Systems and methods for monitoring a fuel cell
A monitoring device tracks a plurality of related metrics, each related metric configured to monitor the same characteristic of a fuel cell system based on different types of data. The related metrics may include a first metric derived from direct measurements of the characteristics and a second metric. The second metric may be inferred from indirect measurements of the characteristic. An analysis module may detect a potential fuel leak based, at least in part, on an error, difference, deviation, and/or other comparative analysis of the related metrics. Alternatively, or in addition, potential fuel leaks may be detected by use of an artificial intelligence, machine learned and/or machine learning (AI/ML) model trained to identify related metrics that are indicative of anomalous operation of the fuel cell system, such as operation under fuel leak conditions.
1 . A method for monitoring a fuel cell system of a vehicle, the method comprising:
deriving a first metric from first monitoring data, the first metric configured to track a specified characteristic of the fuel cell system during a monitoring period;
deriving a second metric from second monitoring data, the second metric configured to track the specified characteristic of the fuel cell system during the monitoring period;
generating features corresponding to a machine-learned (ML) model from the first metric and the second metric, the features corresponding to a rate of change of the first metric and a rate of change of the second metric;
configuring the ML model to determine a leak prediction for the fuel cell system based, at least in part, on the features generated from the first metric and the second metric, the features indicating a difference between the rate of change of the first metric and the rate of change of the second metric at respective times within the monitoring period;
triggering detection of a fuel leak in the fuel cell system based, at least in part, on the leak prediction determined by the ML model; and
implementing a mitigation action at the vehicle in response to detection of the fuel leak, the mitigation action comprising closing a valve of the fuel cell system.
2 . The method of claim 1 , wherein:
the specified characteristic pertains to one or more of: a power output of the fuel cell system and fuel consumption of the fuel cell system.
3 . The method of claim 1 , wherein:
the first monitoring data used to derive the first metric comprises direct measurements of the specified characteristic;
the second monitoring data comprises indirect measurements pertaining to the specified characteristic; and
deriving the second metric comprises determining expected values of the specified characteristic from the indirect measurements.
4 . The method of claim 1 , wherein:
the specified characteristic comprises a fuel consumption of the fuel cell system; and
the first monitoring data comprises measurements acquired from one or more of the fuel cell system, fuel storage of the fuel cell system, and a fuel coupling of the fuel cell system.
5 . The method of claim 4 , wherein the first monitoring data comprises measurements of one or more of: a temperature of fuel within the fuel storage of the fuel cell system, a pressure of the fuel, a mass of the fuel, fuel inflow to the fuel cell system, and fuel outflow from the fuel cell system.
6 . The method of claim 4 , wherein:
the second monitoring data is configured to quantify one or more of an output power of the fuel cell system and an input power supplied to a load coupled to the fuel cell system; and
the second metric is derived by use of power efficiency data pertaining to the fuel cell system.
7 . The method of claim 1 , wherein the specified characteristic corresponds to power generated by the fuel cell system, and wherein the first monitoring data comprises power measurements and the second monitoring data comprises measurements pertaining to fuel consumption by the fuel cell system.
8 . The method of claim 1 , wherein a start time of the monitoring period corresponds to a fueling time of the fuel cell system.
9 . The method of claim 1 , wherein the ML model is configured to determine the leak prediction for the fuel cell system based, at least in part, on fuel cell system (FCS) metadata, the FCS metadata comprising information pertaining to one or more of operating conditions of the fuel cell system, operator inputs to the fuel cell system, and usage of the fuel cell system.
10 . The method of claim 9 , further comprising:
training the ML model to distinguish features corresponding to non-anomalous operation of the fuel cell system from anomalous operation of the fuel cell system, the anomalous operation corresponding to a fuel leak condition.
11 . The method of claim 9 , wherein the ML model comprises one or more of a linear regression predictor and a support vector machine.
