US-12620796-B2 - Data analytics for smart electrical protection systems

US12620796B2US 12620796 B2US12620796 B2US 12620796B2US-12620796-B2

Abstract

A system and method for protecting an electric power distribution system integrated with a hybrid machine-learning model includes a plurality of field protection systems, a plurality of electrical switching circuits, and a local area network. Data from each of the plurality of field protection systems that is connected to the electrical switching circuits that are configured for protecting the electric power distribution system, is transmitted to the cloud. The received data is processed to identify a change in patterns and compute an error related to the HMLM in comparison with the processed data. The HMLM that is implemented in the plurality of FPS is calibrated to minimize the error and send updates to develop a new decision making firmware thereby to control actuation of the plurality of the electrical switching circuits.

Inventors

Mayukha PAL
Alok Kumar Bharati
Satish Shamsundar Belkhode

Assignees

ABB SCHWEIZ AG

Dates

Publication Date: 20260505
Application Date: 20221109
Priority Date: 20200514

Claims (11)

1 . A system for protecting an electric power distribution system integrated with a hybrid machine-learning model comprises: a cloud or central server comprising one or more servers connected to a plurality of field protection systems (FPSs) through a communication network; wherein the plurality of field protection systems is connected to a plurality of electrical switching circuits that are configured for protecting the electric power distribution system; wherein the plurality of field protection systems is configured to transmit, continuously or periodically, data from each of the plurality of field protection systems to the plurality of field protection systems that is connected in a same local area network (LAN) or one or more servers of the central server; wherein the data includes time stamped processed electrical and non-electrical parameters and signals; wherein the one or more servers comprises a processor and a memory; and wherein the processor is configured to: process the received data to identify change in patterns of the data that is time stamped and extract data that is changed in patterns from a predefined list of data provided in the HMLM; perform analysis on the processed data using the HMLM by computing an error related to the HMLM in comparison with the processed data; calibrate the HMLM to minimize the error; update a decision-making firmware or weighted output parameters of the plurality of field protection systems from the calibrated hybrid machine-learning model to develop a new decision making firmware or the weighted output parameters; and send the update regarding the new decision making firmware or the weighted output to the plurality of field protection systems to control actuation of each of the connected plurality of the electrical switching circuits.
2 . The system as claimed in claim 1 , wherein the plurality of field protection systems is further configured to: receive a new decision making firmware or weighted output parameters for the plurality of field protection systems from another plurality of field protection systems that is connected through the same LAN to calibrate the HMLM; and utilize the new decision making firmware or the weighted output parameters to update the HMLM to make future decisions in the FPS.
3 . The system as claimed in claim 1 , wherein the time stamped processed electrical and non-electrical parameters and signals include device ID, phase current, phase voltage, power-factor, bus-bar temperatures, accelerometer data, vibrations recorded, ambient environmental temperature, level of humidity near the electrical switching circuits, geographical location, the ambient Electromagnetic field Intensity (EMI), the EMI source characteristics, arc signatures, load characteristics, building characteristics, nuisance trip and trip or no trip or pick-up or drop out signals.
4 . A method implemented in an electric power distribution protection system integrated with a hybrid machine-learning model (HMLM), comprising: transmitting, continuously or periodically, data from each of the plurality of field protection systems to one or more servers of a central server from a plurality of field protection systems; wherein the data includes time stamped processed electrical and non-electrical parameters and signals; wherein the plurality of field protection systems is connected to a plurality of electrical switching circuits configured for protecting the electric power distribution system; processing the received data to identify change in patterns of the data that is time stamped, and extract data that is changed in patterns from a predefined list of data provided in the HMLM; performing analysis on the processed data using the HMLM by computing an error related to the HMLM in comparison with the processed data; calibrating the HMLM to minimize the error; updating a decision-making firmware or weighted output parameters of the plurality of field protection system, from the calibrated hybrid machine-learning model, to develop a new decision making firmware or a weighted output parameters; and sending an update regarding the new decision making firmware or the weighted output parameters to the plurality of field protection systems to control actuation of each of the connected plurality of the electrical switching circuits.
5 . A field protection system (FPS) integrated with a hybrid machine-learning model (HMLM), wherein the FPS is connected to an electrical switching circuit configured for protecting an electric power distribution system, the FPS comprising: a sensing unit comprising a plurality of sensors configured to sense data that includes a plurality of electrical and non-electrical parameters and signals from each of a plurality of electrical switching circuits; a signal-conditioning unit configured to process the data received from the sensing unit, and transmit the processed data to a controller and to an edge analytic device (EAD); wherein a central server comprising one or more servers is connected to the FPS through a communication network or a Local Area Network; wherein the EAD is configured to: perform supervised or unsupervised machine learning locally from the data that is received from the signal-conditioning unit and extract data that is changed in patterns from a predefined list of data provided in the HMLM; make local decisions based on a Machine-learning (ML) algorithm and the extracted data; time stamp the extracted data; and transmit the time stamped data and the local decision to the controller through a first communication interface; wherein the controller is configured to: receive the time stamped