Learning device, estimation device, learning method, estimation method, learning program product, and estimation program product

Abstract

A learning device is provided, comprising: a characteristic acquisition unit that acquires any one of the voltage and capacity of a battery under multiple conditions in at least one of charging and discharging of the battery, and characteristics corresponding to changes in the capacity and voltage of the battery; and a learning unit that learns the relationship between any one of the voltage and capacity of the battery under the multiple conditions and the characteristics, and generates an estimation model for estimating at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics.

Description

Technical Field

[0001] The present invention relates to a learning device, an estimation device, a learning method, an estimation method, a learning program, and an estimation program. Background Art

[0002] In the past, it is known to use LiFePO 4 and a model of various parameters associated with the phase transition behavior of graphite to estimate the state of the battery (for example, refer to non-patent document 1), and a technology for estimating the state of the battery using a probabilistic neural network (for example, refer to non-patent document 2).

[0003] Prior art literature

[0004] Non-patent literature

[0005] Non-patent document 1: Soichiro Torai, "State-of-health estimation of LiFePO4 / graphite batteries based on a model using differential capacity", Journal of Power Sources 306 (2016)

[0006] Non-patent document 2: Ho-Ta Lin, et al., "Estimation of Battery State of Health Using Probabilistic Neural Network", IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL.9, NO.2, May 2013 Summary of the invention

[0007] Problems to be solved by the invention

[0008] In the technology of non-patent document 1, modeling is performed based on the material level of the battery, so detailed information on the battery material is required. On this basis, since there are a large number of parameters, the calculation burden is also high. In addition, in the technology of non-patent document 2, a lot of time and cost are required to generate the model. However, it is desired to easily estimate the state of the battery based on the data in the operation of the battery.

[0009] General Public

[0010] In order to solve the above-mentioned problems, in a first aspect of the present invention, a learning device is provided. The learning device may include a characteristic acquisition unit, which acquires any one of the voltage and capacity of the battery under multiple conditions in at least one of charging and discharging of the battery, and characteristics corresponding to the capacity change and voltage change of the battery. The learning device may include a learning unit, which learns the relationship between any one of the voltage and capacity of the battery under multiple conditions and the characteristics, and generates an estimation model for estimating at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics.

[0011] The learning unit may generate an estimation model that interpolates the relationship under the different conditions based on the relationship under at least one condition among the plurality of conditions.

[0012] The learning unit may generate an estimation model using a modular network-type self-organizing graph using a neural network as a module.

[0013] The characteristic may be a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage.

[0014] The learning unit may associate the maximum capacity of the storage battery with each of the conditions.

[0015] The learning unit may calculate the maximum capacity of the storage battery associated with each condition based on an integrated characteristic obtained by integrating the capacity differential characteristic with the voltage of the storage battery.

[0016] The characteristic may be a voltage differential characteristic obtained by differentiating the voltage of the battery by the capacity.

[0017] In a second aspect of the present invention, an estimation device is provided. The estimation device may include a model acquisition unit that acquires an estimation model that estimates at least one of the maximum capacity and the remaining capacity of a battery based on the relationship between any one of the voltage and capacity of the battery under multiple conditions in at least one of charging and discharging of the battery, and the characteristics corresponding to the capacity change and voltage change of the battery. The estimation device may include: a data acquisition unit that acquires time series data of the voltage and current of the battery. The estimation device may include a calculation unit that calculates the characteristics based on the time series data of the voltage and current of the battery. The estimation device may include an estimation unit that uses the estimation model to estimate at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the calculated characteristics.

[0018] The estimated model may be an estimated model in which the relationship under a different condition is interpolated based on the relationship under at least one condition among a plurality of conditions.

[0019] The estimation model may be an estimation model learned using a modular network-type self-organizing graph using a neural network as a module.

[0020] The characteristic may be a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage.

[0021] The estimating unit may estimate the maximum capacity of the storage battery based on an integrated characteristic obtained by integrating the capacity differential characteristic with the voltage of the storage battery.

[0022] The characteristic may be a voltage differential characteristic obtained by differentiating the voltage of the battery by the capacity.

[0023] The estimating unit may acquire a model voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity based on the estimation model, and estimate the remaining capacity of the storage battery with reference to the model voltage differential characteristic.

[0024] The estimating unit may estimate the remaining capacity of the battery by fitting a partial voltage differential characteristic obtained by differentiating the battery voltage with the capacity, calculated based on the time series data of the battery voltage and current, using the model voltage differential characteristic as a reference.

[0025] The estimating unit may determine whether or not at least one of the maximum capacity and the remaining capacity of the battery can be estimated based on the time series data of the voltage and current of the battery.

[0026] In a third aspect of the present invention, a learning method for learning by a learning device is provided. The learning method may include: obtaining any one of the voltage and capacity of the battery under multiple conditions in at least one of charging and discharging of the battery, and characteristics corresponding to the capacity change and voltage change of the battery. The learning method may include: learning the relationship between any one of the voltage and capacity of the battery under multiple conditions and the characteristics, respectively, and generating an estimation model for estimating at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics.

[0027] In a fourth aspect of the present invention, an estimation method is provided for estimation by an estimation device. The estimation method may include: obtaining an estimation model, the estimation model estimating at least one of the maximum capacity and the remaining capacity of the battery based on the relationship between any one of the voltage and capacity of the battery under multiple conditions in at least one of charging and discharging of the battery, and the characteristics corresponding to the capacity change and voltage change of the battery. The estimation method may include: obtaining time series data of the voltage and current of the battery. The estimation method may include: calculating the characteristics based on the time series data of the voltage and current of the battery. The estimation method may include: using the estimation model, estimating at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the calculated characteristics.

[0028] In a fifth aspect of the present invention, a learning program is provided. The learning program can be executed by a computer. The learning program can cause the computer to function as a characteristic acquisition unit, which acquires any one of the voltage and capacity of the battery under multiple conditions in at least one of the charging and discharging of the battery, and the characteristics corresponding to the capacity change and voltage change of the battery. The learning program can cause the computer to function as a learning unit, which learns the relationship between any one of the voltage and capacity of the battery under multiple conditions and the characteristics, and generates an estimation model for estimating at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics.

[0029] In a sixth aspect of the present invention, an estimation program is provided. The estimation program can be executed by a computer. The estimation program can make the computer function as a model acquisition unit that acquires an estimation model that estimates at least one of the maximum capacity and the remaining capacity of the battery based on the relationship between any one of the voltage and capacity of the battery under multiple conditions in at least one of charging and discharging of the battery, and the characteristics corresponding to the capacity change and voltage change of the battery. The estimation program can make the computer function as a data acquisition unit that acquires time series data of the voltage and current of the battery. The estimation program can make the computer function as a calculation unit that calculates the characteristics based on the time series data of the voltage and current of the battery. The estimation program can make the computer function as an estimation unit that uses the estimation model to estimate at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the calculated characteristics.

[0030] In addition, the above summary of the invention does not list all the features required by the present invention. In addition, sub-combinations of these feature groups also constitute inventions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The learning device 100 and the battery 10 according to the present embodiment are shown.

[0032] Figure 2 An example of data pre-processing performed by the calculation unit 150 of the learning device 100 according to the present embodiment is shown.

[0033] Figure 3 An example of the estimation model 300 generated by the learning device 100 according to the present embodiment is shown.

[0034] Figure 4 The following is a flow of generating the estimation model 300 by the learning device 100 according to the present embodiment.

[0035] Figure 5 Indicates according to Figure 4 The mnSOM was completed by learning through the learning process.

[0036] Figure 6 An example is shown in which the learning device 100 according to the present embodiment associates the maximum capacity of the battery 10 with each module of the mnSOM.

[0037] Figure 7 An estimation device 700 and a battery 10 according to the present embodiment are shown.

[0038] Figure 8 The following is a flow of estimating the state of battery 10 by estimation device 700 according to the present embodiment.

[0039] Fig. 9 An example of curve fitting performed when the estimation device 700 according to the present embodiment estimates the remaining capacity of the battery 10 is shown.

