Conservation On Auto-Pilot


Written by Brian McConnell, founder of the Worldwide Lexicon project and Der Mundo and a regular guest columnist on GigaOM. He retrofitted his pre-WWII San Francisco home in San Francisco with a 2.5kW net metered solar array. If you live in San Francisco and would like a tour, email brian /at/

The cleantech industry is driving supply-side innovation in energy production. That’s great. Over the next 20 years, we’ll be able to transition to much cleaner energy sources. Of course, we can reduce our energy footprint significantly in the meantime. The problem with conservation is that people associate it with discomfort. It doesn’t have to be that way.

As an engineer, I like to focus on specific, practical examples, so let’s take a look at the household thermostat. At the moment, you have two choices: A dumb analog thermostat (relatively unchanged in a century), or a “smart” thermostat that is so smart you can’t figure out how to program the damned thing. Either way, you end up wasting a lot of energy overheating or overcooling your house, at no benefit to you as a user. So how do you fix this?

A true smart thermostat would be a plug-and-play device. You’d hook it up, enter your Wi-Fi key, give it a preferred temperature range, and then forget it’s there. The thermostat would come with a few low-cost motion sensors that you mount in different zones of your home or office (these are dumb devices that tell the thermostat what zones are occupied). You’d hook these up and forget about them, too.

This thermostat, unlike others, would be able to see the future. It would poll the weather service to fetch weather data and know what outdoor temperatures will be, hour by hour, for the next several days as well as general conditions (sunny or cloudy, dry or wet, etc). With this information, it could do things like keep the heat off on a morning when it’ll be warm or sunny later in the day, or cool the house down early in the morning (during off peak rates) in advance of an especially hot day to reduce peak usage.

These are small decisions that would go unnoticed by the users, but the effect would be a significant reduction in waste with little loss in comfort, as well as peak/off peak load balancing for utilities. (A smart refrigerator that cools itself using outside air at night could also employ this trick to plan energy usage).

Motion sensors, while these are simple low-cost devices, would provide the thermostat with the ability to learn your daily routine. It would learn which rooms are generally occupied for extended times and at what times of day. (Your routine is probably more predictable than you’d care to admit.) This information could be combined with weather forecasts to make heat/cool/do nothing decisions to further conserve energy. A simple AWAY button would tell the thermostat to switch to minimal settings if you’re planning to be away for an extended period of time.

The thermostat would be a Wi-Fi device, so it would be easily able to communicate with a management service that provides additional intelligence and administrative tools. For example, the local electric company might partner with the management service to provide free thermostats to customers so they could remotely manage power consumption during peak periods by asking air conditioning systems in a particular part of the grid to take a break.

Users could go to this web site to remotely manage their own systems, view statistics, etc. The thermostat itself could be a relatively simple device that makes some decisions on its own, and makes others based on instructions from the web service (this approach would allow the decision-making rules to be fine-tuned over time and would be the primary way the user interacts with the HVAC control system).

All of this you could do by replacing only the thermostat, the lowest-cost component of a household or small commercial HVAC system that requires no major work to install. The idea of smart thermostats and remotely controllable thermostats is not new at all. X-10 fans have been improvising systems for years, but this is serious hobbyist territory.

I thought about doing something like this several years ago, but found that X-10 required too much work to set up, and that it produced a lot of HAL-9000 moments (every light in the house turning on at 3 a.m., etc). The basic technology has been out there for years, but ease-of-use issues limit it to the hobbyist and high-end home automation markets. So what I am describing is something that would be baked into standard thermostats so that it becomes a standard practice, and requires no thought on the part of the user.

Because of its simplicity, this type of technique could be deployed on a mass scale, producing quick returns compared with other, more capital-intensive approaches (like retrofitting a home for solar heat) and without requiring utilities to field sophisticated command and control demand response systems (these could be phased in later) .