data and the local decision through a second communication interface from the first communication interface of the EAD; process the time stamped data and the local decision using decision making firmware to make decisions to control actuation of each of the connected plurality of the electrical switching circuits that comprises one or more circuit breakers and one or more trip-units; transmit, continuously or periodically, the time stamped data to the one or more servers of the central server or the cloud or other field protection systems through the local area network (LAN) from the controller or through the first communication interface of the edge analytic device, when the controller is bypassed from the LAN or cloud for future decision making; receive a new decision making firmware or weighted output parameters for the FPS from the one or more servers or the other field protection systems through the LAN based on a calibration of the HMLM; and utilize the new decision making firmware or the weighted output parameters to update the HMLM thereby to make future decisions in the FPS.
6 . The FPS as claimed in claim 5 , wherein the plurality of electrical and non-electrical parameters and signals includes device ID, phase current, phase voltage, power-factor, bus-bar temperatures, accelerometer data, vibrations recorded, ambient environmental temperature, level of humidity near the electrical switching circuits, geographical location, the ambient Electromagnetic field Intensity (EMI), the EMI source characteristics, arc signatures, load characteristics, building characteristics, nuisance trip and trip or no trip or pick-up or drop out signals.
7 . The FPS as claimed in claim 5 , wherein the controller is further configured to make future decisions on whether to actuate the electrical switching circuits based on user feedback by reinforced learning under critical situations, after actuating the one or more circuit breakers; wherein the critical situation includes unsure decisions of the HMLM related to a change in pattern of the data sensed from the sensing unit of the FPS.
8 . A method implemented in a field protection system (FPS) integrated with a hybrid machine-learning model (HMLM), comprising: collecting data that includes a plurality of electrical and non-electrical parameters from a plurality of sensors of each of a plurality of electrical switching circuit that are connected to the FPS; transmitting the data to a signal-conditioning unit; processing, at the signal-conditioning unit, the data received from the plurality of sensors, compatible for communicating with a central server; transmitting, continuously or periodically, the processed data to a controller and an edge analytic device (EAD); wherein the central server comprises one or more servers that are connected to the FPS through a communication network; performing supervised or unsupervised machine learning locally at the EAD, from the data that are received from the signal-conditioning unit, and extracting data that is changed in patterns from a predefined list of data provided in the HMLM; making local decisions based on a Machine-learning (ML) algorithm and data extracted; time stamping the extracted data and encrypting the time stamped data for data security; transmitting the time stamped data and the local decision to the controller through a first communication interface; receiving, at the controller, the time stamped data and the local decision through a second communication interface from the first communication interface of the EAD; processing, at the controller, the time stamped data and the local decision using decision making firmware to make decisions to control actuation of each of the connected plurality of electrical switching circuits that comprises one or more circuit breakers and one or more trip-unit; transmitting, continuously or periodically, the time stamped data from the controller or the EAD to the one or more servers of the central server or the cloud or other field protection systems that are connected through the same LAN; receiving, at the controller, a new decision making firmware or weighted output parameters for the FPS from the one or more servers or the other field protection systems based on a calibration of the HMLM; and utilizing the new decision making firmware or the weighted output parameters to update the HMLM thereby to avoid nuisance tripping in the FPS.
9 . A field protection system (FPS) integrated with a hybrid machine-learning model (HMLM) for protecting the electric power distribution system, the FPS comprising: a sensing unit comprising a plurality of sensors configured to sense data that includes a plurality of electrical and non-electrical parameters from each of a plurality of electrical switching circuits, and transmit the data to an edge analytic device (EAD); wherein the EAD comprises a processor, a sensor interface and a first communication terminal; wherein the processor is configured to: perform supervised or unsupervised machine learning locally from the data that is received from the sensing unit through the sensor interface, and extract data that changed in patterns from a predefined list of data provided in the HMLM; make local decisions based on a Machine-learning (ML) algorithm and the extracted data; time stamp the extracted data; transmit the time stamped data and the local decision to an electrical switching circuit through the first communication terminal via a communication interface; encrypt the time stamped data for data security; communicate the encrypted time stamped data to other field protection systems through the first communication terminal via the LAN to calibrate the HMLM; and utilize the new decision making firmware or the weighted output parameters to update the HMLM to make future decisions in the FPS; wherein the electrical switching circuit comprises at least one circuit breaker and at least one trip-unit; wherein the at least one trip-unit comprises at least one protection controller, a second communication terminal and at least one actuator; wherein the protection controller is configured to: receive the time stamped data and the local decision from the EAD through the second communication terminal via the communication interface; and process the time stamped data and the local decision using decision making firmware to make decisions to control the at least one actuator that is connected to the at least one circuit breaker in the electrical switching circuit.
