Predicting outages with advanced Machine Learning and physics-based algorithms.
Every power distribution company faces unique challenges. Ageing infrastructure and equipment plagues every utility but each has specific pain points and equipment types that are responsible for large amounts of customer outage time.
How does it work?
In order to provide actionable predictions that can be used to mitigate outages, Cascadence solves 3 key problems:
Detecting when anomalous power quality events occur
Event Type Classification
Correlate power quality events to device types and specific causes
Incipient Fault Location
Determine the position of the failing equipment
The Cascadence data pipeline can be custom built to support the requirements of any chosen power quality monitoring device. Live data is fed to our machine learning algorithms which have been trained to pick out precursors to equipment failure. Precursors are classified by equipment type, then the location of the device is determined before the results are displayed in a purpose built, customizable dashboard.
1. Event Detection
Continuously monitoring power quality data at the substation produces large amounts of data that’s just too much to interpret. Cascadence automatically pulls out anomalous events from continuously streamed real time data in order to provide only the most pertinent event data ready for classification and analysis.
2. Event Type Classification
After anomalous events have been detected, they are run through a series of Machine learning models that have been trained to automatically classify events based on their unique signatures. This process of classification is what separates the precursors to failure from the thousands of other signatures caused by the everyday operation of the electrical grid.
3. Incipient Fault Location
Detecting and classifying a precursor to failure indicates an impending failure, but locating the position of that precursor provides the means to mitigate the forced outage. The location algorithm is an impedance-based calculation, capable of working on sub-cycle faults, and crucially providing the distance from the substation for the precursors and the outages themselves. Cascadence can plot a patrol area based on precursor impedance and further refine the area with alarms and smart sensors from other devices on the grid.
Disparate data in one place
The value of collecting data from multiple sources is only fully realised when it is brought together in one place so that all sources are used to inform decision making. Cascadence brings together multiple disparate datasets from multiple business lines in order to maximise the value of each component piece. Once ingested, the predictive power of machine learning algorithms can be applied for classification and our patented impedance based calculations combined with smart grid sensors can preemptively locate faults.