IoT + edge processing is making sci-fi a reality for energy management

Tuesday, March 2, 2021

Source | Smart Energy International

What is the role of IoT and edge computing in optimising the operations of smart buildings, distributed renewable energy resources, remote pipeline management, smart homes and hydro-electric power generation facility control?

The Internet of Things (IoT) is having a dramatic impact on all aspects of our lives. In my home, even the mundane has become magical. I now enjoy connected garage door openers that “automagically” close in the evening at an appointed time. Connected thermostats (more like smart tablets on the wall) know what the weather is and what it will be based on internet-provided weather data, anticipating my heating needs.

The lawn sprinkler controller also leverages Internet-provided weather data to determine if watering is really necessary that day. My home lighting shifts colour based on time of day, and I am surrounded by Siri, Alexa, and Google to play my music, give me answers, and tell corny jokes on demand. I can answer my doorbell from my smartphone, and when I tell my television what I want to watch, it finds it. I even catch my wife, an admitted and proud luddite who loves the smell of a paper book, saying “Alexa, play my favourites”.

But what we see in our own homes and lives is merely the tip of the IoT iceberg – an iceberg that continues to grow. More IoT devices and applications are being created every day, with companies like Microsoft, Amazon Web Services, Google, and IBM providing IoT platform solutions that can be used to collect and process the massive amounts of device-collected data, while determining and directing the appropriate actions. Artificial Intelligence is increasingly being applied to all of this device-collected data, giving us the term AIoT.

Not only are the number of devices and applications growing, but the ways we can connect them to networks is also increasing.

5G wireless technology opens the door for collecting even greater amounts of data from the most prolific connected devices, generally those that capture video. The Facebook Connectivity programme is focused on bringing reliable Internet to places in the world where it does not exist (yes, they will then have a larger ad audience, but connectivity is good).

LoRaWAN radio technology is addressing the need to connect to devices over long distances or in areas where technologies like Wi-Fi and mobile wireless cannot penetrate.

And SpaceX’s Starlink satellite Internet solution promises high-speed, affordable connectivity to some of the most remote parts of our world.

THE DELAY CAUSED BY SENDING EVERYTHING TO THE CLOUD FOR DEEP PROCESSING ESPECIALLY WHEN ONE CONSIDERS THE OCCASIONAL NETWORK CONGESTION THAT SLOWS THINGS DOWN IS AN IMPEDIMENT
TO NEW, REAL-TIME SERVICES…

There is, however, something missing. Have you noticed that you cannot have a back and forth conversation with Alexa or Siri? Or that commands sent from your phone app to control your home devices are not executed immediately – it takes a few seconds. Not a big deal when it is your garage door, but what if it is an urgent medical device where seconds matter, or a machine that relies on network-based control to avoid self-inflicted destruction?

The missing ingredient here is edge computing. Today, when you ask Alexa or Siri “what’s the weather?” your end device uses the “trigger word” as an indicator to begin recording your voice. This digital recording is then sent to “the cloud”, a vast warehouse data centre which could be hundreds of
miles away or more. Since signals travel at light speed, the physical distance is not so much the issue as the “network distance”, the number of different networks and router connections your voice recording needs to travel to get there.

Once your recorded voice arrives, it must be processed by Natural Language Processing (NLP) software to extract the meaning of your request. At this point, an AI engine “understands” that you would like to know the weather forecast and uses its available information about you to determine your location. At that point, the cloud AI can request and receive the forecast in your location. If the forecast is stored as text, it must be converted to voice and sent back to your device, again across multiple networks.

Finally, your device speaks today’s forecast to you. When you consider all that is going on in this simple exchange, you can see why conducting a full conversation with Alexa, Siri, or other AI assistants in real time, at human speed, is not trivial.

The transmission and processing delay based on using the data centre cloud precludes it. It is not a processing problem; it is a distance one. The delay, or latency, caused by sending everything to the cloud for deep processing – especially when one considers the occasional network congestion that slows things down – is an impediment to new, real-time services like conversational AI, Augmented Reality (AR), Virtual Reality (VR), and network-based vehicular control, to name a few. If this latency could be significantly reduced, a new class of network-based applications could be realised.

Edge computing answers this problem, and also addresses a few others. By placing processing resources, or “edge servers” close to the devices that they serve and control, the latency can be dramatically reduced.

This “distributed computing” model not only reduces latency but can also substantially reduce the amount of data that needs to be sent between the central cloud and the devices, saving bandwidth costs. Keeping the data close to the devices also offers security benefits. And finally, with localised edge processing, autonomous control of devices is made possible even when the network is congested, or the connection is lost.

So, what does all of this have to do with smart energy? A great deal, in fact. As just one example, consider electricity-generating wind farms, which may be offshore or in remote locations. Central cloud connections to these locations can be sporadic and offer limited bandwidth, and that bandwidth can be very expensive. But if we combine wireless sensor and control connectivity with local data processing and AI-based applications, the entire operation can be controlled and managed locally and semi-autonomously, while still offering centralised, cloud-based monitoring and analytics for the team at the home office.

Smart Edge Nodes (SENs) can make this a reality. A new networking equipment category, SENs integrate a variety of IoT device connectivity options, WAN connectivity options, and processing resources all in a single device that is about the size of a router. In this case, each wind turbine could have sensors and controls that use LoRaWAN connectivity to address the distances involved. A single compact SEN could connect to all of these turbines and run a local application to manage their operations. It would then use its WAN connection – an integrated mobile wireless radio or ethernet-connected satellite modem – to send periodic, curated analytics data bundles for home office review. It could also capture vibration information from specialised sensors to determine impending failure and take the appropriate control actions to minimise damage. We can take this example a bit further into the sci-fi realm, in which the SEN would also run a control application for drone-based visual inspection.

The video captured by the drone could be locally analysed by the SEN using AI-based computer vision to identify issues, extract the relevant visual and telemetry data that correlates to anomalous vibration statistics, and then send ONLY the data needed for further analysis back to the central cloud. And over time, if additional computation or connectivity resources are required, additional SENs can easily be added to the local mesh network of pooled SEN-provided resources.

Wind farms are just one easily described example, but the benefits of edge computing can open up new possibilities in electric grid management, hydro-electric power generation facility control, environmental and security controls for smart buildings, remote pipeline management and maintenance, drilling and mining operations, and a great deal more.

By deploying SENs where connectivity, local data analysis, and automated control are needed, and combining them with the power of deep central cloud AI processing, an entirely new class of intelligent, autonomous systems is possible. The capabilities, performance, and responsiveness of this edge + central cloud processing model can clearly eclipse what a central-only architecture can achieve. When we also consider the cost savings due to reduced data transmission and distributed processing, it is clear that edge computing will become a core element of the IoT processing model of the future.