Screen shots of the Shiny App:
Global Energy Usage:
Home Energy Usage:
--------------------------------
Introduction
Whatever the cause, it is widely believed that the environmental effects of climate change are having and will continue to have a largely negative impact on people around the world, and in particular, on the poorest and least capable of adjusting to the change through the use of resources. In addition, the higher cost of cleaner alternative energy that will lessen the assumed cause of global warming (i.e., C02 emissions) will be more burdensome to the poorest people and countries. Regardless of one's beliefs about the causes of global warming, the responses to it are a serious matter of social justice.If CO2 emissions are the primary cause, then reducing energy usage (when the burning of fossil fuels for energy production to sustain energy consumption produces those CO2 emissions) is one way of combating global warming and addressing this matter of environmental and social concern. Hence, it is important for businesses and homeowners to understand what factors increase energy usage, what might be done to reduce energy usage, and thereby reduce their contribution to CO2 emissions.
The task of this project is to visually demonstrate and understand the prominent factors that lead to increased energy usage at the local level. To motivate this effort, I first produce visualizations that show the correlation between global energy usage, CO2 emissions, and global relative average temperatures. Having demonstrated the strong correlations among these three, I next produce visualizations that show how home energy usage relates to various factors responsible for my home energy usage increasing and decreasing, including: local temperature, time of day, day of week, month, and (inferred) appliances/devices in use.
Visualizations
The visualizations referred to are located on the “Global Energy Visuals” and “Home Energy Visuals” tabs in this Shiny app.Data Sources
The data sources are:- Home energy usage data from my own house from September 1st, 2017 through September 30, 2018. After having exported from the device (reference: http://eyedro.com/how-to-export-to-excel-from-myeyedro/), the data is now stored in my GitHub repo:
- Global yearly average temperatures:
- Global energy/fuel usage:
- Global CO2 emissions:
- Local weather history to Seattle, WA:
All the above data has been exported for use in the visualizations and is located here:
Parameters of the Data
The data sources above combine into two different data sets with the following parameters:Global
- Energy Production and Consumption
- Nuclear, exports, imports, stock change/other, fuels, renewable, and totals.
- Unit: Quadrillion BTU
- CO2 Emissions
- Gas fuel, liquid fuel, solid fuel, cement production, gas flaring, per capita, and total.
- Unit: million metric tons of Carbon
- Global Temperature Change
- The change in global surface temperature relative to 1951-1980 average temperatures (mean relative temperature), Lowess smoothed mean relative temperature
- Unit: degrees Celsius
- Year
Home
- Calendar
- Year, month, date, hour, day of the week, and day (of the month).
- Watt Hours
- Total and average; port A, port B, and total.
- Unit: Watt hours
- Daily Temperature
- Maximum, minimum, average, difference from 65 degrees.
- Unit: degrees Fahrenheit
- Daily Dewpoint
- Maximum, minimum, average
- Unit: degrees Fahrenheit
- Humidity
- Maximum, minimum
- Unit: percentage %
- Windspeed
- Maximum, minimum
- Unit: mph
- Pressure
- Maximum, minimum
- Unit: Hg
- Precipitation
- Average
- Unit: inches
A Summary of the Current Debate and Challenges with the Data
The current debate around global warming may be summarized as follows. The global warming proponent believes that it is caused by human made CO2 emissions. Consequently, we should take steps to reduce those CO2 emissions to avoid increasing the earth's global average temperature which will be destructive of plant, animal, and human life. The global warming skeptic believes that the warming is part of a natural cycle and processes that create heating and cooling on earth (e.g., volcanic activity, sun activity), and that this is largely independent of human CO2 emissions. Consequently, the increase in CO2 emissions has coincidentally happened at the same time as a natural increase in global average temperatures, but this is mere coincidence. See http://www.longrangeweather.com/global_temperatures.htm for a chart showing this heating and cooling cycle . Furthermore, such heating may actually be beneficial, at least for humans, given that periods in history where humans flourished coincided with periods of warmer than average temperatures. The global warming proponent might accept that natural cycles in heating and cooling do occur and perhaps that we are in such a cycle, but would counter that human activity has dramatically increased the rate and levels at which the earth's natural heating cycle has occurred, and that such an increase is not environmentally sustainable or beneficial.If we had actual measurements of C02 emissions and atmospheric totals going back centuries along with global average temperatures to match, perhaps this debate could be resolved by rigorous data analysis on this data. However, temperatures prior to 1880 are inferred by proxy (see https://en.wikipedia.org/wiki/Proxy_(climate) for details), and different methods (and models) give significant and differing results. The same is true for CO2 emissions and atmospheric totals prior to the 18th century. Consequently, these errors and variances in models of CO2 levels and emissions and global temperatures are further compounded in any model attempting to predict global temperatures from CO2 emissions as distinct from other factors.
