is input to the input device of the first neural network, and is output from the output layer of the first neural network. The device for estimating the state of a secondary battery using artificial intelligence technology is characterized in that the first correction circuit corrects the value of the measured value to estimate deterioration information of the secondary battery to be measured.

A second invention for solving the above-mentioned problem is a secondary battery state estimating device using artificial intelligence technology according to the first invention, further comprising a second invention for estimating the amount of electricity stored in the secondary battery to be measured. a neural network and a second correction circuit, the primary information is input from the detection circuit to the input layer of the second neural network, and the output from the output layer of the second neural network is used for the second correction. The present invention is a secondary battery state estimating device using artificial intelligence technology, characterized in that the amount of charge of the secondary battery to be measured is estimated by being corrected by a circuit.

A third invention for solving the above problem is, in the secondary battery state estimation device using artificial intelligence technology, which is the second invention, the deterioration information corrected by the first correction circuit is transmitted to the second neural network. A secondary battery state estimation device using artificial intelligence technology, characterized in that the input data is input to the input layer of the secondary battery state estimation device.

A fourth invention for solving the above-mentioned problem is a secondary battery state estimation device using artificial intelligence technology, which is the invention described in one of the first to third inventions. It has a third neural network for estimating an abnormality in the secondary battery, and the sudden change value and sudden change period of the primary information are input to the third neural network, and the presence or absence of an abnormality is estimated based on this. This is a secondary battery state estimation device that uses artificial intelligence technology.

According to the present invention, it is an object of the present invention to obtain a secondary battery measuring device using AI technology that can relatively easily cope with secondary batteries that are constantly being improved.

FIG. 2 is an explanatory diagram regarding a secondary battery state estimation device to which the present invention is applied. It is an explanatory diagram explaining a neural network. FIG. 3 is an explanatory diagram illustrating output correction of a neural network. It is a characteristic diagram explaining abnormality of a secondary battery.

1. Regarding an embodiment in which the present invention is applied to an electric vehicle FIG. 1 shows an embodiment in which the present invention is applied to an electric vehicle. Power for driving the electric vehicle is stored in a secondary battery, for example, a lithium ion secondary battery 20, and is transferred from the lithium ion secondary battery 20 to a vehicle drive device 40 that drives the vehicle based on a driving operation by a driver (not shown). Power is supplied. The lithium ion secondary battery 20 is supplied with power from a power generation device (not shown). Furthermore, when a deceleration operation such as a brake operation is performed on the vehicle while the vehicle is running, electric power generated by regenerative braking control is supplied to the lithium ion secondary battery 20 and stored.

The lithium ion secondary battery 20 is repeatedly charged and discharged as the vehicle travels, and gradually deteriorates. Therefore, how much deterioration has progressed in the lithium ion secondary battery 20, how much power is stored in the lithium ion secondary battery 20, and the risk of the lithium ion secondary battery 20 suddenly catching fire? It is desirable that the accurate state of the vehicle is estimated and used for control, etc., and communicated to the driver.

The secondary battery state estimation device 100 includes a primary information detection circuit 120 that detects primary information of the lithium ion secondary battery 20 and the vehicle drive device 40, and a first neural network 130 that outputs basic deterioration data based on the primary information. a second neural network 150 that outputs basic data on the amount of stored electricity based on the primary information and the deterioration degree information corrected by the first neural network 130 or the correction circuit 140; and a third neural network 170 that outputs abnormal basic data. , an input device 190, a correction circuit 140, a correction circuit 160, a correction circuit 180, and a sudden change detection circuit 172. The basic deterioration data, the basic storage amount data, and the basic abnormality data are highly accurate information in themselves, but in order to further improve accuracy, these data can be combined based on input information from the input device 190. , the correction circuit 140, the correction circuit 160, and the correction circuit 180, respectively. Although details will be described below, time is required to construct the first neural network 130, the second neural network 150, and the third neural network 170. On the other hand, the lithium-ion secondary battery 20 whose condition is to be estimated has been subject to various changes, such as new product lineups, small changes to the internal electrode shape, and changes to some of the separator materials. Improvements will be made. If the first neural network 130, second neural network 150, and third neural network 170 themselves are rebuilt each time, it will be difficult to deal with the problem. By correcting these using the correction circuit 140, the correction circuit 160, and the correction circuit 180, sufficiently high accuracy can be maintained. Further, the configurations of the correction circuit 140, the correction circuit 160, and the correction circuit 180 are not particularly specified, and various circuits can be used.