[0040] Fig.10 An example in which the estimation device 700 according to the modification of the present embodiment is realized by cloud computing is shown.

[0041] Fig.11 The computer 2200 is an example in which various aspects of the present invention may be embodied in whole or in part. DETAILED DESCRIPTION

[0042] Hereinafter, the present invention will be described by way of the embodiments of the invention, however, the following embodiments do not limit the invention involved in the claims. In addition, the solution means of the invention do not necessarily require all combinations of the features described in the embodiments.

[0043] Figure 1The learning device 100 and the storage battery 10 according to the present embodiment are shown. The learning device 100 according to the present embodiment generates an estimation model that estimates at least one of the maximum capacity and the remaining capacity of the storage battery 10 that can be repeatedly used by charging.

[0044] Here, the maximum capacity refers to the maximum amount of electricity [Ah] that can be stored in the storage battery 10 within a specified battery voltage range. Generally, as the storage battery is repeatedly charged and discharged, the characteristics of the active material responsible for electrical conduction are deactivated or side reactions occur, resulting in a gradual decrease in maximum capacity. The learning device 100 involved in this embodiment generates a model for estimating the maximum capacity of the storage battery 10 that has deteriorated due to repeated charge and discharge.

[0045] In addition, the so-called residual capacity refers to the remaining amount of electricity [Ah] that can be discharged from the battery from the current time point. The learning device 100 involved in this embodiment generates a model for estimating the residual capacity of such a battery 10. Generally, in order to safely and efficiently operate the power storage system for a long time, it is extremely important to understand these maximum capacities and residual capacities.

[0046] The learning device 100 may be a computer such as a PC (personal computer), a tablet computer, a smart phone, a workstation, a server computer, or a general-purpose computer, or a computer system obtained by connecting multiple computers. Such a computer system is also a computer in a broad sense. In addition, the learning device 100 may also be implemented in a computer through one or more executable virtual computer environments. Alternatively, the learning device 100 may be a dedicated computer designed for learning an inference model, or may be dedicated hardware implemented by a dedicated circuit.

[0047] The learning device 100 involved in the present embodiment is set as a battery 10 that is the object of learning, and supplies power to a load not shown. In recent years, batteries have been widely used in vehicles such as hybrid vehicles and electric vehicles, laptop computers, digital cameras, mobile phones, and smart phones. Generally, these batteries are constructed as a battery pack in which a battery module including a plurality of battery cells is connected in series and parallel. The learning device 100 involved in the present embodiment is set as a battery 10 that is the object of learning, and it may be, for example, at least one battery module in a battery pack constructed in this way. During the period when the learning device 100 obtains training data for generating an inferred model, the battery 10 may also be operated in a tank such as a thermostatic tank whose internal temperature is kept fixed. The battery 10 includes a plurality of battery cells 12 and a current sensor 14.

[0048] The plurality of battery cells 12 are connected in series. The battery cell 12 is a component unit of the battery, and is also called a single cell. By connecting a plurality of such battery cells 12 in series, the storage battery 10 obtains a desired output voltage.

[0049] The current sensor 14 is connected in series to the plurality of battery cells 12 and measures the current flowing through the storage battery 10. The current sensor 14 supplies the measured current to the learning device 100.

[0050] The learning device 100 according to the present embodiment includes a voltage sensor 110 , a temperature sensor 120 , an A / D converter 130 , a data acquisition unit 140 , a calculation unit 150 , a characteristic acquisition unit 160 , a learning unit 170 , and a model output unit 180 .

[0051] The voltage sensor 110 measures the voltage value obtained when all the battery cells 12 are connected in series, that is, the voltage between the terminals of the battery 10. Alternatively, the voltage sensor 110 may measure the voltage between the terminals of each of the battery cells 12. The voltage sensor 110 supplies the measured voltage to the A / D converter 130.

[0052] The temperature sensor 120 measures the temperature of the battery 10. The temperature sensor 120 measures the temperature of the battery 10 at a position of any one of the plurality of battery cells 12 as a representative point. The temperature sensor 120 may also measure the temperature of each of the plurality of battery cells 12. The temperature sensor 120 supplies the measured temperature to the A / D converter 130.

[0053] The A / D converter 130 samples the current measured by the current sensor 14, the voltage measured by the voltage sensor 110, and the temperature measured by the temperature sensor 120 at a predetermined sampling period, and converts the sampled analog data into digital data. Then, the A / D converter 130 supplies the voltage V, current I, and temperature T of the battery 10 converted into digital data to the data acquisition unit 140.

[0054] The data acquisition unit 140 acquires the time series data of the voltage V and the current I of the battery 10 converted into digital data by the A / D converter 130. In addition, the data acquisition unit 140 acquires the temperature T of the battery 10 converted into digital data by the A / D converter 130 together with the time series data of the voltage V and the current I of the battery 10. When the voltage sensor 110 measures the terminal voltage of each of the plurality of battery cells 12, the data acquisition unit 140 may acquire the voltage V of the battery 10 from, for example, the sum of the terminal voltages of each of the plurality of battery cells 12. In addition, when the temperature sensor 120 measures the temperature of each of the plurality of battery cells 12, the data acquisition unit 140 may acquire the temperature T of the battery 10 based on, for example, statistics (for example, average) of the temperatures of each of the plurality of battery cells 12. Alternatively, the data acquisition unit 140 may acquire the set temperature of the thermostatic chamber as the temperature T of the battery 10 instead of using the temperature measured by the temperature sensor 120. However, there is a case where the set temperature of the thermostatic chamber is different from the actual surface temperature of the battery cell 12. Therefore, it is preferable that the temperature T of the battery 10 is the temperature actually measured by the temperature sensor 120 .

[0055] The calculation unit 150 performs preprocessing of the time series data of the voltage V and the current I of the battery 10 acquired by the data acquisition unit 140. The calculation unit 150 calculates, for example, the voltage V and the capacity Q of the battery 10 under a plurality of conditions in at least one of the charging and discharging of the battery 10, and the characteristics corresponding to the capacity change and the voltage change of the battery 10 based on the time series data of the voltage V and the current I of the battery 10. Here, the characteristics may be capacity differential characteristics obtained by differentiating the capacity of the battery 10 by the voltage.

[0056] The characteristic acquisition unit 160 acquires any one of the voltage V and the capacity Q of the battery 10 under a plurality of conditions in at least one of charging and discharging of the battery 10 , and characteristics corresponding to changes in capacity and voltage of the battery 10 .

[0057] The learning unit 170 learns the relationship between the voltage V and the capacity Q of the battery 10 under a plurality of conditions and the characteristics corresponding to the capacity change and the voltage change of the battery 10, and generates an estimation model for estimating at least one of the maximum capacity and the remaining capacity of the battery 10 based on the voltage V and the capacity Q of the battery 10 and the characteristics. This will be described later.

[0058] The model output unit 180 outputs the estimated model learned by the learning unit 170. The model output unit 180 supplies the estimated model to other functional units in the device. In addition, the model output unit 180 may supply the estimated model to other devices via a network, for example.

[0059] In the above description, the case where the learning device 100 includes the voltage sensor 110, the temperature sensor 120, and the A / D converter 130 is described as an example, but the present invention is not limited to this. At least one of the voltage sensor 110, the temperature sensor 120, and the A / D converter 130 may be included in a device different from the learning device 100.

[0060] Figure 2 An example of data preprocessing performed by the calculation unit 150 of the learning device 100 according to the present embodiment is shown. The data acquisition unit 140 acquires the voltage V and current I of the storage battery 10 under a plurality of conditions, and the time t. In addition, the plurality of conditions mentioned here refers to a condition in which at least one of the temperature T, the charge and discharge rate, and the SOH (State of Health) of the storage battery 10 is different. Here, the charge and discharge rate refers to the current amount during charging and discharging, with the current amount that fully charges the storage battery in 1 hour being 1C. In addition, the so-called SOH refers to the maximum capacity at the current time point relative to the initial maximum capacity expressed as a percentage. On the left side of this figure, as an example, the voltage V and current I of the storage battery 10, and the time t acquired by the data acquisition unit 140 under a certain condition (temperature T = ○○℃, charge and discharge rate = ○○C, SOH = ○○%) are shown.

