Open "What if..." Challenge

Machine Learning Home Energy Use Optimization System

"What if" Challenge

 

 

 

 

 

 

 

 

 

 

Machine Learning Home Energy Use Optimization System (MLHEUOS)

 

 

 

 

 

 

 

 

Chris Goettler

Southern Illinois University Carbondale

10/14/2018 

Introduction:

In 2017, the United States' total primary energy consumption was roughly 97.7 quadrillion BTUs. 6.2% of this, or approximately 6 quadrillion BTUs, was used to power residential buildings (U.S. Energy Facts Explained). With the demand for energy expected to continue to grow, optimizing energy consumption becomes imperative. The proposed machine learning system has the potential to decrease residential energy consumption by nearly 20%.

46 percent of the energy used for homes is used in areas such as heating, cooling, lighting, water heating, etc. (How is Electricity Used). These areas of energy consumption represent tremendous energy savings potential. With machine learning implemented into the energy consuming applications in homes, it is possible to maintain our standard of living while at the same time slashing electricity consumption.

 

Statement of Problem:

The world is headed for an energy crisis. One reason is that the world's population is growing at an alarming rate. The world's population in 2017 was 7.6 billion but is expected to reach 9.8 billion by the year 2050 (World Population). With an increase of roughly 2.2 billion people in the next 30 years, it is clear that energy demands will also increase. However, in addition to population growth, energy consumption per capita must also be considered. Since 1970, the average energy consumption per capita has increased by approximately 45% (Ritchie). With the world's quickly growing population and ever-increasing standard of living in mind, an energy crisis is a very real problem in our not-so-distant future.

 

Objectives:

The objective of the proposed system is to decrease a typical home's energy consumption by nearly 20 percent.

 

Proposal:

To obtain a nearly 20% decrease in energy consumption for an average home, a new machine learning home energy use optimization system (MLHEUOS) is proposed. This proposal outlines how a 20% decrease in energy consumption is possible, a tracking system, how to reduce utility energy consumption, and potential problems.

A 20% decrease in energy consumption is possible by optimizing the energy consumption of major energy consuming applications such as climate control, water heating, and miscellaneous electronics such as lighting and TVs. Tables 1-4 show the percent of energy that can be saved for each application by implementing the MLHEUOS. Each section (Overnight, at work, at home) represents an 8-hour block of time. Table 1 shows the potential energy savings during the week for each energy consuming application.

 

Table 1: Percent of energy consumption saved for each application during the week.

Weekdays

Utility Percent Saved

Overnight At work At home Average

 

Water Heating 100.0% 100.0% 0.0% 66.7%

Climate Control 75.0% 100.0% 25.0% 66.7%

Lighting, TV, electronics 5.0% 5.0% 5.0% 5.0%

 

To clarify, 100% of overnight water heating energy consumption could be saved because there is no reason to run the water heater during this time. 75% of overnight climate control energy consumption could be saved because only 1 or 2 bedrooms would need to be heated instead of the entire house. Likewise, during the day, only the living areas and not the bedrooms would need to be heated or cooled, resulting in an approximate 25% savings. Table 2 shows similar data but for weekends.

 

Table 2: Percent of energy consumption saved for each application during the weekend.

Weekends

Utility Percent Saved

Overnight At work At home Average

 

Water Heating 100.0% 0.0% 0.0% 33.3%

Climate Control 75.0% 0.0% 25.0% 33.3%

Lighting, TV, electronics 5.0% 5.0% 5.0% 5.0%

 

By combining table 1 and table 2 and applying a weighted average, it is possible to estimate the average savings for each application on a weekly basis. The weighted average of savings is shown in table 3 below.

 

Table 3: Weighted average of energy consumption saved for each application over an entire week.

Week (Weighted Average)

Utility Percent Saved

Overnight At work At home Average

 

Water Heating 100.0% 71.4% 0.0% 57.1%

Climate Control 75.0% 71.4% 25.0% 57.1%

Lighting, TV, electronics 5.0% 5.0% 5.0% 5.0%

 

Each value in the right hand column is the percent of energy to be saved for each application over a week and in turn the entire year. By knowing the total home energy consumption percentage for each application (Water heating, climate control, etc), it is easy to calculate the percentage of total home energy consumption that can be saved as seen in table 4.

 

Table 4: Total percentage of home energy consumption available to be saved.

Home energy consumption saved

Utility Percentage of home energy consumption Percent saved (figure 3) Percentage of home energy consumption saved

 

Water Heating 10% 57.1% 5.7%

Climate Control 21% 57.1% 12.0%

Lighting, TV, electronics 15% 5.0% 0.8%

 

Total 18.5%

 

As can be seen, based on these estimates an 18.5% savings in energy consumption is possible. Percentage of home energy consumption values obtained from How is Electricity Used in U.S. Homes from EIA website.

The tracking system for the proposed MLHEUOS will be used to track where in the home residents are throughout the day. The system will track residents' whereabouts within the home throughout the day and learn tendencies for each day of the week, holidays, etc. By understanding these tendencies, the MLHEUOS will be able to optimize home energy use by selectively running appliances only when required. For example, if the resident sleeps from 10pm – 7am, the MLHEUOS will turn off the climate control for all other rooms in the house except for the bedroom during this time.

