Big data is mostly associated with industries such as retail, healthcare, insurance, banking & finance, and customer communication businesses. I would have believed it to be true, until I started acquiring and analyzing data coming from connected buildings. Without getting into the wiki definition of “Big data,” I can convincingly state that the data received from conventional data sources pertaining to buildings such as, Closed Circuit TVs, standalone controllers, fire alarm system, boilers, hot water heaters, weather stations, solar inverters, sub-meters, and smart meters, is enough to flood your existing database infrastructure with Gigabytes of data you may or may not (mostly may not!) need. Hence, the industry buzz around, “cloud”, “fog,” and “edge” computing becomes very pertinent (A colleague of mine sent me this very interesting article to check out, http://online.wsj.com/news/articles/SB10001424052702304908304579566662320279406).
Expansion of internet (quite literally) due to wide adoption of Internet Protocol (IP) version 6 from IP version 4, faster & cheaper internet bandwidths, faster & cheaper chips, and general economies of scale have driven this big push towards “Internet of Things (IoT).” Buildings are not left untouched. On the contrary, industry gurus and analyst believe IoT may be most impactful in the building industry. Why so? While, the Gurus and analysts may have their own sophisticated reasoning, my feeling is that IoT brings out an impressive contrast between the utility of technologies which “exist” in buildings, as opposed to their utility when upgraded/replaced/retrofitted with IoTs. Most common example can be a web-enabled wall thermostat that learns of your habits and can be controlled over your smart phone, as opposed to the old analog thermostat which proves to be so complicated to operate, it hurts your ego.
The ubiquity of connected devices comprising the IoT is so astounding that today, just about everything you see and feel in the building may be communicating over the internet. Smoke alarms, door locks, supply vents, steam radiators, space sensors, occupancy sensors, actuators, light bulbs, power strips, wall receptacles, window shades, and household appliances (from refrigerators to egg minders!) are just few of the devices which can “talk,” as I would call it. My experience interfacing with many of these devices has taught me some valuable lessons, which is what I’d like you to take back from this read.
- What are your objectives from data: Before doing anything, brainstorm, facilitate a workshop, or go on trek, to figure out the business objectives you intend to achieve with building data gathering and analyzing efforts. In general, I use data to enable the building reduce its energy footprint, but this is a very broad objective. Instead, tread further and list down each objective in crisp and clear fashion. For example, one of my objectives is to identify and turn down/off non-critical loads while making sure tenant comfort is not impeded.
- Quantify data retrieval efforts: As easy as it may seem, retrieving data coming from various IoT and storing it in one place is time and resource consuming. Due to different reasons, IoT provide data in different ways and forms. Spending the time interacting with the data provider, understanding and parsing relevant data, cleaning up the data (if not already done by the data provider) can prove to be a costly affair. Prior to interfacing with a new device, quantify the value of data retrieved and compare it with the level of effort put into it, all in monetary terms. Proceed only if it provides quantifiable returns.
- Don’t be “Techwashed”: Yes, I coined the term! In building industry, many old (and respectable) firms in fear of losing market dominance have started to promote their devices as internet-compatible in order to resonate with the iPhone-crazed consumer of today. While, they may not be totally wrong in their statement, I would still call their technology half-baked. For instance, their Application Programming Interface (APIs) may provide limited data resources, raw data may need lot of processing to get them into decent shape, customer service of their database teams might be far from pleasant, and so on. You’ll be amazed to find the “Techwashers” out there whose data is a tad better than smoke signals.
- Learn to say “No” to data: While, it may feel very enticing to grab all the easy data available and think about it later, maintain an equal distance from the data as well as your excited database guys, and ponder over whether the data directly helps your data analysis objectives or not.
- Make room for flexibility: Often times, I have seen businesses bogged down by the limitations of the black box of data analysis they have created themselves. The IoT is changing (for the good) so rapidly that by the time you are done developing a data acquisition, processing, storage, analysis, and communication infrastructure, the underlying form of data, business objectives, building type, communication medium, and/or sources of data, may have changed. For instance, I witnessed a company develop a front-end on Flash few years back only to be boxed-out when Steve Jobs stopped supporting Flash on Apple devices. Another example is of a company which spent years of time and effort developing an infrastructure around a particular data source, only to find the same data source made available freely and easily to people (as a result of regulatory changes) right at the time they finished their data acquisition architecture.
Try deep learning using MATLAB