s <- suppressPackageStartupMessages
s(library(zoo))
s(library(data.table))
s(library(lubridate))
s(library(knitr))
s(library(ggplot2))

1 Introduction

This notebook uses the outdoor temperature, indoor temperature and indoor airspeed to calculate:

Outdoor Rolling Average over last 14 days
Upper and Lower Bound of Comfort Level
Number of Indoor Temperature Degrees Off the Upper or Lower Comfort Bound

# Parameters
n_hours_in_every_step <- 366 #Rolling mean will be based of the previous 14 days measurements

2 Method

2.1 Perform Average by Hour

The first step to the code is to construct the outdoor rolling mean. In order to do that, first the outdoor hourly measurement is averaged in hours.

# Load and rename columns for consistency -- no space and no upper case column names
outdoor <- fread("~/Github/erdl/thermal_comfort/adaptive_method/data/houses_outdoor_temperature.csv")
colnames(outdoor) <- c("timestamp","outdoor_temperature_f")

# Identify the right types for each column
outdoor$timestamp <- ymd_hms(outdoor$timestamp,tz="HST")
outdoor$outdoor_temperature_f <- as.numeric(outdoor$outdoor_temperature_f)
outdoor$ymd_h <- outdoor$timestamp

#Ensure timestamps are ordered. Database does not guarantee that.
outdoor <- outdoor[order(timestamp)]

# Reset the minute and second to 0, so we can group by ymd_h. 
minute(outdoor$ymd_h) <- 0
second(outdoor$ymd_h) <- 0
outdoor <- outdoor[,.(outdoor_temperature=mean(outdoor_temperature_f)),by="ymd_h"]
kable(head(outdoor))

ymd_h	outdoor_temperature
2017-06-09 14:00:00	74.96525
2017-06-09 15:00:00	74.63462
2017-06-09 16:00:00	74.31538
2017-06-09 17:00:00	74.83525
2017-06-09 18:00:00	74.88375
2017-06-09 19:00:00	74.44012

2.2 Perform 14 Days Hourly Rolling Average

Next, the hourly temperature is used to calculate the 14 days temperature backwards.

# Invert table so slide window goes backwards
inverted_outdoor <- outdoor[order(-ymd_h)]

# Calculate rolling mean and reverts back the output from the rolling mean. 
rolling_mean_outdoor_temperature <- rev(rollapply(inverted_outdoor$outdoor_temperature,
                                            width=n_hours_in_every_step,
                                            FUN=mean))

#
outdoor <- outdoor[n_hours_in_every_step:.N]  
outdoor$rolling_mean_outdoor_temperature <- rolling_mean_outdoor_temperature
colnames(outdoor) <- c("timestamp","outdoor_temperature","rolling_mean_outdoor_temperature")

2.3 Simulation Data

To calculate the boundaries and acceptance, besides the outdoor rolling average, we also need the indoor temperature, and air speed. Because the tables are different, i.e. indoor’s temperature and airspeed are hourly, wherewas the outdoor was originally at a minute sampling rate, there is a chance after the rolling average is calculated the number of rows will not match between both tables. This script matches the minimum number of rows between them both.

# load indoor data
indoor <- fread("~/Github/erdl/thermal_comfort/adaptive_method/data/house2_livingroom_temperature.csv")[,.(reading_timestamp,value)]
colnames(indoor) <- c("timestamp","indoor_temperature")
indoor$timestamp <- ymd_hms(indoor$timestamp,tz="HST")

## Date in ISO8601 format; converting timezone from UTC to "HST".

# Identify the right types for each column
indoor$timestamp <- ymd_hms(indoor$timestamp,tz="HST")
indoor$outdoor_temperature_f <- as.numeric(outdoor$indoor_temperature)
indoor$ymd_h <- indoor$timestamp

#Ensure timestamps are ordered. Database does not guarantee that.
indoor <- indoor[order(timestamp)]

