| CPC H02H 7/26 (2013.01) [G01D 21/02 (2013.01); G05B 13/0265 (2013.01); G05B 19/0428 (2013.01); H02H 1/0007 (2013.01); H02H 1/0092 (2013.01); G05B 2219/25257 (2013.01)] | 10 Claims |
|
1. A system for nuisance-trip decision management in an electric power system using data analytics comprising:
a protection system (PS) and at least one electrical protection circuit (EPC);
wherein the PS comprises:
a plurality of sensors configured to measure data from a plurality of non-electrical parameters and electrical parameters sensors; wherein the measured data includes a plurality of non-electrical parameters and electrical parameters that include ambient temperature, vibrations, electromagnetic interference (EMI) signal, voltage and current waveforms;
a memory configured to store an updated data of nuisance-trip events and a plurality of nuisance-trip parameters of a plurality of protection functions detected or generated on the electric power system over a period;
a processor configured to perform hybrid machine learning (HML) based on the measured data from plurality of sensors for a nuisance-trip condition and communicate with a neighboring protection system through Local Area Network (LAN), central server/cloud via a communication interface-1;
a protection microcontroller configured to use the measured data from a plurality of electrical parameter sensors to identify an electrical fault and allow or avoid tripping at least one electronic tripping circuit provided in the at least one electronic protection circuit by communicating with the at least one EPC, the processor and the neighboring protection systems through the LAN or the central server via a communication interface-2;
wherein the processor is further configured to:
collect the measured data from the plurality of sensors;
generate a nuisance-trip parameter for the plurality of protection functions at a hybrid machine learning model (HMLM) provided in the processor based on the measured data using weighted output parameters thereby to identify the nuisance-trip condition that is mapped with a predefined list of a training data stored in the memory;
wherein the training data is updated to be utilized by a physics aware reinforced machine learning for field calibration; wherein the plurality of protection functions includes a ground fault, an over current, a voltage, a frequency, an arc fault, a zone selective interlock (ZSI), a Reduced Energy Let Through (RELT), and a Power reversal; and
communicate the generated nuisance-trip parameter to the protection microcontroller for an intelligent nuisance-trip decision making via the communication interface-1;
wherein the protection microcontroller is further configured to:
receive the generated nuisance-trip parameter from the processor;
compare the received nuisance-trip parameter with a stored nuisance-trip parameter of the plurality of protection functions to identify the nuisance-trip condition and decide to:
allow tripping an electronic tripping circuit (ETC) of the at least one EPC, if a valid trip condition of the at least one ETC is identified;
avoid tripping the ETC of the at least one EPC, if a confirmed nuisance-trip condition of the ETC is identified; and
allow tripping the ETC of the at least one EPC, if an unconfirmed nuisance-trip condition of the ETC is identified and simultaneously request a user feedback to validate the unconfirmed nuisance-trip condition by:
receiving a confirmation from the user to allow or avoid tripping the ETC; and
updating the weighted output parameters of the HMLM by reinforced machine learning, based on the confirmation from the user to implement a tripping decision in future.
|