Computing device, calculation method, and program

Abstract

The system utilizes new sensors and existing trained models to perform diagnostic and other computational processing appropriately. [Solution] The conversion unit obtains pseudo-first time series data by inputting second time series data measured by the second sensor into a conversion model that converts the input time series data into pseudo-first time series data that resembles the first time series data measured by the first sensor. The conversion model is a generative adversarial network conversion model in which an identification model that identifies whether or not the input time series data is the first time series data measured by the first sensor and the conversion model are trained adversarially against each other.

Claims

1. A generative adversarial network (AGAD) is trained adversarially on an identification model that identifies whether the input time series data is first time series data measured by a first sensor, and on a conversion model that converts the input time series data into pseudo-first time series data that resembles the first time series data. The AGAD unit obtains pseudo-first time series data by inputting second time series data measured by a second sensor that measures the same type of state quantity as the first sensor into the conversion model of the AGAD. Equipped with, computing device.

2. The system includes a computing unit that performs computational processing using a pre-trained computational model trained with a training dataset that uses the first time-series data measured by the first sensor as input samples. The calculation unit performs the calculation by inputting the pseudo-first time series data into the calculation model. The computing device according to claim 1.

3. The first sensor and the second sensor measure waves generated by partial discharge of power equipment. The calculation model outputs a value related to partial discharge to the power equipment based on the one-time series data. The computing device according to claim 2.

4. The system includes a display unit that displays the second time series data and the pseudo-first time series data obtained by converting the second time series data. The computing device according to claim 1.

5. The system includes a display unit that displays the feature quantities of the second time series data and the feature quantities of the pseudo-first time series data obtained by transforming the second time series data. The computing device according to claim 1.

6. The system includes a correction unit that corrects the second time-series data according to the difference in characteristics between the first sensor and the second sensor. The conversion unit obtains the pseudo-first time series data from the corrected second time series data. The computing device according to claim 1.

7. Computers The steps include: obtaining pseudo-first time series data by inputting second time series data measured by a second sensor measuring the same type of state quantity as the first sensor into the transformation model of a generative adversarial network, which is trained adversarially with a recognition model that identifies whether the input time series data is first time series data measured by a first sensor and a transformation model that transforms the input time series data into pseudo-first time series data that resembles the first time series data; and obtaining pseudo-first time series data by inputting second time series data measured by a second sensor that measures the same type of state quantity as the first sensor into the transformation model of the generative adversarial network. A calculation method that includes the following features.

8. On the computer, The steps include: obtaining pseudo-first time series data by inputting second time series data measured by a second sensor measuring the same type of state quantity as the first sensor into the transformation model of a generative adversarial network, which is trained adversarially with a recognition model that identifies whether the input time series data is first time series data measured by a first sensor and a transformation model that transforms the input time series data into pseudo-first time series data that resembles the first time series data; and obtaining pseudo-first time series data by inputting second time series data measured by a second sensor that measures the same type of state quantity as the first sensor into the transformation model of the generative adversarial network. A program to execute.

Images