Strengths and Weaknesses of this Particular Dataset
The global data is primarily limited by the years available. While the carbon emissions data dates to 1751, the global temperature data is not available until 1880. The energy data isn't available until 1949. Consequently, comparison among these variables is limited by the variable with the fewest number of data points. However, while limited in the number of years of data available, the global data does have useful breakouts by type of carbon emissions and type of energy production and consumption. This allows for a more improved analysis to determine precisely which kinds of carbon emissions and which kinds of energy production and consumption are most associated with global temperature increases.The home energy data is extremely detailed and there are lots of data points. The weather data is complete and could only be improved by having the data at an hourly or minute level of granularity. The biggest limitation is in the energy usage data. While I have the totals for given points in time, I cannot directly observe which appliances are using the energy. For example, I have no data showing the energy usage breakouts for the refrigerator, the heaters/AC, the hot water tank, or the stove/oven. I can infer which appliances are in use based on the day and time of the energy usage and my knowledge of our family's home appliance and electrical usage, but I cannot directly and conclusively show where the energy is being used the most.
Analysis: Visual
Global
The data shows that there is a strong correlation between carbon emissions and increasing average global temperatures since 1880 (in my visualization, a simple linear model returns an R2 of 0.82). This supports the widely accepted view that carbon emissions are the cause of increasing average global temperatures.Those carbon emissions are (presumably) the direct result of energy production and consumption (my simple linear model returns an R2 of 0.89 between energy consumption and carbon emissions). Energy production largely consists of burning fossil fuels which produces CO2 emissions.
Home
To understand energy usage at the local home or business level, I used my own home energy usage over the course of a year and related that information to the local weather, which I assumed to be among the major factors associated with energy usage. Based on the visualization below, I can immediately see that there is a rough negative linear relationship between temperature and energy consumption (lower temperature = higher energy usage; R2 = 0.5).However, knowing that I have air conditioning, I can see that energy usage goes up a bit when the temperature is above 70 degrees and the air conditioning is turned on. Thus, a main energy consumer in my home are heaters and air conditioning, where increasing energy usage is associated with more extreme (hot or cold) temperatures outside. This can be more easily shown comparing the absolute temperature difference between the daily temperature and 65 degrees with total energy usage. A simple linear model returns an adjusted R2 of 0.62.
Another source of energy consumption can be inferred from the two bottom charts on the “Home Energy Visuals” tab. The highest peaks in energy usage are in the morning, followed by a lull midday with some lesser peaks around dinner time. The peaks correspond to water usage, more specifically, hot water usage, and kitchen appliance usage. The morning peaks correspond to daily showers while the dinner time peak corresponds to hand washing dishes, bath-time for the kids, running the dishwasher, and on certain days, doing laundry. The stove and oven are also primarily used at dinner time.
Analysis: Regression
Global
When predicting global mean relative change in temperature, a linear model that considered the following variables returned a model with an adjusted R2 of 0.88:- CarbonEmissions_GasFuel
- CarbonEmissions_LiquidFuel
- CarbonEmissions_SolidFuel
- CarbonEmissions_CementProd
- CarbonEmissions_GasFlaring
- EnergyConsumption_Nuclear
- EnergyProd_Nuclear
- EnergyStockChangeOther
- EnergyConsumption_Fuels
- EnergyProd_Fuels
- EnergyConsumption_Renewable
- EnergyProd_Renewable
Using StepAIC to automatically reduce the variables produced a model with an adjusted R2 of 0.89, and contained the following variables, each of which was statistically significant at 0.05 or lower:
- CarbonEmissions_LiquidFuel
- CarbonEmissions_CementProd
- CarbonEmissions_GasFlaring
- EnergyConsumption_Renewable
Liquid fuel and cement production emissions were positively correlated with global mean temperature relative change, while gas flaring and energy consumption from renewables were negatively correlated.