12 . The method of claim 1 , wherein the mitigation action comprises one or more of disabling operation of the fuel cell system, closing an input valve of the fuel cell system, closing an output valve of the fuel cell system, alerting an operator to detection of the fuel leak, and transmitting an alert pertaining to detection of the fuel leak through an external network.
13 . The method of claim 1 , further comprising selecting the ML model based, at least in part, on one or more of an age of the fuel cell system and a usage level of the fuel cell system.
14 . The method of claim 1 , further comprising training the ML model to identify metrics indicative of anomalous operation of the fuel cell system.
15 . An apparatus to monitor a fuel cell system of a vehicle, the apparatus comprising:
a processor operatively coupled to a memory;
a monitoring module configured for operation on the processor, the monitoring module configured to:
determine a first metric configured to track a specified operational characteristic of the fuel cell system, the first metric derived from first monitoring data comprising measurements of the specified operational characteristic, and
determine a second metric configured to track the specified operational characteristic of the fuel cell system, the second metric comprising estimates of the specified operational characteristic derived from second monitoring data having a defined physical relationship to the specified operational characteristic;
a feature extraction module configured to generate features corresponding to a machine-learned model (ML) from the first metric and the second metric, the features corresponding to a rate of change of the first metric and a rate of change of the second metric;
an analysis module comprising the ML model, the analysis module configured to trigger detection of a fuel leak based, at least in part, on a leak prediction determined for the fuel cell system by the ML model, the leak prediction based on the features generated from the first metric and the second metric, the features indicating a difference between the rate of change of the first metric and the rate of change of the second metric at respective times within the monitoring period; and
a mitigation module configured to implement a mitigation action at the vehicle in response to detection of the fuel leak, the mitigation action comprising closing a valve of the fuel cell system.
16 . The apparatus of claim 15 , wherein the specified operational characteristic pertains to fuel consumption of the fuel cell system, the apparatus further comprising:
a first monitoring unit configured to receive first monitoring data, the first monitoring data acquired from one or more of the fuel cell system, fuel storage of the fuel cell system, and a fuel coupling of the fuel cell system;
wherein the first monitoring data comprises direct measurements pertaining to the fuel consumption of the fuel cell system, the direct measurements including one or more of a temperature of fuel stored within the fuel storage, a pressure of the fuel, a mass of the fuel, fuel inflow to the fuel cell system, and fuel outflow from the fuel cell system; and
wherein the first metric is derived from the direct measurements pertaining to the fuel consumption of the fuel cell system.
17 . The apparatus of claim 16 , further comprising:
a second monitoring unit configured to acquire the second monitoring data from one or more of the fuel cell system and a load coupled to the fuel cell system, the second monitoring data comprising indirect measurements pertaining to the fuel consumption of the fuel cell system, the indirect measurements including one or more of power generated by the fuel cell system and power supplied to the load coupled to the fuel cell system;
wherein the second metric is derived from the indirect measurements and power efficiency data pertaining to the fuel cell system.
18 . The apparatus of claim 15 , wherein the specified operational characteristic pertains to a power output of the fuel cell system, the apparatus further comprising:
a first monitoring unit configured to acquire the first monitoring data from one or more of the fuel cell system and a load coupled to the fuel cell system, the first monitoring data comprising direct measurements pertaining to the power output of the fuel cell system, the direct measurements comprising one or more of power generated by the fuel cell system and power supplied to the load coupled to the fuel cell system, wherein the first metric is derived from the direct measurements; and
a second monitoring unit configured to receive the second monitoring data, the second monitoring data acquired from one or more of the fuel cell system, fuel storage of the fuel cell system, and a fuel coupling of the fuel cell system, the second monitoring data comprising indirect measurements pertaining to the power output of the fuel cell system, the indirect measurements comprising one or more of a temperature of fuel stored within the fuel storage, a pressure of the fuel, a mass of the fuel, fuel inflow to the fuel cell system, and fuel outflow from the fuel cell system;
wherein the second metric is derived from the indirect measurements and power efficiency data pertaining to the fuel cell system.
