Online health monitoring and fault detection for high voltage DC distribution networks

1. A system comprising:

a first power supply coupled to a power channel;

a load coupled to the power channel;

a transmitting sensor coupled to the power channel between the first power supply and the load;

a receiving sensor coupled to the power channel between the transmitting sensor and the load;

a controller in electronic communication with the transmitting sensor and the receiving sensor, the controller configured to:

operate the transmitting sensor to provide an alternating current (AC) test signal to the power channel, wherein the AC test signal comprises a predefined test signal pattern;

operate the receiving sensor to sense, from the power channel, a continuous power signal including the AC test signal;

analyze the AC test signal to determine a fault of the power channel based on comparing the predefined test signal pattern with a predefined nominal probe signal pattern corresponding to a specific network configuration;

wherein the AC test signal to determine a fault of the power channel based on comparing the predefined test signal pattern with a predefined nominal probe signal pattern corresponding to a specific network configuration comprises:

transforming, using a continuous wavelet transform, the continuous power signal to a first scalogram;

generating, using a machine learning model, a first feature vector comprising a plurality of features extracted from the first scalogram;

plotting the first feature vector in a feature space to determine the fault in the power channel.

2. The system of claim 1 , wherein the feature space comprises a first set of feature vectors indicating a normal operation of the power channel based on the predefined test signal pattern.

3. The system of claim 2 , wherein determining the fault in the power channel comprises:

determining a distance value between the first feature vector and the first set of feature vectors indicating the normal operation of the power channel;

determining the fault in the power channel based on the distance value exceeding a pre-defined threshold value.

4. The system of claim 3 , wherein the feature space comprise a second set of feature vectors indicating a first fault type of the power channel.

5. The system of claim 4 , wherein the controller is further configured to:

determine a second distance value between the first feature vector and the second set of feature vectors indicating the first fault type of the power channel; and

determining the first fault type based on the second distance value being within a second pre-defined threshold value.

6. The system of claim 1 , wherein the feature space is generated using labeled training data for the predefined test signal pattern.

7. The system of claim 1 , wherein the feature space is generated based on unlabeled training data comprising historical data for the continuous power signal and the predefined test signal pattern.

8. The system of claim 1 , further comprising:

a second load coupled to the power channel; and

a second receiving sensor coupled to the power channel between the transmitting sensor and the second load, wherein the controller is in electronic communication with the second receiving sensor.

9. The system of claim 8 , wherein the controller is further configured to:

operate the second receiving sensor to sense, from the power channel, the continuous power signal; and

analyze the continuous power signal to determine the fault in the power channel based on comparing the continuous power signal to the predefined signal pattern.

10. The system of claim 1 , wherein the first power supply comprises a direct current (DC) power supply, wherein the DC power supply comprises an AC generator and rectification circuit.

11. The system of claim 1 , wherein the load comprises an AC motor in an aircraft.

12. A method comprising

providing a first power supply coupled to a power channel;

providing a load coupled to the power channel;

providing a transmitting sensor coupled to the power channel between the first power supply and the load;

providing a receiving sensor coupled to the power channel between the transmitting sensor and the load;

operating, by a controller, the transmitting sensor to provide an alternating current (AC) test signal to the power channel, wherein the AC test signal comprises a predefined test signal pattern;

operating, by the controller, the receiving sensor to sense, from the power channel, a continuous power signal including the AC test signal;

analyzing, by the controller, the AC test signal to determine a fault of the power channel based on comparing the predefined test signal pattern with a predefined nominal probe signal pattern corresponding to a specific network configuration;

wherein analyzing the AC test signal to determine the fault in the power channel based on comparing the predefined test signal pattern with a predefined nominal probe signal pattern corresponding to a specific network configuration comprises:

transforming, using a continuous wavelet transform, the continuous power signal pattern to a first scalogram;

generating, using a machine learning model, a first feature vector comprising a plurality of features extracted from the first scalogram;

plotting the first feature vector in a feature space to determine the fault in the power channel.

13. The method of claim 12 , wherein the feature space comprises a first set of feature vectors indicating a normal operation of the power channel based on the predefined test signal pattern.

14. The method of claim 13 , wherein determining the fault in the power channel comprises:

determining a distance value between the first feature vector and the first set of feature vectors indicating the normal operation of the power channel;

determining the fault in the power channel based on the distance value exceeding a pre-defined threshold value.

15. The method of claim 14 , wherein the feature space comprise a second set of feature vectors indicating a first fault type of the power channel.

16. The method of claim 15 , further comprising:

determining a second distance value between the first feature vector and the second set of feature vectors indicating the first fault type of the power channel; and

determining the first fault type based on the second distance value being within a second pre-defined threshold value.

17. The method of claim 12 , wherein the feature space is generated using labeled training data for the predefined test signal pattern.

18. The method of claim 12 , wherein the feature space is generated based on unlabeled training data comprising historical data for the continuous power signal and the predefined test signal pattern.