A model that considered only energy consumption and production in predicting global mean temperature relative change yielded a model with an adjusted R2 of 0.87. The variables that remained in the model after StepAIC were:
- EnergyConsumption_Nuclear
- EnergyStockChangeOther
- EnergyConsumption_Fuels
- EnergyProd_Renewable
A model that only considered carbon emissions in predicting global mean temperature relative change yielded a model with an adjusted R2 value of 0.85. After StepAIC, each of the following variables was statistically significant at 0.001:
- CarbonEmissions_GasFuel
- CarbonEmissions_LiquidFuel
- CarbonEmissions_CementProd
- CarbonEmissions_GasFlaring
To summarize, in comparing energy production and consumption to carbon emissions as factors in predicting global mean temperature relative change, it appears that carbon emissions are more statistically significant. This should not surprise us as most energy production and consumption produces CO2 emissions, and thus energy production and consumption will be highly correlated with increasing global mean relative temperatures if CO2 emissions are in fact the cause of increasing global temperatures. However, energy production and consumption are presumably the remote causes of global warming, while carbon emissions are presumably the primary proximate cause of global warming. Thus, it should not surprise us that carbon emissions, being more directly related, should be more statistically significant in predicting the global mean relative temperature.
Home
When predicting total Watt hours (Wh_Total) for a given day at my house, a linear model consisting of the following variables returned a model with an adjusted R2 of 0.68:- DayOfWeek
- Month
- Temp_Max
- Temp_Min
- Temp_Avg
- Temp_Avg_AbsDiff65
- DewPoint_Avg
- DewPoint_Max
- DewPoint_Min
- Humidity_Min
- Humidty_Max
- WindSpeed_Max
- WindSpeed_Min
- Pressure_Max
- Pressure_Min
- Precip_Avg
Most variables were not significant, although the following were:
- DayOfWeekMonday - positively correlated and added much more than other days. This is our primary laundry day.
- Months: August, December, January, February - the hottest month and the coldest months in which the AC or heat is working the hardest. All were positively correlated and had much higher magnitudes than other months.
- Temperature: while the min and average were significant, the most important variable was Temp_Avg_AbsDiff65, that is, the absolute difference between the day's average temperature and 65 degrees. I had determined 65 degrees to be the temperature at which the AC/heater had the lowest energy usage, so any deviation above (AC) or below (heat) would be drawing more energy.
After having binarized the character variables, I ran StepAIC on the above model to reduce the variables to those that were most significant. A model with an adjusted R2 of 0.69 resulted. Most significant were:
- Temp_Avg_AbsDiff65 - positively correlated
- DayOfWeek_Monday - positively correlated and adding a large amount in magnitude.
- Months: September, October, March, April, May: negatively correlated, accounting for the milder weather and reduced need for heating and cooling (and thus, less need for energy usage).
- Intercept: 31,972 watts per day
- Temp_Avg_AbsDiff65: 738 watts added per degree away from 65 degrees
- Dewpoint_Avg: -395 watts added per degree
- Humidity_Max: 197 watts added per humidity percent
- Days:
- DayOfWeek_Monday: 5637 watts added
- DayOfWeek_Wednesday: 1995 watts added
- DayOfWeek_Saturday: 2125 watts added
- Months
- September through November: negatively correlated, with October contributing -9131 watts to the model.
- March through June: negatively correlated, with March and May contributing -8366 and -8026 watts respectively.
- Temperature/weather (including dewpoint, humidity, and month)
- Hot water usage (inferred from day of week):
- Monday - primary laundry day
- Wednesday and Saturday - primary bath days for kids
Conclusion
The impact of increased global temperatures and ways of responding to these increased global temperatures is a matter of environmental concern and of social justice. Consequently, one should care about energy production and consumption in relationship to global warming, for if one can reduce energy production and consumption (at least where such production and consumption comes from CO2 emission producing processes), one should be able to reduce one's impact on global warming if CO2 emissions are the primary cause of increasing global temperatures.If this is true, what can a responsible average person do to reduce global warming? A straightforward thing to do is to reduce one's energy consumption, thus reducing the need to produce that energy by burning fossil fuels that produce CO2 emissions. How can one reduce one's energy consumption? One must first understand where that energy consumption is largely coming from and what factors play into increasing or decreasing energy consumption. Once understood, one can determine how best to reduce one's energy consumption by targeting these contributing factors.
Understanding these factors is what I have attempted to do in this project. While a more detailed analysis could be undertaken to more specifically understand the causes of energy consumption in my home, it seems that the primary sources of energy consumption come from heating/cooling the air inside the house (related to the temperature outside) and heating the water (related to hot water usage inferred from day and time of energy usage).
Having identified these factors, these can be targeted for reducing energy consumption in my home through:
- Better insulation, which would counteract the need for heating and cooling the home through heaters and air conditioners. This would make the energy usage less dependent on the outside temperature.
- More efficient appliances, especially, a hot water tank, that use less energy to produce the same result (e.g., hot water, a heated stove and oven).
No comments:
Post a Comment