[0061] The calculation unit 150 performs interval averaging processing and moving average processing on the data acquired by the data acquisition unit 140, and smoothes the data. Next, the calculation unit 150 multiplies the current I by the time t and integrates the data as the time t passes, thereby calculating the capacity Q [Ah] as shown in the center of this figure. Then, the calculation unit 150 differentiates the capacity Q with the voltage V, thereby converting it into a capacity differential characteristic dQ / dV [Ah / V] obtained by differentiating the capacity Q of the storage battery 10 with the voltage V. The calculation unit 150 performs such pre-processing on each of the data acquired by the data acquisition unit 140 under each condition, thereby obtaining the relationship between the voltage V of the storage battery 10 and the capacity differential characteristic dQ / dV under each condition as shown on the right side of this figure.

[0062] Figure 3An example of an inferred model 300 generated by the learning device 100 involved in the present embodiment is shown. The learning unit 170 generates an inferred model that interpolates the relationship under different conditions based on the relationship under at least one condition among multiple conditions. As an example, the learning unit 170 can also use a modular network self-organizing map (mnSOM: modular network Self Organizing Map) with a neural network as a module to generate an inferred model. However, it is not limited to this. The learning unit 170 can also generate an inferred model using clustering such as the k-means method, principal component analysis, and other unsupervised learning such as independent component analysis, and can also generate an inferred model using regression such as the k-nearest neighbor method, regression tree, random forest, neural network, support vector regression, and projection tracking regression.

[0063] SOM (Self Organizing Map) is a known model that expresses the similarity of given input information by distance on a graph. SOM is composed of a plurality of regularly arranged units. Each unit has a reference vector of the same dimension as the input vector. Furthermore, SOM compares the input vector with the reference vector and transforms the input vector into the coordinates on the graph of the unit having the reference vector closest to the input vector. That is, SOM maps the input vector from the vector space where the input vector exists to the graph space where the units are arranged.

[0064] SOM itself learns in such a way that the relationship between input vectors in the vector space is also maintained in the mapped graph space. That is, SOM obtains an in-phase mapping from the vector space to the graph space through unsupervised learning. Specifically, if an input vector is given to SOM, the unit with the reference vector closest to the input vector is determined to be the winning unit. At this time, learning is performed in such a way that the unit closer to the winning unit on the graph obtains the right to learn more strongly with respect to the input vector, and the reference vector is brought closer to the input vector according to the strength. Regarding the strength of learning at this time, the winning unit is the strongest, and the farther away from the winning unit on the graph, the weaker it becomes. By repeatedly updating the reference vector in this way for multiple input vectors, SOM can obtain a graph that preserves the relationship between the input vectors in the vector space.

[0065] Thus, SOM is a two-layer structured unsupervised learning neural network consisting of an input layer and an output layer, and maps input data to an output layer of any dimension through unsupervised learning. That is, SOM is a model that maps input data to an output layer in which a plurality of units are regularly arranged through unsupervised learning, and the plurality of units are respectively associated with vectors of the same dimension as the input data, and SOM is a model that learns by updating the vectors associated with each unit.

[0066] The mnSOM replaces the reference vector unit in such a SOM with a module such as a neural network. The learning unit 170 of the learning device 100 involved in this embodiment uses the mnSOM to generate an estimation model 300. The estimation model 300 has a plurality of modules 310. In this embodiment, each module 310 is a 1-input 1-output neural network structure based on a multilayer perceptron, and the number of layers, the number of nodes in the hidden layer, and the activation function in each module 310 can be set to be arbitrary. The learning unit 170 sets the voltage V of the battery 10 as an input vector and the capacity differential characteristic dQ / dV as an output vector, and learns each module 310 through a neural network to generate the estimation model 300. However, it is not limited to this. Each module 310 is not limited to a neural network, and can also be a structure such as linear regression, k-nearest neighbor method, regression tree, random forest, support vector regression, decision tree, Bayesian network, hidden Markov model, etc. that can be learned in a manner that the voltage V of the battery 10 is set as an input and the capacity differential characteristic dQ / dV is set as an output.

[0067] Figure 4The flow of generating the estimation model 300 by the learning device 100 according to the present embodiment is shown. In this flow, as an example, the learning unit 170 uses the training data under different SOH conditions (e.g., SOH=64%, 69%, 86%, 92%, and 100%) in a common temperature condition (e.g., temperature T=25°C) and a common charge and discharge rate condition (e.g., charge and discharge rate=0.2C charging) to form a graph. In this case, the learning unit 170 uses the training data under different conditions of at least one of the temperature condition and the charge and discharge rate condition to form another graph different from the above graph. That is, the learning unit 170 forms a plurality of graphs corresponding to a plurality of conditions that are different from at least one of the temperature condition and the charge and discharge rate condition. Alternatively, the learning unit 170 may also use training data under different temperature conditions (temperature T = 25°C and 45°C) and different SOH conditions (for example, SOH = 64%, 69%, 86%, 92%, and 100%) in a common charge and discharge rate condition (for example, charge and discharge rate = 0.2C) to form a graph. That is, the learning unit 170 may also use training data under 10 conditions, for example, "temperature T = 25°C, charge and discharge rate = 0.2C, SOH = 64%, 69%, 86%, 92%, and 100%", and "temperature T = 45°C, charge and discharge rate = 0.2C, SOH = 64%, 69%, 86%, 92%, and 100%", to form a graph.

[0068] In step 410 , the data acquisition unit 140 acquires time series data of the voltage V and the current I of the storage battery 10 under different SOH conditions.

[0069] In step 420, the calculation unit 150 calculates Figure 2 The data is pre-processed according to the description, and the relationship between the voltage V and the capacity differential characteristic dQ / dV of the battery 10 under each SOH condition is obtained.

[0070] In step 430, the learning unit 170 determines the winning module based on the relationship between the voltage V and the capacity differential characteristic dQ / dV of the battery 10 under one of the different SOH conditions (for example, SOH=64%). The learning unit 170, for example, inputs the training data, i.e., the vector of the voltage V under SOH=64%, to each module, and obtains the output of the capacity differential characteristic dQ / dV value vector calculated by the neural network in each module. Next, the learning unit 170 calculates the Euclidean distance between the calculated capacity differential characteristic dQ / dV value vector and the training data, i.e., the capacity differential characteristic dQ / dV value vector under SOH=64%, for each module. Then, the learning unit 170 determines the module with the smallest Euclidean distance as the winning module.

[0071] In step 440, the learning unit 170 learns each module 310 based on the distance on the graph from the winning module determined in step 430. For example, the learning unit 170 performs the learning most strongly so that the module determined as the winning module among the plurality of modules 310 is close to the capacity differential characteristic dQ / dV of the training data, and performs the learning more weakly as the position on the graph is farther from the winning module.

[0072] In step 450, the learning unit 170 determines whether there is other training data. If it is determined that there is other training data, the learning unit 170 returns the process to step 430, uses the training data under the next SOH condition (for example, SOH = 69%) to determine the next winning module, and learns each module. In this way, the learning unit 170 repeatedly processes until the other training data disappears, that is, until each module is learned using the training data under SOH = 64%, 69%, 86%, 92%, and 100%.

[0073] When it is determined in step 450 that there is no other training data, the learning unit 170 associates the maximum capacity of the battery 10 with each condition in step 460. That is, the maximum capacity of the battery 10 is associated with each module 310. At this time, the learning unit 170 calculates the maximum capacity of the battery 10 associated with each condition based on the integral characteristic obtained by integrating the capacity differential characteristic dQ / dV using the voltage V of the battery 10. Specifically, the learning unit 170 can also calculate the maximum capacity of the battery 10 associated with each condition based on the following formula. Here, in (Formula 1), Q max represents the maximum capacity of the battery 10, Upper V represents the upper limit voltage that specifies the maximum capacity of the battery 10, and Lower V represents the lower limit voltage that specifies the maximum capacity of the battery 10. That is, the learning unit 170 calculates the area value in the range from the lower limit voltage to the upper limit voltage that specifies the maximum capacity of the battery 10 for the capacity differential characteristic dQ / dV of each module, and associates the area value with each module. Then, the learning unit 170 ends the flow processing of this figure.