2. About driving an electric vehicle Engine vehicles are powered by the output of a gasoline or diesel engine. Due to environmental concerns, hybrid cars and electric cars have come into use in recent years, in which electric power is stored in a lithium-ion secondary battery 20 mounted on the vehicle, and a motor is driven by the stored electric power. Whether it is a hybrid car or an electric car, a lithium ion secondary battery 20 is required to store the power necessary for driving. An example of an electric vehicle will be explained using the lithium ion secondary battery 20 and vehicle drive device 40 shown in FIG. Note that even if it is a hybrid car, the basic parts are the same.

Electric power for running is stored in a lithium ion secondary battery 20. The vehicle drive device 40 is provided with a driving motor, an accelerator pedal, a brake pedal, and a steering mechanism. When the driver depresses the accelerator pedal, power is supplied to the driving motor, and the vehicle accelerates and decelerates based on the amount of operation. During acceleration, the power necessary for acceleration is supplied from the lithium ion secondary battery 20 to the vehicle drive device 40, and during deceleration, the power recovered by regenerative braking is stored from the vehicle drive device 40 in the lithium ion secondary battery 20. . In this way, the value of the current flowing out from the lithium ion secondary battery 20 changes moment by moment.

When the lithium ion secondary battery 20 is used for a long time and is repeatedly charged and discharged, its deterioration progresses, resulting in problems such as, for example, not being able to extract sufficient power even after being charged. Conventionally, when the internal resistance of the lithium ion secondary battery 20 increases, it is determined that the battery has deteriorated, and it is assumed that the battery life has come to an end and the battery is replaced. Inside the lithium ion secondary battery 20, phenomena such as movement of ions occur, and charging and discharging are performed due to phenomena such as movement of ions. When attempting to measure the deterioration of the lithium ion secondary battery 20, highly accurate measurement cannot be performed unless the state of the internal ions is stable. It takes a long time for the state of the lithium ion secondary battery 20 to become stable. However, responsiveness is important for performing various controls such as driving control, and if you wait until the system stabilizes, you will not be able to use the measured information for control. As described above, this problem can be solved by constructing the first neural network 130, the second neural network 150, and the third neural network 170, which will be described below.

The constantly changing states of the lithium ion secondary battery 20 and the vehicle drive device 40 are captured as primary information by the primary information detection circuit 120. The primary information 124 regarding the lithium ion secondary battery 20 taken into the primary information detection circuit 120 is information including, for example, terminal voltage and current, internal resistance of the lithium ion secondary battery 20, outside temperature, and the like. Further, the primary information 122 regarding the vehicle drive device 40 that is taken in by the primary information detection circuit 120 is, for example, information such as the vehicle speed, the amount of change in acceleration/deceleration, and the travel distance. After these pieces of information are measured by the primary information detection circuit 120, they are sent to the respective input layers as inputs to the first neural network 130, second neural network 150, and third neural network 170.

The basic configurations of the first neural network 130, the second neural network 150, and the third neural network 170 are almost the same, and the first neural network 130 will be used as a representative example of these neural networks. As shown in FIG. 2, the first neural network 130 has an input layer 132, a hidden layer 134, and an output layer 136. This is the same for the other second neural network 150 and third neural network 170, and the configurations described below are also the same. The input layer 132, intermediate layer 134, and output layer 136 each include a neuron group composed of neurons indicated by circles. When the driving-related primary information 122 and the secondary battery-related primary information 124 are input to the input layer 132, even though the lithium-ion secondary battery 20 is not actually in a stable state, the degree of deterioration measured in the stable state is displayed. Degraded basic data representing . . . is estimated and output from the output layer 136. For example, in the example of the first neural network 130, the driving-related primary information 122 and the secondary battery-related primary information are obtained so that such outputs can be obtained from the first neural network 130, the second neural network 150, and the third neural network 170. The parameters of the information 124 are changed in advance, and the deterioration state of the lithium ion secondary battery 20 under the changed state is measured by waiting for the lithium ion secondary battery 20 to reach a stable state over time. , the coefficient value of each neuron in the intermediate layer 134 is determined from the states of the running-related primary information 122 and the secondary battery-related primary information 124 so that the deterioration information in the stable state can be obtained. By doing this, when the driving-related primary information 122 or the secondary battery-related primary information 124, which is unstable and changes from moment to moment, is input, the lithium ion secondary battery 20 is immediately stabilized based on these information. Deterioration information can be obtained with high accuracy. The estimation of the amount of stored electricity by the second neural network 150 and the above estimation by the third neural network 170 are also the same.

