Electricity is one of the great luxuries of the 21st century. Flick a switch, and your home has light, heat, a variety of entertainment options, not to mention an instant connection to the corners of the world. Behind the scenes, there is a frenzy of generation, forecasting, and aggregation of power sources, but all a consumer cares about is that the switch turns on reliably––their safety and livelihood depend upon it.
The behind-the-scenes aspects of power plants have made the utilities industry primed for artificial intelligence (AI) implementation. Utilities are heavily asset-driven, and a power company’s priorities are to maintain uptime and ensure reliability. Now that sensors are ubiquitous, assets and plants themselves are producing ridiculous amounts of data that must be translated to actions––a perfect opportunity to leverage machine learning.
It is easy to forget the power grid is the largest and yet the most complex real-time machine ever built. Relying upon human expertise alone to enhance security and reliability of this machine is no longer an option.
Currently, many power plants use rule-based feedback loop to ensure asset health and condition––but these static systems are proving insufficient. If an unexpected mechanical problem occurs and a generation unit goes down, humans are required to intervene and rebalance the grid. This dependence on human input creates long lag times during which consumers are typically left in the dark.
A smarter, connected power plant means utilizing actionable data insights to ensure that assets are fully understood by operators, creating reliable generation and grid stability. Artificial intelligence algorithms sort through historical and new data to create models for both asset behaviors and demand forecast nearly in real time. It’s the first step to autonomous plants.
AI models of asset behaviors predict when an asset will require maintenance, detect anomalies, and help utilities plan maintenance schedules and optimize operations. For example, SparkPredict, alerted one utility company to a never-before-seen manufacturing defect on a turbine that was missed by other monitoring systems. The program captured this anomaly due to AI’s capability to find new relationships between the asset’s real-time data and its normal operation. With new strides in AR/VR, even the way that assets are serviced could look completely different in a connected plant.
Though machine learning and AI make operations more predictable, not every outage can be predicted. Dynamic AI models powered by IIoT data respond more quickly to new information from real-time events and suggest adjustments to the grid operators or could even make them automatically.
This is not to say a connected plant will operate without humans. Rather, expertise from the workforce will be captured in algorithms, and employees can shift focus to other areas (such as maintaining assets rather than operating them). For example, SparkCognition is working with Mitsubishi-Hitachi Power Systems’ Tomoni technology to identify anomalies in power plants, recognizing the need for input from OEMs and plant operators. This knowledge retention is also a significant benefit in an industry where subject matter experts are reaching retirement age en masse, and there is great need to get new hires up to speed quickly.
In addition, a connected power plant will make work safer for its employees. There’s a world of wearables and IoT devices that can gather data from environmental conditions, provide safety alerts, or monitor conditions to determine human and machine behaviors that may cause accidents.
To learn more about SparkCognition’s view on the connected plant, come find me at the Connected Plant Conference next week (Stuart Gillen, our Senior Director of Business Development, Customer Success, and Partnerships, will be speaking), or email us email@example.com