Made in quantity, there is no reason a smart thermostat should be any more expensive than a Wi-Fi hub (which is itself, a small computer). If a device like this can reduce a typical user’s HVAC bill by 10 percent, that would equate to a six-month payback in many parts of the country.

Why hasn’t this happened on a large scale already? Wi-Fi is still relatively new for most people, only becoming a standard home technology once it’s embedded in residential routers and other low-cost devices. Ten years ago it wasn’t clear that it would become as widely used as it’s become. Ultra-cheap, embedded computers are also relatively new. The convergence of these two technologies makes this practical, so I think that this approach to reducing HVAC waste will begin to take off.

This is also an example in which the private sector could take the lead without waiting for massive state subsidies (although it may require the government to provide a kick in the pants to get the utilities moving). All it takes is an innovative utility company to create a large market and sales channel for these devices. In California alone this is a huge market.

The idea of a smart grid, of course, is not new. Utilities have been developing a variety of strategies for matching supply and demand, conservation, etc. The benefit of this approach is that it is based on the Stupid Network principle, smart endpoints attached to a dumb network. The thermostat doesn’t need to wait for PG&E to build a sophisticated command and control system. It just needs to figure out how to connect to NOAA or Weather Underground to get the data it needs to make autonomous decisions (which most of the time will also happen to be beneficial for the utility in curbing peak demand as well as tailored to the user’s behavior). Being an internet appliance, the thermostats can be field upgraded to take advantage of new web services.

The key point is that the “smart grid” is a misnomer, just as the phone companies pushed the concept of an “advanced intelligent network” in the 90s. All you really need are smart net-aware devices that can make autonomous decisions, and consult centralized web services as needed. The grid itself can be dumb as a box of rocks as long as the devices can obtain information about the grid and overall environment.

This is a small first step. Once smart HVAC controllers are standard equipment in homes and businesses, it would be possible to extend them, just as computers are extensible. As users added new features, like solar-assisted heating, or in ground heat exchangers for cooling, the smart controller could leverage these systems to maximum benefit. The point, though, is to get the camel’s nose under the tent by offering something that is cheap, easy to install and extensible.

So how long will it be until we see Linksys thermostats (and what’s the incremental cost of adding a temperature sensor to Wi-Fi hubs)? This is just one example of how conversation can be baked into autonomous consumer appliances in a way that requires little thought and can produce significant savings from day one. There are plenty of others from CF bulbs with built in light sensors, to dryers that can switch to outside air on warm dry days. The savings from any one decision these devices make may be small, but the cumulative effect will be significant.



this sounds like it would have tons of possibilities for hackers/ prankers

Buck Smith

I very much agree with Brian’s general take on this but am confused why WiFi has to play such a large role? Is that because it’s a more common denominator than protocols like Z-Wave and, therefore, more accessible in a DIY fashion?

My company currently provides much of the feature set Brian describes above and, for some commercial customers, we’re seeing electricity consumption reductions of up to 20%.


It´s a great idea, but the problem with thermostats is usually the time wasted to get at a certain temperature.


I agree with you that smart thermostats should be plug-and-play. A smart grid is, however, necessary to provide MORE value to consumers choosing to install smart thermostats by enabling a more robust market for demand-response. Unlike the telephone network, which could be dumb so long as it connects servers to endpoints with broadband (and many disagree about that, this was part of the national network-neutrality argument 2 yrs ago), the power grid will need to balance it’s load and adapt in real-time to the changes that distributed generation (renewables like solar panels and small wind, and PHEVs) will mean for the grid. That is the reason the grid itself needs to be smart. Also, having a dumb grid leaves on the table another 1% to 3% drop in total energy pushed through the grid that can be saved with load balancing, volt-var and other smart grid strategies. WiFi is okay, but managing millions of smart appliances over an entire city will require a lot of bandwidth, the kind you can get with a BPL (broadband over powerline) system that Xcel Energy is using for its Smart Grid City in Boulder, Colorado (see Xcel webpage).

Comments are closed.