Arc fault detection through mixed-signal machine learning and neural networks

A method of detecting presence of an arc fault occurring within a circuit including a line terminal, the method comprising:

• measuring, with a wideband current sensor, a current flowing through the line terminal;
• applying, via a microcontroller including an electronic processor, a digital filter to a line current measurement signal received from the wideband current sensor;
• estimating, via the microcontroller, a spectral density of the filtered line current measurement signal;
• calculating, via the microcontroller, an arcing characteristic based on the spectral density;
• determining, via the microcontroller, whether an arc fault is present within the circuit based on a comparison of the arcing characteristic to one or more thresholds;
• and
• activating, via the microcontroller, an interrupting device when an arc fault is present;
• wherein the microcontroller is supported by a first printed circuit board and the wideband current sensor is supported by a second printed circuit board.

A system comprising:

• a circuit interrupting device including:
• a line terminal;
• a wideband current sensor configured to measure a current flowing through the line terminal;
• and
• a communication circuit configured to wirelessly transmit a signal including current measurements taken by the wideband current sensor;
• and
• an external device including:
• a second communication circuit configured to wirelessly receive the signal;
• and
• a controller including an electronic processor configured to:
• estimate a spectral density of the current measurements;
• calculate at least one arcing characteristic associated with the spectral density;
• develop a machine learning model based on the spectral density and at least one arcing characteristic;
• and
• deploy the machine learning model to the circuit interrupting device.

A method of deploying a machine learning model to a circuit interrupting device, the method including:

• receiving, via a communication circuit, a first set of line current measurements that were taken when an arc fault was present;
• receiving, via the communication circuit, a second set of line current measurements that were taken when an arc fault is not present;
• estimating, via a controller including an electronic processor, a first spectral density of the first set of line current measurements;
• estimating, via the controller, a second spectral density of the second set of line current measurements;
• calculating, via the controller, a first set of arcing features associated with the first spectral density;
• calculating, via the controller, a second set of non-arcing features associated with the second spectral density;
• executing, via the controller, a training algorithm to create the machine learning model based on the first set of arcing features and the second set of non-arcing features;
• and
• deploying the machine learning model to the circuit interrupting device.

A circuit interrupting device comprising:

• a line terminal;
• a wideband current sensor configured to measure a current flowing through the line terminal;
• a microcontroller including an electronic processor, the microcontroller configured to:
• estimate a spectral density of a line current measurement taken by the wideband current sensor;
• calculate an arcing feature based on the spectral density;
• calculate, by executing an inference algorithm, a probability of an arc fault occurring based on the arcing feature and the spectral density;
• set a first flag when the probability exceeds a threshold for a first amount of time;
• and
• activate an interrupting device when the first flag is set.

A method of detecting presence of an arc fault occurring within a circuit including a line terminal, the method comprising:

• measuring, with a wideband current sensor, a current flowing through the line terminal;
• estimating, via a microcontroller having an electronic processor, a spectral density of a line current measurement signal generated by the wideband current sensor;
• calculating, via the microcontroller, an arcing feature based on the spectral density;
• calculating, via executing an inference algorithm, a probability of an arc fault occurring based on the arcing feature and the spectral density;
• setting a first flag when the probability exceeds a threshold for a first amount of time;
• and
• activating an interrupting device when the first flag is set.