3. Regarding the measurement of the amount of stored electricity, the second neural network 150 and the third neural network 170 basically have the same configuration as the first neural network 130 shown in FIG. 2, and the method of constructing the intermediate layer 134 is also the same. In the case of the second neural network 150 for obtaining power storage information, the input layer 132 of the second neural network 150 includes the output of the first neural network 130 in addition to the driving-related primary information 122 and the secondary battery-related primary information 124. Alternatively, the output of the correction circuit 140 is input. The reason for inputting the output of the first neural network 130 or the output of the correction circuit 140 to the second neural network 150 is as follows.

When the amount of stored electricity is estimated based on the secondary battery-related primary information 124, even if the estimated amount is the same, the amount of power that can actually be taken out from the lithium ion secondary battery 20 depends on the deterioration of the lithium ion secondary battery 20. It varies depending on the condition. As the lithium ion secondary battery 20 deteriorates, the amount of movement of ions within the lithium ion secondary battery 20 due to discharge decreases. Also, the moving speed of ions decreases. Furthermore, there are various other factors that increase internal resistance. As a result, the amount of power that can be extracted from the lithium ion secondary battery 20 is smaller than in a state where the degree of deterioration is low. For this reason, the output of the first neural network 130 or the output of the correction circuit 140 is input to the input layer 132 of 150. The second neural network 150 outputs basic power storage amount data. Further, as will be explained below, the third neural network 170 outputs abnormality information regarding whether an abnormal state occurs.

4. Regarding the correction circuit 140, the correction circuit 160, and the correction circuit 180, the first neural network 130, the second neural network 150, and the third neural network 170 have the configurations shown in FIG. The state of the output for the intermediate layer 134 is constructed from the accurate measurement of the state of the output. Building this intermediate layer 134 requires time and is difficult to change easily. On the other hand, the lithium ion secondary battery 20 is constantly being improved in terms of electrolyte materials, separator materials, and the like. We also offer products with different capacities and sizes. In such a case, the operating state of the lithium ion secondary battery 20 is mostly the same, and high accuracy can often be maintained by changing only a part of it. For these reasons, by correcting the outputs of the first neural network 130, second neural network 150, and third neural network 170 using the correction circuit 140, correction circuit 160, and correction circuit 180, it is possible to follow changes in the product. , it is possible to obtain highly accurate estimation results. These estimation results are each displayed on the display device 200.

FIG. 3 shows the basic configurations of the correction circuit 140, the correction circuit 160, and the correction circuit 180. The correction circuit 140 will be described as a representative of these correction circuits, but the same applies to the correction circuit 160 and the correction circuit 180. From the input device 190, the manufacturer name 192 of the lithium ion secondary battery 20 that is the target of state estimation, a name (hereinafter referred to as type name) 194 that can identify the type of product, such as the product name or model, and the lithium ion 2 The power capacity 196 of the secondary battery 20 is input into the table 142 for searching the correction coefficient k of the correction circuit 140. Note that FIG. 3 is just an example, and any circuit or method may be used. Further, the correction may be performed by calculation or by other methods.

When the outputs of the first neural network 130, the second neural network 150, and the third neural network 170 are input i, the arithmetic circuit 144 performs calculations such as multiplication using the correction coefficient k retrieved from the table 142 and the input i. Accordingly, it is possible to correct input i for small-scale changes in the lithium ion secondary battery 20. By doing so, it is possible to estimate the degree of deterioration, the amount of stored electricity, and abnormalities with high accuracy even when various changes are made to the lithium ion secondary battery 20 that are frequently performed, and the display device 200 can be displayed.

5. Regarding estimation of the abnormal state of the lithium ion secondary battery 20 If the lithium ion secondary battery 20 causes a short circuit accident, it will become high temperature and there is a danger of it catching fire. However, signs of such an abnormality often appear before it occurs. An example is shown in FIG. FIG. 4 is a graph showing changes in terminal voltage when charging is performed. Characteristic 224 is a normal characteristic. When charging and discharging are repeated, characteristics 226 or 220 may appear momentarily based on deterioration of the lithium ion secondary battery 20. However, characteristic 226 or characteristic 220 quickly disappears and returns to normal characteristic 224. Repeating this process eventually leads to the occurrence of major abnormalities.

The sudden change detection circuit 172 in FIG. 1 captures the instantaneous characteristic changes, such as the characteristics 220 and 226 in FIG. 4, and also captures the period of occurrence thereof. The third neural network 170 determines whether this will lead to a major abnormality or whether it is just noise. In this case, when the third neural network 170 determines that there is an abnormality, the abnormality may be displayed on the display device 200 without providing the correction circuit 180. Of course, the correction circuit 180 may be adapted to accommodate small-scale changes to the lithium ion secondary battery 20 or the addition of new products.