The proposed tracking system for the MLHEUOS consists of a tracking module capable of determining distance and angle between the tracking module and individual trackers for each resident in the household. The most convenient tracker would be a cell phone, although trackers implemented into items such as watches or worn on the waist like a Fitbit would also work. By having the tracker with them throughout the day, the MLHEUOS will learn which rooms the resident occupies and when.

Once the system has learned the tendencies of the resident, it is simple to save considerable amounts of energy. Decreasing home energy consumption by nearly 20% is a realistic goal by optimizing the power consumption of utilities such as climate control, water heating, and electronics such as lights, TVs, and computers. By understanding the tendencies of the resident, the MLHEUOS can turn these utilities on only when necessary and turn them off the remaining time.

An enormous amount of energy is wasted on climate control. Many people continue to heat and cool their homes while they are at work or school. This means these utilities are running close to 8 hours for absolutely no reason. In addition, when the resident is home they are typically only in 1 room or moving between a few rooms, not the entire home. Similarly, when the resident is asleep only the bedrooms need to be heated or cooled, not the entire house.

There are two ways to control which rooms the home heats or cools. The first is to alter the homes HVAC ductwork to a design that can remotely open and close ducts to certain rooms. This would mean that the system can control which rooms are heated or cooled by opening and closing ducts. This approach is likely expensive and quite a hassle, although feasible.

The second, and more economical approach is to install remote controllable floor vent covers on all floor vents in the home. The MLHEUOS would then open and close these vents depending on which rooms need to be heated or cooled. These remote controllable vent covers would be relatively inexpensive. At approximately $10/vent cover, the initial cost would be $200 for a home with 20 floor vents. In the long run this would be worth it, but some consumers might be unable to look past these initial costs.

When it comes to water heaters, all that would be required would be a remote controllable switch that is controlled by the MLHEUOS and turns the water heater on and off. Similar smart home switches on amazon cost roughly $10 so this is not a significant investment (Smart Plug 2-Pack). The MLHEUOS would shut the water heater off in times of no use such as overnight and during the workday.

Controlling the lights, TVs, and other electronics such as computers and printers becomes a little more complex. These devices would also need to be connected to smart home switches, although multiple devices could be connected to the same smart home switch. These electronics can also be integrated into existing smart home systems, saving the consumer money. The MLHEUOS would know when a resident is about to enter a room and turn on all lights in that room. Likewise, MLHEUOS could detect when the resident has left a room for a certain amount of time and would shut off all the devices in that room. Assuming one smart home switch per room at $10/switch, consumers can expect to pay roughly $50 for this feature. Between the floor vents and home switches, the total upfront cost for this system would be roughly $260.

 

Problems and Solutions:

With the potential savings of this system in mind, it is important to consider potential problems. Such problems include decreased effectiveness for larger households, the resident needing to keep the tracker with them while the MLHEUOS learns their tendencies, possible inconvenience for the resident when breaking routine, and the initial cost.

The effectiveness of a system such as the one proposed here will decrease when more people are in a household due to being able to turn less appliances off. This means the potential energy savings might drop from just under 20% to somewhere around 10%-15%.

Another potential problem is requiring the resident to keep the tracking device with them while the MLHEUOS learns their tendencies. For example, if the resident uses a phone as the tracker and charges it while moving around the home, the MLHEUOS system will obtain a false location of the resident. Once the MLHEUOS learns these tendencies however this is no longer a problem.

Third, breaking routine could lead to inconvenience if not planned for in advance. For example, if the resident is normally asleep from 10pm – 7am, the machine would shut off the water heater during these times to save energy. As a result, if for some reason the resident got home very late (3am) and wanted to take a shower, there would be little to no hot water to do so.

One last possible problem with this system is the initial cost. While consumers can expect to save roughly $40/month on utility bills assuming a $200/month energy bill, the upfront cost of $260 might scare many people away. Despite only needing 7 months to break even, some consumers might still be scared away.

 

Conclusion:

In conclusion, reducing the average home energy consumption by almost 20% is more than doable, although there remain a few obstacles to overcome when implementing a system such as the one proposed. Considering 6.2% of energy consumption in the U.S. is for residential applications, an 18.5% decrease in residential energy consumption would save over 1% of the United States' annual energy consumption.

 

 

Works Cited

"U.S Energy Facts Explained." EIA, 16 May 2018. Accessed 14 Oct. 2018.

"How Is Electricity Used in U.S. Homes?" EIA, 8 Feb. 2018. Accessed 14 Oct. 2018.

"World Population Projected to Reach 9.8 Billion in 2050, and 11.2 Billion in 2100." United Nations, 21 June 2017. Accessed 14 Oct. 2018.

Ritchie, Hannah, and Max Roser. "Energy Production & Changing Energy Sources." Our World in Data. Accessed 14 Oct. 2018.

"Smart Plug 2-Pack." Amazon, Accessed 14 Oct. 2018.

Tags

Attachments

- [ Show all ]

Voting

1 vote
1 up votes
0 down votes
Idea No. 280