Lately, I've been thinking a lot about artificial intelligence – specifically, how machine learning (which is rooted in analytics) is a subset of this technology. Couldn't follow that line of thought? Simon Chan, CEO of PredictionIO provided a pretty good explanation in his blog.
In a way, predictive analytics and machine learning are one in the same. The former technology focuses on estimating the future condition of certain places, situations or people based on historical and real-time activity. In comparison, machine learning occurs when a device collects and analyzes data and draws conclusions regarding certain factors. Apple's Siri and Microsoft's Cortana often use behavioral and predicative algorithms to self-educate.
"Machine learning is one component of a much greater grid infrastructure scheme."
How utilities are using machine learning
As we're a big fan of how utilities are using analytics, let me present a hypothetical situation to you. Suppose a server running a machine learning program connects to thousands of intelligent electronic devices (IEDs), substation monitoring applications and smart meters dispersed across an entire grid. Based on customer demand data, power plant output information and distributed energy production rates, the machine learning application develops an "understanding" of how certain events will positively and negatively impact services.
From a deeper perspective, the technology is capable of solving the problems posed by the radial topologies power companies have used pretty much since the inception of the industry. Radial architectures organize resources so that power flows in one direction, that is, from power plants to consumers. Now, distributed energies are rendering this model obsolete, and machine learning is one component of a much greater infrastructure scheme that will sanction multi-directional electricity distribution.
eMeter Director of Product Management Krishan Gupta spoke with Siemens on the matter, describing the pattern recognition elements of machine learning and how they can be applied to the smart grid.
"For instance, if you've identified some instances of energy theft, then you can feed into your analytics system data from those cases and have it look for similar patterns in current customer data," said Gupta, as quoted by the source.
"It's important to keep expectations tempered in regard to artificial intelligence."
Analytics on wings
If you'll please excuse the corny subheading, take a gander at how airline companies are paying for aircraft engines. Tata Consultancy Services wrote a white paper on machine learning's impact on various industries, noting the technology's applicability to air services. Nowadays, airlines are paying for engines based on a time on wing metric, which identifies the operational reliability of an aircraft engine.
Due to this payment model, companies manufacturing such machines are obligated to make these engines more dependable. This is where the pattern recognition capabilities of machine learning come into play. These algorithms can isolate vulnerabilities within these implementations and accurately identify what kind of maintenance is required to repair them. The assessment doesn't stop there, multiple operational components caused by frequent use and external factors can be analyzed through the intelligent eye of a computer.
Now, we start getting into artificial intelligence. It's important to keep expectations tempered in regard to AI. The Von Neumann architecture can't support the kind of functionality you'd see in "I, Robot" or "Terminator," but the advent of quantum computing may bring this technology to fruition.