[0074] [Mathematical formula 1]

[0075]

[0076] Figure 5 Indicates according to Figure 4The mnSOM that has been learned by the process of learning. As shown in this figure, for example, in the learning using the training data under SOH=100%, the module at the lower left end of the figure, that is, module (0, 9), is determined as the winning module, and is learned in a manner that the capacity differential characteristic dQ / dV of module (0, 9) is closest to the capacity differential characteristic dQ / dV under SOH=100%. In addition, for example, in the learning using the training data under SOH=64%, the module at the lower right end of the figure, that is, module (14, 9), is determined as the winning module, and is learned in a manner that the capacity differential characteristic dQ / dV of module (14, 9) is closest to the capacity differential characteristic dQ / dV under SOH=64%. Here, although the learning unit 170 uses only 5 SOH conditions as training data, as shown in this figure, 10×15=150 functions are obtained through module learning, and functions with similar characteristics are arranged close to each other in a self-organizing manner.

[0077] In this way, the learning unit 170 of the learning device 100 according to this embodiment follows Figure 4 According to the process, the estimated model 300 is generated using mnSOM, so that the module determined as the winning module can be trained with a teacher using the training data, and on the other hand, the entire network can be clustered in a self-organizing manner through untrained learning. Therefore, the learning unit 170 can generate an estimated model based on a relatively small amount of training data, and can reduce the time and cost involved in the preparation for generating the model. In addition, the learning unit 170 generates the estimated model based on the data in operation of the battery 10, so the model can be generated without using detailed information such as the physical property values ​​inherent to the electrode material of the battery 10 or the reaction principle of the battery.

[0078] Figure 6 The figure shows an example in which the learning device 100 according to the present embodiment associates the maximum capacity of the battery 10 with each module of the mnSOM. In this figure, the numerical value recorded in each module indicates the maximum capacity of the battery 10 associated with each module, and the figure shows the size of the maximum capacity by the shade of each module. The learning unit 170 associates the maximum capacity of the battery 10 with each module based on the above formula (Mathematical formula 1). Figure 5 The functions of the modules shown are integrated using the voltage V of the battery 10 , thereby calculating the maximum capacity of the battery 10 and associating it with each module.

[0079] As described above, the learning unit 170 of the learning device 100 according to the present embodiment associates the maximum capacity of the battery 10 with each module, and thus can provide an estimation model 300 that facilitates estimation of the maximum capacity of the battery 10 as described later.

[0080] In the above description, as an example, the case where the learning device 100 learns the capacity differential characteristic dQ / dV obtained by differentiating the capacity Q of the battery 10 with the voltage V to generate the estimation model is shown, but the present invention is not limited to this. The learning device 100 may also learn the voltage differential characteristic dV / dQ obtained by differentiating the voltage V of the battery 10 with the capacity Q to generate the estimation model. That is, the calculation unit 150 may calculate the voltage differential characteristic dV / dQ obtained by differentiating the voltage V of the battery 10 with the capacity Q as a characteristic corresponding to the capacity change and voltage change of the battery 10, and the learning unit 170 may learn each module 310 with the capacity Q of the battery 10 as an input vector and the voltage differential characteristic dV / dQ of the battery 10 as an output vector to generate the estimation model 300.

[0081] In the above description, as an example, the learning device 100 generates the estimation model 300 using data actually obtained from the battery 10 as training data, but the present invention is not limited thereto. The learning device 100 may generate the estimation model 300 using the output of a numerical simulation model that simulates the behavior of the battery 10 as training data.

[0082] Figure 7 2 shows an estimation device 700 and a battery 10 according to the present embodiment. Figure 7 In the Figure 1 Components having the same functions and structures are given the same reference numerals, and descriptions thereof will be omitted except for the following differences. The estimation device 700 according to the present embodiment acquires the estimation model 300 , and estimates at least one of the maximum capacity and the remaining capacity of the storage battery 10 using the acquired estimation model 300 .

[0083] The estimation device 700 may be configured as follows: Figure 1 The same device as the learning device 100 shown may also be configured as another device different from the learning device 100. The estimation device 700 may be a computer such as a PC (personal computer), a tablet computer, a smart phone, a workstation, a server computer, or a general-purpose computer, or a computer system obtained by connecting multiple computers. Such a computer system is also a computer in a broad sense. In addition, the estimation device 700 may also be implemented in a computer through one or more executable virtual computer environments. Alternatively, the estimation device 700 may be a dedicated computer designed for estimating at least one of the maximum capacity and the remaining capacity, or it may be dedicated hardware implemented by a dedicated circuit.

[0084] The estimation device 700 according to the present embodiment includes a voltage sensor 110, a temperature sensor 120, an A / D converter 130, a data acquisition unit 140, a calculation unit 150, a characteristic acquisition unit 160, a model acquisition unit 710, an estimation unit 720, and a result output unit 730. The voltage sensor 110, the temperature sensor 120, the A / D converter 130, the data acquisition unit 140, the calculation unit 150, and the characteristic acquisition unit 160 are described in detail below. Figure 1 The same, therefore the description is omitted.

[0085] The model acquisition unit 710 acquires an estimation model that estimates at least one of the maximum capacity and the remaining capacity of the battery 10 based on the relationship between any one of the voltage V and the capacity Q of the battery 10 under a plurality of conditions in at least one of the charging and discharging of the battery 10 and the characteristics corresponding to the capacity change and the voltage change of the battery 10. For example, the model acquisition unit 710 acquires an estimation model that estimates at least one of the maximum capacity and the remaining capacity of the battery 10 based on the relationship between any one of the voltage V and the capacity Q of the battery 10 under a plurality of conditions in at least one of the charging and discharging of the battery 10 and the characteristics corresponding to the capacity change and the voltage change of the battery 10. Figure 1 The mnSOM that has been learned and associated with the maximum capacity by the learning device 100 shown is acquired via a network or the like.

[0086] The estimation unit 720 estimates at least one of the maximum capacity and the remaining capacity of the battery 10 based on the voltage V and the capacity Q of the battery 10 in use and the characteristics corresponding to the calculated capacity change and voltage change of the battery 10, using the estimation model 300 obtained by the model acquisition unit 710. At this time, the estimation unit 720 may determine whether or not the maximum capacity and the remaining capacity of the battery 10 can be estimated based on the time series data of the voltage V and the current I of the battery 10 in use. In addition, the estimation unit 720 may estimate the SOH and SOC (State of Charge) of the battery 10 using the estimated maximum capacity and the remaining capacity of the battery 10. In addition, the so-called SOC is expressed as a percentage of the remaining capacity relative to the maximum capacity at the current time point of the battery 10. Various estimations in the estimation unit 720 will be described later. The estimation unit 720 supplies the estimated result to the result output unit 730.

[0087] The result output unit 730 supplies at least one of the maximum capacity, remaining capacity, SOH, and SOC of the battery 10 estimated by the estimation unit 720 to other functional units or other devices. The result output unit 730 may also output the estimation results estimated by the estimation unit 720 when there is an inquiry about the state of the battery 10 from the battery control system, for example.

[0088] Figure 8The following is a flow chart of the estimation device 700 of the present embodiment estimating the state of the battery 10. In step 810, the data acquisition unit 140 acquires the time series data of the voltage V and the current I of the battery 10 in use. In addition, the data acquisition unit 140 acquires the temperature T of the battery 10 together with the time series data of the voltage V and the current I of the battery 10.