As described above, according to the present embodiment, it is possible to easily and efficiently respond to small-scale changes in products and the addition of new products, and moreover, it is possible to estimate with high accuracy.

20... Lithium ion secondary battery, 40... Vehicle drive device, 100... Secondary battery state estimation device, 120... Primary information detection circuit, 122... Driving related primary information, 124... - Secondary battery related primary information, 130... First neural network, 132... Input layer, 134... Intermediate layer, 136... Output layer, 140... Correction circuit, 142... Table , 144... Arithmetic circuit, 150... Second neural network, 160... Correction circuit, 170... Third neural network, 172... Sudden change detection circuit, 180... Correction circuit, 190... ...Input device, 200...Display device, 220...Characteristics, 224...Characteristics, 226...Characteristics.

Claims (6)

a primary information detection circuit that detects primary information including terminal voltage and current of a lithium-ion secondary battery that is repeatedly charged and discharged;
A first neural comprising an input layer, an intermediate layer, and an output layer, and outputs degraded basic data from the output layer by inputting the primary information detected by the primary information detection circuit to the input layer. network and
By correcting the basic deterioration data output from the output layer of the first neural network, obtaining an estimation result regarding the degree of deterioration of the lithium ion secondary battery following changes to the lithium ion secondary battery. a first correction circuit;
Equipped with
The primary information of the lithium ion secondary battery is measured by the primary information detection circuit,
The measured primary information is input to the input layer of the first neural network, and by inputting the primary information to the first neural network, the lithium ion secondary information is transmitted from the output layer of the first neural network to the input layer of the first neural network. The basic battery deterioration data is output,
The deterioration basic data output from the output layer of the first neural network is input to the first correction circuit,
In the first correction circuit, the input basic deterioration data is corrected to obtain the estimation result regarding the degree of deterioration of the lithium ion secondary battery that follows the change to the lithium ion secondary battery. be able to
The change to the lithium ion secondary battery is handled by correcting the basic deterioration data output from the first neural network using the first correction circuit.
A secondary battery state estimation device using artificial intelligence technology. A secondary battery state estimation device using artificial intelligence technology according to claim 1,
A product specifying information input device is further provided for inputting product specifying information for specifying the product of the lithium ion secondary battery to the first correction circuit,
By inputting the product specific information to the first correction circuit, the first correction circuit searches for a correction coefficient of the lithium ion secondary battery,
In the first correction circuit, by correcting the basic deterioration data from the first neural network using the correction coefficient, the lithium ion secondary battery can be adjusted to follow the change to the lithium ion secondary battery. the estimation result regarding the degree of deterioration is obtained;
A secondary battery state estimation device using artificial intelligence technology. A secondary battery state estimation device using artificial intelligence technology according to claim 2,
The first correction circuit includes a table for searching the correction coefficient based on the product specification information from the product specification information input device, and a search function for the first neural network based on the correction coefficient searched by the table. It is equipped with an arithmetic circuit for correcting the deterioration basic data from the
In the first correction circuit, the deterioration basic data from the first neural network is corrected by the arithmetic circuit using the correction coefficient, so that the lithium the estimation result regarding the degree of deterioration of the ion secondary battery is obtained;
A secondary battery state estimation device using artificial intelligence technology. In the secondary battery state estimation device using artificial intelligence technology according to claim 3,
further comprising a display device;
displaying the estimation result regarding the degree of deterioration obtained in the first correction circuit on the display device ;
A device for estimating the state of a secondary battery using artificial intelligence technology. In the secondary battery state estimation device using artificial intelligence technology according to claim 3,
a second neural network that outputs basic data on the storage amount of the lithium ion secondary battery by receiving the primary information from the primary information detection circuit; and the storage amount output from the second neural network. A second correction circuit for correcting the basic data is further provided,
The second correction circuit corrects the basic power storage amount data output from the second neural network, thereby determining the power storage amount of the lithium ion secondary battery that follows the change to the lithium ion secondary battery. so that estimation results can be obtained,
The change to the lithium ion secondary battery is handled by correcting the basic storage amount data output from the second neural network using the second correction circuit.
A device for estimating the state of a secondary battery using artificial intelligence technology. A secondary battery state estimation device using artificial intelligence technology according to any one of claims 1 to 5,
The lithium ion secondary battery is a lithium ion secondary battery for supplying power to a drive device of a vehicle.
A secondary battery state estimation device using artificial intelligence technology.