10 . A method implemented in a field protection system (FPS) integrated with a hybrid machine-learning model (HMLM), comprising: collecting data that includes a plurality of electrical and non-electrical parameters from a plurality of sensors of each of a plurality of electrical switching circuit that are connected to the FPS; transmitting the data to an edge analytic device (EAD); performing supervised or unsupervised machine learning locally at the EAD, from the data that is received from the plurality of sensors through the sensor interface, and extracting data that is changed in patterns from a predefined list of data provided in the HMLM; making local decisions based on a Machine-learning (ML) algorithm and the extracted data; time stamping the extracted data and encrypting the time stamped data at the EAD for data security; transmitting the time stamped data and the local decision to an electrical switching circuit through a first communication terminal of the EAD via a communication interface; receiving, at a second communication terminal of a protection controller of a trip unit of the electrical switching circuit, the time stamped data and the local decision from the EAD through the communication interface; and processing the time stamped data and the local decision using decision making firmware to make decisions to control at least one actuator that is connected to at least one circuit breaker in the electrical switching circuit; communicating the encrypted time stamped data to other field protection systems through the first communication terminal via the LAN to calibrate the HMLM; and receiving the new decision making firmware or the weighted output parameters to update the HMLM thereby to avoid nuisance tripping in the FPS.
11 . The system as claimed in claim 1 , wherein the time stamped processed electrical and non-electrical parameters and signals include ambient Electromagnetic field Intensity (EMI) and EMI source characteristics obtained by an EMI sensing unit, and wherein each of the field protection systems is configured to implement a machine learning (ML) algorithm to characterize EMI signatures by generating EMI parameters using the ambient EMI and the EMI source characteristics and tag the EMI source characteristics with a corresponding EMI source.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This patent application claims priority to International Patent Application No. PCT/IB2021/054091, filed on May 13, 2021, and to Indian Patent Application No. 202041020403, filed on May 14, 2020, each of which is incorporated herein in its entirety by reference. FIELD OF THE DISCLOSURE The present disclosure generally relates to smart electrical protection systems and, more particularly, to a system and method for protecting an electric power distribution system integrated with a hybrid machine-learning model (HMLM). BACKGROUND OF THE INVENTION Generally, protection devices such as circuit breakers or switch gears are in operation for safety of a customer property. These protection devices are utilized to minimize a delay in fault detection and isolation of faulty equipment from an electrical circuit. Early detection and localization of faults, and prompt removal from service of the faulty equipment, is performed to safeguard the entire system, ensure continuity of supply, and minimize damage and repair costs. At some instances, due to field noise like EMI conditions or other operating load conditions, noises are injected into the electrical circuit. The protection devices pick up those noises, which may lead to certain trip conditions like GF, RELT etc. as nuisance trips, and as a result corresponding electrical circuits or elements may get isolated. Some noises are injected due to human error by introducing a potential arc flash before operating a switch on a panel, which can lead to tripping of the circuit breaker (CB). In many cases, these nuisance trips may create inconvenience to the customer or undermine critical true safety events. Conventionally, filters are used to check the injected noises to make decisions accordingly on tripping the electrical circuits. But those filters are designed based on laboratory testing results and cannot be adapted to environmental changes. Especially, for noises such as electromagnetic interference (EMI), the filters are not capable enough to characterize sources of the EMI, according to the environmental changes. In some internet of things (IOT) based tripping circuits, the decision is made by considering loads of the electrical circuit for isolation. Such IOT based tripping of electrical circuits will not characterize the load and need a Cloud or central server to communicate between a breaker fleet within a home or building, before making the tripping decision. In most of the cases, data such as trip data, signatures of events, sensor data, historical events, non-electrical parameter data such as temperature, vibrations, humidity and so on are not analyzed in making decisions before isolating the electrical circuit. BRIEF SUMMARY OF THE INVENTION Based on the foregoing, there is a need for a system and method for protecting an electric power distribution system integrated with a hybrid machine-learning model (HMLM) to make decisions based on a real-time data and also by learning nuisance trip events and its environmental conditions for isolating noise parameters and adapting to ignore them. The present disclosure is directed to a system and method configured and operating to mitigate, alleviate or eliminate one or more of the above-identified deficiencies and disadvantages in the prior art and solve at least the above-mentioned problem. Accordingly, an embodiment herein provides a first aspect of a system for protecting an electric power distribution system integrated with a hybrid machine-learning model (HMLM) and a second aspect of a method for protecting an electric power distribution system integrated with a HMLM. Further, in view of the foregoing, another embodiment herein provides a third aspect of a field protection system (FPS) integrated with a hybrid machine-learning model (HMLM) and a fourth aspect of a method implemented in a field protection system (FPS) integrated with a HMLM. Furthermore, in view of the foregoing, yet another embodiment herein provides a fifth aspect of a field protection system (FPS) integrated with a hybrid machine-learning model (HMLM) and a sixth aspect of a method implemented in a field protection system (FPS) integrated with a HMLM. BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS The detailed description is set forth with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different figures indicates similar or identical items. FIG. 1 is a block diagram for a system for protecting an electric power distribution system integrated with a hybrid machine-learning model (HMLM), according to the present disclosure. FIG. 2 is a functional diagram of a field protection system (FPS) integrated with a hybrid machine-learning model (HMLM) according to the present disclosure. FIG. 3 is a block diagram of a field protection system (FPS) inte