[0089] In step 830, the estimation unit 720 determines whether the maximum capacity of the battery 10 can be estimated based on the time series data of the voltage V and the current I of the battery 10. The estimation unit 720 determines whether the maximum capacity of the battery 10 can be estimated, for example, by whether appropriate data for estimating the maximum capacity is accumulated, for example, whether a region with a large change relative to the SOH is accumulated. In addition, the estimation unit 720 determines whether the maximum capacity of the battery 10 can be estimated, for example, by whether a sufficient amount of data for estimating the maximum capacity is accumulated, for example, whether data of a capacity range corresponding to a predetermined SOC range (SOC: 20% equivalent, etc.) is accumulated. In addition, the estimation unit 720 determines whether the maximum capacity of the battery 10 can be estimated, for example, by whether a feature point that is advantageous for estimating the maximum capacity is captured, for example, whether a peak point of the capacity differential characteristic dQ / dV is captured. The estimation unit 720 may determine that the maximum capacity of the battery 10 can be estimated when all of these are satisfied. Alternatively, estimating unit 720 may determine that the maximum capacity of battery 10 can be estimated when at least one of the above conditions is satisfied.

[0090] If it is determined in step 830 that the maximum capacity can be estimated, in step 835, the calculation unit 150 calculates the maximum capacity in accordance with step 420. Figure 2 Then, the characteristic acquisition unit 160 stores the relationship between the voltage V and the capacity differential characteristic dQ / dV of the storage battery 10 in use in a database (not shown) for a fixed period of time.

[0091] Then, in step 840, the winning module is determined based on the relationship between the voltage V of the battery 10 and the capacity differential characteristic dQ / dV, and the winning module is updated. In addition, any module may be set as the winning module at the initial stage. The estimation unit 720 selects a graph of the mnSOM corresponding to the conditions in use from a plurality of graphs. For example, the estimation unit 720 selects a graph of the mnSOM corresponding to the temperature conditions and the charge and discharge rate conditions in use. Next, the estimation unit 720 inputs the vector of the accumulated voltage V to each module of the selected mnSOM, and obtains the output of the capacity differential characteristic dQ / dV value vector calculated by the neural network in each module. Next, the estimation unit 720 calculates the Euclidean distance between the calculated capacity differential characteristic dQ / dV value vector and the accumulated capacity differential characteristic dQ / dV value vector for each module. Then, the estimation unit 720 determines the module with the smallest Euclidean distance as the winning module.

[0092] In step 850, the estimation unit 720 estimates the maximum capacity of the battery 10 based on the integrated characteristic obtained by integrating the capacity differential characteristic dQ / dV in the module determined as the winning module with the voltage V of the battery 10. For example, in the estimation model 300 acquired by the model acquisition unit 710, when the maximum capacity of the battery 10 is previously associated with each module, the estimation unit 720 may call the maximum capacity associated with the module determined as the winning module and estimate it as the maximum capacity of the battery 10. In the case where the maximum capacity of the battery 10 is not associated with each module in the estimation model 300, the estimation unit 720 may calculate the maximum capacity based on the above-mentioned (Formula 1) and estimate it as the maximum capacity of the battery 10. The estimation device 700 stores the maximum capacity of the battery 10 estimated in this way as the latest maximum capacity estimated value. Thus, when there is an inquiry about the maximum capacity value from other functional units or other devices, the result output unit 730 can provide the maximum capacity estimated value.

[0093] In step 860, the estimation unit 720 estimates the SOH of the storage battery 10 using the maximum capacity estimation value estimated in step 850. The estimation unit 720 may estimate the SOH of the storage battery 10 by expressing, for example, a value obtained by dividing the maximum capacity estimation value by the initial maximum capacity, that is, the maximum capacity of SOH=100% as a percentage. The estimation device 700 stores the SOH of the storage battery 10 estimated in this way as the latest SOH estimation value. Thus, when there is an inquiry about the SOH from other functional units or other devices, the result output unit 730 can provide the SOH estimation value.

[0094] In addition, in the above description, as an example, the estimation device 700 estimates both the maximum capacity and the SOH of the battery 10, but the present invention is not limited thereto. When the estimation of the SOH is not necessary, the estimation device 700 may omit the processing of step 860. In addition, when the estimation of the maximum capacity and the SOH is not necessary, the estimation device 700 may omit the processing from step 830 to step 860. Thus, the estimation device 700 can omit the estimation of unnecessary states, and can reduce the calculation load.

[0095] In step 870, the estimation unit 720 determines whether the remaining capacity of the battery 10 can be estimated based on the time series data of the voltage V and the current I of the battery 10. The estimation unit 720 determines whether the remaining capacity of the battery 10 can be estimated, for example, by whether appropriate data for estimating the remaining capacity is accumulated, for example, whether a region with a large change relative to the SOC is accumulated. In addition, the estimation unit 720 determines whether the remaining capacity of the battery 10 can be estimated, for example, by whether a sufficient amount of data for estimating the remaining capacity is accumulated, for example, whether data of a capacity range corresponding to a predetermined SOC range (SOC: 20% equivalent, etc.) is accumulated. In addition, the estimation unit 720 determines whether the remaining capacity of the battery 10 can be estimated, for example, by whether a feature point that is advantageous for estimating the remaining capacity is captured, for example, whether a peak point of the capacity differential characteristic dQ / dV is captured. The estimation unit 720 may determine that the remaining capacity of the battery 10 can be estimated when all of these conditions are met. Alternatively, estimating unit 720 may determine that the remaining capacity of battery 10 can be estimated when at least one of the above conditions is satisfied.

[0096] If it is determined in step 870 that the remaining capacity can be estimated, the estimation device 700 performs data preprocessing in step 875, and the estimation unit 720 obtains the partial voltage differential characteristic dV / dQ calculated based on the time series data of the voltage V and current I of the battery 10 in use, which is obtained by differentiating the voltage V of the battery 10 with the capacity Q. At this time, for example, Figure 2Similarly to the left figure and the center figure of , the calculation unit 150 performs data smoothing and calculation of the capacity Q based on the time series data of the voltage V and the current I of the storage battery 10 in use, and then differentiates the voltage V with the capacity Q to convert it into a voltage differential characteristic dV / dQ [V / Ah] obtained by differentiating the voltage V of the storage battery 10 with the capacity Q, and supplies it to the estimation unit 720. Alternatively, for example, after it is determined in step 830 that the maximum capacity can be estimated, that is, in step 875 after the pre-processing of the data in step 835 is performed, the estimation unit 720 converts the capacity differential characteristic dQ / dV of the storage battery 10 obtained in step 835 into a voltage differential characteristic dV / dQ obtained by differentiating the voltage V of the storage battery 10 with the capacity Q, thereby obtaining a partial voltage differential characteristic dV / dQ.

[0097] Then, in step 880, the estimation unit 720 obtains the model voltage differential characteristic dV / dQ obtained by differentiating the voltage V of the battery 10 with the capacity Q based on the estimation model 300, and estimates the remaining capacity of the battery 10 using the model voltage differential characteristic dV / dQ as a reference. For example, the estimation unit 720 converts the capacity differential characteristic dQ / dV in the module determined as the winning module in step 840 into the model voltage differential characteristic dV / dQ obtained by differentiating the voltage V of the battery 10 with the capacity Q, and uses the model voltage differential characteristic dV / dQ as a reference.

[0098] Then, the estimation unit 720 estimates the remaining capacity of the battery 10 by fitting the partial voltage differential characteristic dV / dQ calculated based on the time series data of the voltage V and the current I of the battery 10 in use, which is obtained by differentiating the voltage V of the battery 10 with the capacity Q, with the model voltage differential characteristic dV / dQ as a reference. This will be described later. In addition, when it is determined in step 830 that the maximum capacity cannot be estimated, the estimation unit 720 may use the winning module at the previous maximum capacity estimation for the estimation of the remaining capacity. In addition, when the processing of steps 830 to 860 is omitted as described above, the estimation unit 720 may use the winning module at the previous maximum capacity estimation or the module pre-set as the winning module for the estimation of the remaining capacity. The estimation device 700 stores the remaining capacity of the battery 10 estimated in this way as the latest remaining capacity estimation value. Thus, when there is an inquiry regarding the remaining capacity value from other functional units or other devices, the result output unit 730 can provide the remaining capacity estimated value.

