Remaining capacity estimation apparatus, model generation apparatus, and non-transitory computer-readable medium

Patent metadata

Field	Value
Publication number	US-12443152-B2
Application number	US-202117996277-A
Country	US
Kind code	B2
Filing date	Apr 12, 2021
Priority date	Apr 17, 2020
Publication date	Oct 14, 2025
Grant date	Oct 14, 2025

Abstract

A remaining capacity estimation apparatus includes a storage processing unit and a calculation unit. The storage processing unit acquires a model from a model generation apparatus and stores the model in a model storage unit. When data for updating the model are acquired from the model generation apparatus, the storage processing unit updates the model stored in the model storage unit. The calculation unit calculates a remaining capacity of a storage battery managed by the remaining capacity estimation apparatus by using the model stored in the model storage unit. At this time, data (measurement data for calculation) input to the model include a current, a voltage, and a temperature of the storage battery. When the input data when generating the model are only a current, a voltage, and a temperature, the measurement data for calculation are only a current, a voltage, and a temperature.

First claim

The invention claimed is:

1. A model generation apparatus comprising:
at least one memory storing instructions; and
at least one processor configured to execute the instructions to perform operations comprising:
acquiring a plurality of pieces of training data using, as input values, measurement data for training including a current, a voltage, and a temperature of a storage battery and using, as a target value, a remaining capacity value for training being a remaining capacity value of the storage battery; and
generating a model for calculating a remaining capacity of an object storage battery from measurement data for calculation including a current, a voltage, and a temperature of the object storage battery, the object storage battery being an object of processing, by performing machine learning on the plurality of pieces of training data, wherein the operations further comprise:
processing n sets of the training data into an object matrix of 3×n;
performing processing on the object matrix, thereby generating one-dimensional data consisting of L pieces of data;
generating the model using the one-dimensional data as an input value; and
generating the one-dimensional data by performing, at least once, conversion processing of, while expanding one of rows or columns of the object matrix to the number of dimensions larger than the number of dimensions of the one, the one being associated to the current, the voltage, and the temperature, reducing the other of the rows or the columns of the object matrix to the number of dimensions smaller than the number of dimensions of the other by processing a digital filter on the object matrix.

2. A model generation apparatus comprising:
at least one memory storing instructions; and
at least one processor configured to execute the instructions to perform operations comprising:
acquiring a plurality of pieces of training data using, as input values, measurement data for training including a current, a voltage, and a temperature of a storage battery and using, as a target value, a remaining capacity value for training being a remaining capacity value of the storage battery;
processing n sets of the training data into an object matrix of m×n (where m is the number of types of data included in measurement data for training) and performing processing on the object matrix, thereby generating one-dimensional data consisting of z pieces of data; and
generating a model for calculating a remaining capacity of an object storage battery from measurement data for calculation including a current, a voltage, and a temperature of the object storage battery, the object storage battery being an object of processing, by performing machine learning using the one-dimensional data as an input value, wherein the operations comprise:
generating the one-dimensional data by performing, at least once, conversion processing of, while expanding one of rows or columns of the object matrix to the number of dimensions larger than the number of dimensions of the one, the one being associated to the current, the voltage, and the temperature, reducing the other of the rows or the columns of the object matrix to the number of dimensions smaller than the number of dimensions of the other by processing a digital filter on the object matrix.

3. The model generation apparatus according to claim 1 , wherein
the digital filter is a matrix, and
the operations comprise performing following (1) and (2) at least once as the conversion processing:
(1) cutting out a partial matrix consisting of the same number of rows and the same number of columns as the digital filter, from the object matrix; and
(2) setting a value acquired by operating the digital filter on the partial matrix and adding each element of a result of the operation, as an element of the object matrix after processing.

4. The model generation apparatus according to claim 3 , wherein
the operations comprise performing processing of expanding at least one of rows or columns of the object matrix by adding a dummy value to an outer periphery of the object matrix before the (1) in the conversion processing.

5. The model generation apparatus according to claim 1 , wherein
the operations comprise processing n number of the remaining capacity values for training being associated to the n sets of training data, thereby generating the target value consisting of k pieces of data (where k<n).

6. The model generation apparatus according to claim 1 , wherein
the machine learning is performed by using a neural network.

7. The model generation apparatus according to claim 1 , wherein
actual result data including measurement values of a current, a voltage, a temperature, and a remaining capacity value of a storage battery are prepared,
the training data are generated by using a part of the actual result data, and
the operations comprise verifying the model by using at least one piece of the remaining actual result data.

8. The model generation apparatus according to claim 1 , wherein the operations comprise:
acquiring the training data for each type of the storage battery; and
generating the model for each type of the storage battery.

9. A non-transitory computer-readable medium storing a program causing a computer to perform operations comprising:
acquiring a plurality of pieces of training data using, as input values, measurement data for training including a current, a voltage, and a temperature of a storage battery and using, as a target value, a remaining capacity value for training being a remaining capacity value of the storage battery; and
generating a model for calculating a remaining capacity of an object storage battery from measurement data for calculation including a current, a voltage, and a temperature of the object storage battery by performing machine learning on the training data, the object storage battery being an object of processing, wherein the operations further comprise:
processing n sets of the training data into an object matrix of 3×n;
performing processing on the object matrix, thereby generating one-dimensional data consisting of L pieces of data;
generating the model using the one-dimensional data as an input value; and
generating the one-dimensional data by performing, at least once, conversion processing of, while expanding one of rows or columns of the object matrix to the number of dimensions larger than the number of dimensions of the one, the one being associated to the current, the voltage, and the temperature, reducing the other of the rows or the columns of the object matrix to the number of dimensions smaller than the number of dimensions of the other by processing a digital filter on the object matrix.

10. A non-transitory computer-readable medium storing a program causing a computer to perform operations comprising:
acquiring a plurality of pieces of training data using, as input values, measurement data for training including a current, a voltage, and a temperature of a storage battery and using, as a target value, a remaining capacity value for training being a remaining capacity value of the storage battery;
processing n sets of the training data into an object matrix of m×n and performing processing on the object matrix, thereby generating one-dimensional data consisting of z pieces of data; and
generating a model for calculating a remaining capacity of an object storage battery from measurement data for calculation including a current, a voltage, and a temperature of the object storage battery, the object storage battery being an object of processing, by performing machine learning using the one-dimensional data as an input value, wherein the operations comprise:
generating the one-dimens

Assignees

• Aesc Japan Ltd

Inventors

• Shimawaki Hidenori
• Chen Jiuting

Classifications

• H02J7/82 Primary

  Control of state of charge [SOC] · CPC title

• H02J7/80

  including monitoring or indicating arrangements · CPC title

• G01R31/367

  Software therefor, e.g. for battery testing using modelling or look-up tables · CPC title

• G01R31/387

  Determining ampere-hour charge capacity or SoC · CPC title

• G06N3/09

  Supervised learning · CPC title