19 . The apparatus of claim 15 , wherein the ML model is configured to determine the leak prediction for the fuel cell system based, at least in part, on fuel cell system (FCS) metadata, the FCS metadata comprising information pertaining to one or more of operating conditions of the fuel cell system, operator inputs to the fuel cell system, and usage of the fuel cell system.
20 . The apparatus of claim 15 , wherein the ML model comprises one or more of a linear regression predictor and a support vector machine.
21 . The apparatus of claim 15 , wherein the mitigation action comprises one or more of disabling operation of the fuel cell system, closing an input valve of the fuel cell system, closing an output valve of the fuel cell system, alerting an operator to detection of the fuel leak, and transmitting an alert pertaining to detection of the fuel leak through an external network.
22 . The apparatus of claim 15 , wherein the analysis module is further configured to select the ML model from a plurality of ML models based, at least in part, on one or more of an age of the fuel cell system and a usage level of the fuel cell system.
23 . The apparatus of claim 15 , further comprising a training module configured to train the ML model to identify metrics indicative of anomalous operation of the fuel cell system.
24 . A non-transitory computer-readable storage medium comprising instructions configured to cause a processor of a computing device to implement operations for monitoring a fuel cell system of a vehicle, the operations comprising:
deriving a first metric from first monitoring data, the first metric configured to track fuel consumed by the fuel cell system, the first monitoring data comprising measurements of one or more of a temperature of fuel stored within fuel storage of the fuel cell system, a pressure of the fuel, a mass of the fuel, an inflow of fuel to the fuel cell system, and an outflow of fuel from the fuel cell system;
deriving a second metric from second monitoring data, the second metric comprising estimates of the fuel consumed by the fuel cell system derived from indirect measurements of the second monitoring data and power efficiency data pertaining to the fuel cell system, the indirect measurements comprising one or more of a power output of the fuel cell system and power supplied to a load coupled to the fuel cell system;
generating features corresponding to a machine-learned (ML) model from the first metric and the second metric, the features corresponding to a rate of change of the first metric and a rate of change of the second metric;
configuring the ML model to determine a leak prediction for the fuel cell system based, at least in part, on the features generated from the first metric and the second metric, the features indicating a difference between the rate of change of the first metric and the rate of change of the second metric at respective times within the monitoring period;
triggering detection of a leak in the fuel cell system based, at least in part, on the leak prediction determined by the ML model; and
implementing a mitigation action at the vehicle in response to detection of the leak in the fuel cell system, the mitigation action comprising closing a valve of the fuel cell system.
25 . The non-transitory computer-readable storage medium of claim 24 , the operations further comprising:
receiving the first monitoring data, the first monitoring data acquired from one or more of the fuel cell system, the fuel storage of the fuel cell system, and a fuel coupling of the fuel cell system; and
receiving the second monitoring data, the second monitoring data acquired from one or more of the fuel cell system and the load coupled to the fuel cell system.
26 . The non-transitory computer-readable storage medium of claim 24 , wherein the ML model is configured to determine the leak prediction for the fuel cell system based, at least in part, on fuel cell system (FCS) metadata, the FCS metadata comprising information pertaining to one or more of operating conditions of the fuel cell system, operator inputs to the fuel cell system, and usage of the fuel cell system.
27 . The non-transitory computer-readable storage medium of claim 24 , wherein the ML model comprises one or more of a linear regression predictor and a support vector machine decision boundary.
28 . The non-transitory computer-readable storage medium of claim 24 , the operations further comprising selecting the ML model based, at least in part, on one or more of an age of the fuel cell system and a usage level of the fuel cell system.
29 . The non-transitory computer-readable storage medium of claim 24 , the operations further comprising training the ML model to identify metrics indicative of anomalous operation of the fuel cell system.