[0099] In step 890, the estimation unit 720 estimates the SOC of the battery 10 using the estimated remaining capacity value estimated in step 880. The estimation unit 720 may estimate the SOC of the battery 10 by expressing, for example, a value obtained by dividing the estimated remaining capacity value by the estimated maximum capacity value in percentage. The estimation device 700 stores the SOC of the battery 10 estimated in this way as the latest estimated SOC value. Thus, when there is an inquiry about the SOC from other functional units or other devices, the result output unit 730 can provide the estimated SOC value.

[0100] In addition, in the above description, as an example, the estimation device 700 estimates the remaining capacity and SOC of the battery 10, but the present invention is not limited to this. When the estimation of the SOC is not necessary, the estimation device 700 may omit the processing of step 890. In addition, when the estimation of the remaining capacity and SOC is not necessary, the estimation device 700 may omit the processing of steps 870 to 890. Thus, the estimation device 700 can omit the estimation of unnecessary states and reduce the calculation load.

[0101] Fig. 9 An example of curve fitting performed when the estimation device 700 involved in the present embodiment estimates the remaining capacity of the storage battery 10 is shown. The left side of this figure shows the model voltage differential characteristic dV / dQ and the partial voltage differential characteristic dV / dQ before curve fitting. In this figure, the model voltage differential characteristic dV / dQ is represented as a waveform 910. In addition, the partial voltage differential characteristic dV / dQ is represented as a waveform 920.

[0102] In this state, the estimation unit 720 fits the partial voltage differential characteristic dV / dQ with the model voltage differential characteristic dV / dQ as a reference in order to estimate the remaining capacity. Specifically, the estimation unit 720 shifts the waveform 920 representing the partial voltage differential characteristic dV / dQ in the horizontal axis direction (X-axis direction) so as to fit the waveform 910 representing the model voltage differential characteristic dV / dQ.

[0103] The right side of the figure shows the model voltage differential characteristic dV / dQ and the partial voltage differential characteristic dV / dQ after curve fitting. The estimation unit 720 estimates the remaining capacity of the battery 10 based on the partial voltage differential characteristic dV / dQ after curve fitting. For example, the estimation unit 720 estimates the value of the X-axis of the partial voltage differential characteristic dV / dQ after curve fitting (for example, the value of the X-axis at the left end of the waveform 920) as the remaining capacity of the battery 10.

[0104] In this way, the estimation device 700 in the present embodiment uses the acquired estimation model 300 to estimate at least one of the maximum capacity and the remaining capacity of the battery 10 based on the voltage of the battery 10 in use and the characteristics corresponding to the capacity change and the voltage change of the battery 10. As a result, the estimation device 700 according to the present embodiment can estimate at least one of the maximum capacity and the remaining capacity of the battery 10 without using detailed information such as the physical property values ​​inherent to the electrode material of the battery 10 and the reaction principle of the battery.

[0105] In the above description, as an example, the processing of the estimation device 700 is shown in the case where the estimation model 300 is a model generated by learning each module 310 using the capacity differential characteristic dQ / dV as an output vector, but the present invention is not limited to this. In the case where the estimation model 300 is a model generated by learning each module 310 using the voltage differential characteristic dV / dQ as an output vector, the estimation device 700 may determine the winning module of the battery 10 by using the capacity Q of the battery 10 as an input vector and the voltage differential characteristic dV / dQ of the battery 10 as an output vector, and estimate the maximum capacity and the remaining capacity.

[0106] Fig.10 An example of an estimation device 700 according to a variation of the present embodiment implemented by cloud computing is shown. In this figure, as an example, data on the operation of each storage battery 10 in a large number of small-scale virtual power plants dispersedly arranged in each household, such as a VPP (virtual power plant), is obtained, and the status of each storage battery 10 is provided to an aggregator that performs comprehensive control of the VPP.

[0107] In this figure, each home is provided with a storage battery 10, a power regulation subsystem 1010, a gateway 1020, and a home network 1030. The power regulation subsystem 1010 mainly has the following functions: converting the AC power of the system into DC power to charge the storage battery 10, converting the DC power of the storage battery 10 into AC power to supply the devices in the home, monitoring the state of the storage battery 10 and safely controlling the storage battery 10, and providing the information of the storage battery 10 to the upper level.

[0108] The gateway 1020 is a network node for connecting a computer network to a network of different protocols. The gateway 1020 is connected to the power conditioning subsystem 1010 via a home network 1030 .

[0109] The home network 1030 connects the power regulation subsystem 1010 to the gateway 1020. The representative communication protocol of the home network 1030 is ECHONET Lite (registered trademark). ECHONET Lite is a communication specification that IoT-izes energy-saving devices such as sensors, white appliances, and equipment machines to implement services such as energy management and remote maintenance. Therefore, the gateway 1020 supplies various data provided from the power regulation subsystem 1010 via such communication specifications to the cloud environment via the IP network. In addition, the gateway 1020 supplies various data obtained from the cloud environment via the IP network to the power regulation subsystem 1010 via the home network.

[0110] In the cloud environment, the estimation device 700 according to the present embodiment, a server computer 1050 , and a measurement database 1060 are installed.

[0111] The server computer 1050 provides various services to clients that can access the cloud environment. The server computer 1050 can be, for example, a cloud server in cloud computing or an edge server in edge computing. The server computer 1050 in this figure obtains various data including the voltage and current of the storage battery 10 of the power conditioning subsystem 1010 via the gateway 1020 and the IP network, and supplies the obtained various data to the measurement database 1060.

[0112] The measurement database 1060 accumulates various data supplied from the server computer 1050 , and supplies the accumulated various data to the estimation device 700 .

[0113] The data acquisition unit 140 of the estimation device 700 in this modification acquires various data required for estimating the maximum capacity and the remaining capacity of the storage battery 10 from the measurement database 1060. Then, the estimated state of each storage battery 10 is provided to the aggregator of the VPP, etc. In this way, in this modification, the estimation device 700 is realized by cloud computing, so that services related to the capacity estimation of the storage battery 10 can be realized in various environments that can access the cloud.

[0114] Various embodiments of the present invention may be described with reference to flow charts and block diagrams, where a block may represent (1) a stage of a process in which an operation is performed or (2) a portion of a device having the function of performing an operation. Specific stages and portions may be implemented by dedicated circuits, programmable circuits supplied with computer-readable instructions stored on a computer-readable medium, and / or processors supplied with computer-readable instructions stored on a computer-readable medium. Dedicated circuits may include digital and / or analog hardware circuits, and may also include integrated circuits (ICs) and / or discrete circuits. Programmable circuits may also include reconfigurable hardware circuits that include logical AND, logical OR, logical XOR, logical NAND, logical NOR, and other logical operations, flip-flops, registers, field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and other memory elements.

[0115] The computer-readable medium may also include any tangible device capable of storing commands executed by appropriate devices, and as a result, the computer-readable medium having the commands stored therein becomes an article of manufacture containing the following commands, wherein the commands are commands that can be executed to generate means for performing the operations specified in the flowchart or block diagram. As examples of computer-readable media, electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, etc. may also be included. As further specific examples of computer-readable media, Floppy (registered trademark) disks, floppy disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), electrically erasable programmable read-only memories (EEPROM), static random access memories (SRAM), compact disk read-only memories (CD-ROM), digital versatile disks (DVD), Blu-ray (RTM) disks, memory sticks, integrated circuit cards, etc. may also be included.

[0116] The computer-readable instructions may also include any source code or object code described by any combination of one or more programming languages, wherein the one or more programming languages ​​include assembly commands, instruction set architecture (ISA) commands, machine commands, machine-related commands, microcode, firmware commands, state setting data, or object-oriented programming languages ​​such as Smalltalk, JAVA (registered trademark), C++, and previous procedural programming languages ​​such as the "C" programming language or similar programming languages.

