Estimation system and estimation method

Patent metadata

Field	Value
Publication number	US-12289184-B2
Application number	US-202218278062-A
Country	US
Kind code	B2
Filing date	Mar 8, 2022
Priority date	May 19, 2021
Publication date	Apr 29, 2025
Grant date	Apr 29, 2025

Abstract

An estimation system generates, from a value of a time series of an estimation object in a past period, a plurality of patterns of a transition of a value of the estimation object. Based on the plurality of generated patterns and a value of a time series of a factor in the past period, the estimation system specifies a dependency relationship between a transition pattern and a value of the factor and a transition pattern at a past (or future) time point and identifies a model in accordance with the specified dependency relationship. By inputting a value of a time series of the factor in a future period to the estimation model, the estimation system specifies a time series of a value in the future period of the estimation object using at least one transition pattern.

First claim

The invention claimed is:

1. An estimation system, comprising:
a first computer including a first network interface and a storage device, the storage device storing sample estimation object data, sample factor data, and estimation factor data, the sample estimation data being time series data of electric power consumption, and the sample factor data being time series data of weather values which affect the time series data of electric power consumption, the estimation factor data including, for each of the one or more factors, a data set representing a value of a time series of the factor in a future period;
a second computer including a second network interface coupled to the first network interface via a network; and
an electric power generator that consumes fuel,
wherein the second computer is programmed to:
generate, from the sample estimation object data, a plurality of estimation object transition patterns, each of which being a pattern of a transition of a value of the estimation object,
specify a pattern factor dependency relationship which is a dependency relationship between (X) and (Y) based on the sample factor data and the plurality of estimation object transition patterns, (X) being an estimation object transition pattern, and (Y) being a value of the one or more factors and an estimation object transition pattern at a past or future time point,
identify an estimation model which is a model in accordance with the pattern factor dependency relationship and which is a model that accepts the value of the one or more factors and the estimation object transition pattern at a past or future time point as input and which produces an estimation object transition pattern as output, and
generate, by inputting the estimation factor data to the estimation model, estimation result data which is data including one or more data sets representing a time series of a value in the future period of the estimation object using at least one of the plurality of estimation object transition patterns,
determine, based on the estimation result data, an amount of fuel for the generator which corresponds to the future period, as part of an operation plan of the generator, and
transmit a command to control the generator based on the operation plan to operate the generator,
wherein a period of transition represented by each of the plurality of estimation object transition patterns is shorter than the future period and is a time period of minutes, hours, days, or weeks.

2. The estimation system according to claim 1 ,
wherein the second computer is programmed to generate, by inputting at least the estimation factor data among the sample factor data and the estimation factor data to the estimation model, estimation result data which is data including one or more data sets representing a time series of a value in a predetermined period including the future period of the estimation object, and
wherein the one or more data sets included in the estimation result data is a data set of which a difference from a time series of an observed value with respect to an estimation object is equal to or smaller than a threshold among one or a plurality of data sets generated by the processor using the estimation model.

3. The estimation system according to claim 1 ,
wherein with respect to each of the one or more data sets included in the estimation result data, a difference between a value in the data set and a predicted data set including one or more values predicted by a predetermined method regarding a part of time points in the future period is equal to or smaller than a threshold.

4. The estimation system according to claim 1 ,
wherein when at least one data set in the estimation result data has a first difference that is a difference between a value in the data set and a value predicted by a predetermined method with respect to at least a part of time points in the future period, the processor performs a change for reducing the first difference as a change to the identified estimation model.

5. The estimation system according to claim 4 ,
wherein when at least one data set in the estimation result data has a second difference that is a difference between a value in the data set and a value specified from the sample estimation object data instead of or in addition to the first difference with respect to at least a part of time points in the future period, a change for reducing the second difference instead of or in addition to the first difference as a change to the identified estimation model is performed.

6. The estimation system according to claim 1 ,
wherein the storage device stores sample element data and estimation element data,
wherein the sample element data is data including, for each of one or more elements defined to be potentially capable of affecting a value of the at least one factor, a data set representing a value of a time series of the element in the past period,
wherein the estimation element data is data including, for each of the one or more elements, a data set representing a value of a time series of the element in the future period,
wherein the second computer is programmed to:
generate, with respect to at least one factor among the one or more factors, a plurality of factor transition patterns, each of which being a pattern of a transition of a factor value, from a data set in the sample factor data,
specify a factor element dependency relationship which is a dependency relationship between (x) and (y) based on the sample element data and the plurality of factor transition patterns, (x) being a factor transition pattern, and (y) being a value of the one or more elements and a factor transition pattern at a past or future time point,
identify a factor model which is a model in accordance with the factor element dependency relationship and which is a model that accepts the value of the one or more elements and the factor transition pattern in the past or the future as input and which produces a factor transition pattern as output, and
output, by inputting the estimation element data to the factor model, the estimation factor data that is data including one or more data sets representing a time series of a value in the future period of the at least one factor using at least one of the plurality of factor transition patterns.

7. An estimation method of an estimation system, the estimation system including a first computer including a first network interface and a storage device, the storage device storing sample estimation object data, sample factor data, and estimation factor data, the sample estimation data being time series data of electric power consumption, and the sample factor data being time series data of weather values which affect the time series data of electric power consumption, the estimation factor data including, for each of the one or more factors, a data set representing a value of a time series of the factor in a future period; a second computer including a second network interface coupled to the first network interface via a network; and an electric power generator that consumes fuel, the method comprising:
generating, from sample estimation object data that is data representing a value of a time series of an estimation object in a past period, a plurality of estimation object transition patterns, each of which being a pattern of a transition of a value of the estimation object;
specifying a pattern factor dependency relationship which is a dependency relationship between (X) and (Y) based on sample factor data and the plurality of estimation object transition patterns, the sample factor data being data including, for each of one or more factors defined to be potentially capable of affecting a value of the estimation object, a data set repre

Assignees

• Hitachi Ltd

Inventors

• Utsumi Masato
• Watanabe Tohru
• Namba Kazuki
• Shigemori Ikuo
• Iimura Hiroshi
• Ogawa Hiroaki
• Hamaba Daisuke
• Yamazaki Jun

Classifications

• G06F18/241

  relating to the classification model, e.g. parametric or non-parametric approaches · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06Q10/04

  Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202) · CPC title