[0117] Regarding computer-readable instructions, a processor or programmable circuit provided to a general-purpose computer, a special-purpose computer, or other data processing device capable of executing a program locally or via a wide area network (WAN) such as a local area network (LAN) or the Internet executes the computer-readable instructions in order to generate a means for performing the operations specified in the flowchart or block diagram. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

[0118] Fig.11 An example of a computer 2200 in which multiple embodiments of the present invention can be embodied in whole or in part is shown. The program installed in the computer 2200 can make the computer 2200 function as an operation associated with the device involved in the embodiment of the present invention or one or more parts of the device, or can make it perform the operation or the one or more parts, and / or can make the computer 2200 perform the process involved in the embodiment of the present invention or the stage of the process. Such a program is executed by the CPU 2212 in order to make the computer 2200 perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described in this specification.

[0119] The computer 2200 according to the present embodiment includes a CPU 2212, a RAM 2214, a graphics controller 2216, and a display device 2218, which are connected to each other via a host controller 2210. The computer 2200 further includes input / output units such as a communication interface 2222, a hard disk drive 2224, a DVD-ROM drive 2226, and an IC card drive, which are connected to the host controller 2210 via an input / output controller 2220. The computer further includes a conventional input / output unit such as a ROM 2230 and a keyboard 2242, which are connected to the input / output controller 2220 via an input / output chip 2240.

[0120] The CPU 2212 controls each unit by operating according to the program stored in the ROM 2230 and the RAM 2214. The graphic controller 2216 acquires image data generated by the CPU 2212 through a frame buffer provided in the RAM 2214 or in the graphic controller 2216 itself, and displays the image data on the display device 2218.

[0121] The communication interface 2222 communicates with other electronic devices via a network. The hard disk drive 2224 stores programs and data used by the CPU 2212 in the computer 2200. The DVD-ROM drive 2226 reads programs or data from the DVD-ROM 2201 and provides the programs or data to the hard disk drive 2224 via the RAM 2214. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0122] The ROM 2230 stores therein a boot program etc. executed by the computer 2200 when activated, and / or a program depending on the hardware of the computer 2200. The input / output chip 2240 can also connect various input / output units to the input / output controller 2220 via a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0123] The program is provided via a computer-readable medium such as a DVD-ROM 2201 or an IC card. The program is read from the computer-readable medium and installed to a hard disk drive 2224, a RAM 2214, or a ROM 2230, which are also examples of computer-readable media, and executed by a CPU 2212. The information processing described in these programs is read into the computer 2200, bringing about a connection between the program and the various types of hardware resources described above. The device or method can also be constructed by implementing the operation or processing of information according to the use of the computer 2200.

[0124] For example, when communication is performed between the computer 2200 and an external device, the CPU 2212 may also execute a communication program loaded into the RAM 2214, and based on the process described in the communication program, issue a communication processing command to the communication interface 2222. Under the control of the CPU 2212, the communication interface 2222 reads the transmission data stored in the transmission buffer processing area provided in the RAM 2214, the hard disk drive 2224, the DVD-ROM 2201, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes the reception data received from the network to the reception buffer processing area provided on the recording medium.

[0125] In addition, the CPU 2212 can also read all or a necessary part of a file or database stored in an external recording medium such as the hard disk drive 2224, the DVD-ROM drive 2226 (DVD-ROM 2201), or an IC card into the RAM 2214, and perform various types of processing on the data on the RAM 2214. The CPU 2212 then writes the processed data back to the external recording medium.

[0126] Various types of information such as various types of programs, data, tables, and databases are stored in the recording medium and receive information processing. CPU2212 can also perform various types of processing recorded in various places of this disclosure, including various types of operations specified by the command sequence of the program, information processing, conditional judgment, conditional branching, unconditional branching, information retrieval / replacement, etc., for the data read from RAM2214, and write the results back to RAM2214. In addition, CPU2212 can also retrieve information in files, databases, etc. in the recording medium. For example, in the case where multiple entries with attribute values ​​of a first attribute, each of which is associated with an attribute value of a second attribute, are stored in the recording medium, CPU2212 can also retrieve the entry specified by the attribute value of the first attribute from the multiple entries that is consistent with the condition, and read the attribute value of the second attribute stored in the entry, thereby reading the attribute value of the second attribute associated with the first attribute that meets the predetermined condition.

[0127] The program or software module described above may also be stored in a computer-readable medium on or near the computer 2200. In addition, a recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can be used as a computer-readable medium, thereby providing the program to the computer 2200 via the network.

[0128] The present invention has been described above using the embodiments, however, the technical scope of the present invention is not limited to the scope described in the above embodiments. It is obvious to those skilled in the art that various changes or improvements can be added to the above embodiments. The various changes or improvements added can also be included in the technical scope of the present invention, which is clearly derived from the description of the claims.

[0129] Regarding the execution order of the actions, processes, steps, and stages of the devices, systems, programs, and methods shown in the claims, specifications, and drawings, it should be noted that as long as it is not explicitly stated as "first", "before", etc., or the output of the previous process is not applied to the next process, it can be implemented in any order. Regarding the action flow in the claims, specifications, and drawings, even if it is explained using "first", "second", etc. for convenience, it does not mean that it must be implemented in this order.

[0130] Description of symbols

[0131] 10. Battery

[0132] 12 Battery Cells

[0133] 14 Current sensor

[0134] 100 Learning Devices

[0135] 110 Voltage sensor

[0136] 120 Temperature Sensor

[0137] 130 A / D converter

[0138] 140 Data Acquisition Department

[0139] 150 Computing Department

[0140] 160 Characteristic Acquisition Department

[0141] 170 Learning Department

[0142] 180 Model output unit

[0143] 300 Estimated Model

[0144] 310 Module

[0145] 700 Prediction Device

[0146] 710 Model Acquisition Department

[0147] 720 Presumption Department

[0148] 730 Result Output Unit

[0149] 1010 Power Regulation Subsystem

[0150] 1020 Gateway

[0151] 1030 Home Network

[0152] 1050 Server Computer

[0153] 1060 Measurement Database

[0154] 2200 Computer

[0155] 2201 DVD-ROM

[0156] 2210 Host Controller

[0157] 2212 CPU

[0158] 2214 RAM

[0159] 2216 Graphics Controller

[0160] 2218 Display Devices

[0161] 2220 Input / Output Controller

[0162] 2222 Communication Interface

[0163] 2224 Hard Drive

[0164] 2226 DVD-ROM drive

[0165] 2230 ROM

[0166] 2240 Input / Output Chip

[0167] 2242 Keyboard.

Claims

1. A learning device comprising: a characteristic acquisition unit that acquires any one of a voltage and a capacity of the storage battery under a plurality of conditions in at least one of charging and discharging of the storage battery, and a characteristic corresponding to a change in the capacity and a change in the voltage of the storage battery; and a learning unit that learns the relationship between the characteristics and any one of the voltage and capacity of the battery under the plurality of conditions, and generates an estimation model that estimates at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics; The learning unit generates the estimation model using a modular network type self-organizing graph using a neural network as a module. The self-organizing graph is composed of a plurality of regularly arranged cells, each of which has a reference vector of the same dimension as the input vector. The modular network type self-organizing graph replaces the unit having the reference vector in the self-organizing graph with a module as a neural network, The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage or a voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity. The inference model has multiple modules. The learning unit sets the voltage V of the battery as an input vector and the capacity differential characteristic dQ / dV of the battery as an output vector, or sets the capacity Q of the battery as an input vector and the voltage differential characteristic dV / dQ of the battery as an output vector, and learns each module through a neural network to generate the estimated model.

2. The learning device according to claim 1, wherein: The learning unit generates the estimation model that interpolates the relationship under different conditions based on the relationship under at least one condition among the plurality of conditions.

3. The learning device according to claim 1 or 2, wherein: The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with a voltage.

4. The learning device according to claim 3, wherein: The learning unit associates the maximum capacity of the storage battery with each of the conditions.

5. The learning device according to claim 4, wherein: The learning unit calculates the maximum capacity of the storage battery associated with each condition based on an integrated characteristic obtained by integrating the capacity differential characteristic with a voltage of the storage battery.

6. The learning device according to claim 1 or 2, wherein: The characteristic is a voltage differential characteristic obtained by differentiating the voltage of the battery by capacity.

7. An estimation device comprising: a model acquisition unit that acquires an estimation model that estimates at least one of a maximum capacity and a remaining capacity of the battery based on a relationship between any one of a voltage and a capacity of the battery under a plurality of conditions in at least one of charging and discharging of the battery and characteristics corresponding to a capacity change and a voltage change of the battery; a data acquisition unit that acquires time series data of the voltage and current of the storage battery; a calculation unit that calculates the characteristic based on time series data of the voltage and current of the storage battery; as well as an estimating unit that estimates at least one of a maximum capacity and a remaining capacity of the battery based on either one of a voltage and a capacity of the battery and the calculated characteristic using the estimation model, The estimation model is an estimation model learned using a modular network type self-organizing graph using a neural network as a module. The self-organizing graph is composed of a plurality of regularly arranged cells, each of which has a reference vector of the same dimension as the input vector. The modular network type self-organizing graph replaces the unit having the reference vector in the self-organizing graph with a module as a neural network, The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage or a voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity. The inference model has multiple modules. The estimated model is generated by setting the voltage V of the battery as an input vector and the capacity differential characteristic dQ / dV of the battery as an output vector, or setting the capacity Q of the battery as an input vector and the voltage differential characteristic dV / dQ of the battery as an output vector, and learning each module through a neural network.

8. The estimation device according to claim 7, wherein: The estimated model is an estimated model in which the relationship under a different condition is interpolated based on the relationship under at least one condition among the plurality of conditions.

9. The estimation device according to claim 7 or 8, wherein: The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with a voltage.

10. The estimation device according to claim 9, wherein: The estimating unit estimates the maximum capacity of the storage battery based on an integrated characteristic obtained by integrating the capacity differential characteristic with a voltage of the storage battery.

11. The estimation device according to claim 7 or 8, wherein: The characteristic is a voltage differential characteristic obtained by differentiating the voltage of the battery by capacity.

12. The estimation device according to claim 9, wherein: The estimating unit acquires a model voltage differential characteristic obtained by differentiating a voltage of the storage battery with a capacity based on the estimation model, and estimates a remaining capacity of the storage battery using the model voltage differential characteristic as a reference.

13. The estimation device according to claim 12, wherein: The estimating unit estimates the remaining capacity of the battery by fitting a partial voltage differential characteristic calculated based on time series data of the voltage and current of the battery and obtained by differentiating the voltage of the battery with the capacity, using the model voltage differential characteristic as a reference.

14. The estimation device according to claim 7 or 8, wherein: The estimating unit determines whether or not at least one of a maximum capacity and a remaining capacity of the storage battery can be estimated based on time series data of a voltage and a current of the storage battery.

15. A learning method is a learning method performed by a learning device, comprising: acquiring any one of a voltage and a capacity of the storage battery under a plurality of conditions in at least one of charging and discharging of the storage battery, and a characteristic corresponding to a change in capacity and a change in voltage of the storage battery; and learning the relationship between the characteristics and any one of the voltage and capacity of the battery under the plurality of conditions, respectively, and generating an estimation model, wherein the estimation model estimates at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics, The inference model is generated using a modular network-type self-organizing graph using a neural network as a module, The self-organizing graph is composed of a plurality of regularly arranged cells, each of which has a reference vector of the same dimension as the input vector. The modular network type self-organizing graph replaces the unit having the reference vector in the self-organizing graph with a module as a neural network, The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage or a voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity. The inference model has multiple modules. The voltage V of the battery is set as an input vector, and the capacity differential characteristic dQ / dV of the battery is set as an output vector. Alternatively, the capacity Q of the battery is set as an input vector, and the voltage differential characteristic dV / dQ of the battery is set as an output vector. Each module is learned through a neural network to generate the estimated model.

16. A method of inference, which is a method of inference performed by an inference device, obtaining an estimation model for estimating at least one of a maximum capacity and a remaining capacity of the battery based on a relationship between any one of a voltage and a capacity of the battery under a plurality of conditions in at least one of charging and discharging of the battery and characteristics corresponding to a capacity change and a voltage change of the battery; Obtaining time series data of the voltage and current of the battery; calculating the characteristic based on time series data of the voltage and current of the battery; as well as using the estimation model, estimating at least one of a maximum capacity and a remaining capacity of the battery based on either one of a voltage and a capacity of the battery and the calculated characteristic, The estimation model is an estimation model learned using a modular network type self-organizing graph using a neural network as a module. The self-organizing graph is composed of a plurality of regularly arranged cells, each of which has a reference vector of the same dimension as the input vector. The modular network type self-organizing graph replaces the unit having the reference vector in the self-organizing graph with a module as a neural network, The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage or a voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity. The inference model has multiple modules. The estimated model is generated by setting the voltage V of the battery as an input vector and the capacity differential characteristic dQ / dV of the battery as an output vector, or setting the capacity Q of the battery as an input vector and the voltage differential characteristic dV / dQ of the battery as an output vector, and learning each module through a neural network.

17. A learning program product, comprising a computer program, wherein the computer program is executed by a computer to cause the computer to function as: a characteristic acquisition unit that acquires any one of a voltage and a capacity of the storage battery under a plurality of conditions in at least one of charging and discharging of the storage battery, and a characteristic corresponding to a change in the capacity and a change in the voltage of the storage battery; and a learning unit that learns the relationship between the characteristics and any one of the voltage and capacity of the battery under the plurality of conditions, and generates an estimation model that estimates at least one of the maximum capacity and the remaining capacity of the battery based on any one of the voltage and capacity of the battery and the characteristics; The learning unit generates the estimation model using a modular network type self-organizing graph using a neural network as a module. The self-organizing graph is composed of a plurality of regularly arranged cells, each of which has a reference vector of the same dimension as the input vector. The modular network type self-organizing graph replaces the unit having the reference vector in the self-organizing graph with a module as a neural network, The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage or a voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity. The inference model has multiple modules. The learning unit sets the voltage V of the battery as an input vector and the capacity differential characteristic dQ / dV of the battery as an output vector, or sets the capacity Q of the battery as an input vector and the voltage differential characteristic dV / dQ of the battery as an output vector, and learns each module through a neural network to generate the estimated model.

18. A presumption program product, comprising a computer program, wherein the computer program is executed by a computer to cause the computer to function as: a model acquisition unit that acquires an estimation model that estimates at least one of a maximum capacity and a remaining capacity of the battery based on a relationship between any one of a voltage and a capacity of the battery under a plurality of conditions in at least one of charging and discharging of the battery and characteristics corresponding to a capacity change and a voltage change of the battery; a data acquisition unit that acquires time series data of the voltage and current of the storage battery; a calculation unit that calculates the characteristic based on time series data of the voltage and current of the storage battery; as well as an estimating unit that estimates at least one of a maximum capacity and a remaining capacity of the battery based on either one of a voltage and a capacity of the battery and the calculated characteristic using the estimation model, The estimation model is an estimation model learned using a modular network type self-organizing graph using a neural network as a module. The self-organizing graph is composed of a plurality of regularly arranged cells, each of which has a reference vector of the same dimension as the input vector. The modular network type self-organizing graph replaces the unit having the reference vector in the self-organizing graph with a module as a neural network, The characteristic is a capacity differential characteristic obtained by differentiating the capacity of the storage battery with the voltage or a voltage differential characteristic obtained by differentiating the voltage of the storage battery with the capacity. The inference model has multiple modules. The estimated model is generated by setting the voltage V of the battery as an input vector and the capacity differential characteristic dQ / dV of the battery as an output vector, or setting the capacity Q of the battery as an input vector and the voltage differential characteristic dV / dQ of the battery as an output vector, and learning each module through a neural network.

Images