# Reset the minute and second to 0, so we can group by ymd_h. 
minute(indoor$ymd_h) <- 0
second(indoor$ymd_h) <- 0
indoor <- indoor[,.(indoor_temperature=mean(indoor_temperature)),by="ymd_h"]
colnames(indoor) <- c("timestamp","indoor_temperature")

dt <- merge(outdoor,indoor,by="timestamp")

# Add Rolling Mean Vector to Simulation Data Column-Wise. Notice this is NOT an inner join.
#min_rows <- min(length(rolling_temperature_f_mean),nrow(simulation_data))

#dt <- simulation_data[1:min_rows]
#dt$rolling_temperature_f_mean <- rolling_temperature_f_mean[1:min_rows]

2.4 Calculate boundaries

The comfort level boundaries are calculated using the outdoor rolling average adjusted by the indoor airspeed.

calculate_air_speed_adjustment <- function(in_air_speed){
  return(0.0153*in_air_speed+0.4333)
}

dt$upper_bound <- 0.31*dt$rolling_mean_outdoor_temperature + 60.5
dt$upper_bound_0_fpm <- dt$upper_bound + calculate_air_speed_adjustment(0)
dt$upper_bound_120_fpm <- dt$upper_bound + calculate_air_speed_adjustment(120)
dt$upper_bound_200_fpm <- dt$upper_bound + calculate_air_speed_adjustment(200)


dt$lower_bound <- 0.31*dt$rolling_mean_outdoor_temperature + 47.9

2.5 Calculate Acceptance

With the comfort upper and lower bound, we then compare the indoor temperature against them.

dt$is_acceptable_80 <- NA_character_
dt[indoor_temperature > upper_bound]$is_acceptable_80 <- "Unacceptable Hot"
dt[indoor_temperature < lower_bound]$is_acceptable_80 <- "Unacceptable Cold"
dt[indoor_temperature > lower_bound & indoor_temperature < upper_bound]$is_acceptable_80 <- "Acceptable"

2.6 Calculate Degrees Off

In addition, how far higher or lower than the upper and lower bound respectively can also be calculated.

dt$degrees_off <- NA_real_

dt[indoor_temperature > upper_bound]$degrees_off <- dt[indoor_temperature > upper_bound]$indoor_temperature - dt[indoor_temperature > upper_bound]$upper_bound

dt[indoor_temperature < lower_bound]$degrees_off <- dt[indoor_temperature < lower_bound]$indoor_temperature - dt[indoor_temperature < lower_bound]$lower_bound

dt[indoor_temperature > lower_bound & indoor_temperature < upper_bound]$degrees_off <- 0

3 Output Table

Finally, we the output table containing all the calculated information.

kable(head(dt))

timestamp	outdoor_temperature	rolling_mean_outdoor_temperature	indoor_temperature	upper_bound	upper_bound_0_fpm	upper_bound_120_fpm	upper_bound_200_fpm	lower_bound	is_acceptable_80	degrees_off
2017-10-06 12:00:00	77.18779	82.29032	84.73325	86.01000	86.44330	88.27930	89.50330	73.41000	Acceptable	0.0000000
2017-10-06 13:00:00	76.74112	82.27104	85.97350	86.00402	86.43732	88.27332	89.49732	73.40402	Acceptable	0.0000000
2017-10-06 14:00:00	76.38004	82.25701	87.15250	85.99967	86.43297	88.26897	89.49297	73.39967	Unacceptable Hot	1.1528280
2017-10-06 15:00:00	76.25629	82.24737	87.30000	85.99669	86.42999	88.26599	89.48999	73.39669	Unacceptable Hot	1.3033147
2017-10-06 16:00:00	76.15875	82.23663	87.22050	85.99336	86.42666	88.26266	89.48666	73.39336	Unacceptable Hot	1.2271438
2017-10-06 17:00:00	75.58500	82.22458	86.82250	85.98962	86.42292	88.25892	89.48292	73.38962	Unacceptable Hot	0.8328792

4 Plot

n_acceptable <- nrow(dt[is_acceptable_80 == "Acceptable"])
n_unacceptable <- nrow(dt[is_acceptable_80 != "Acceptable"])

st <- min(dt$timestamp)
et <- max(dt$timestamp)
p_start_time <- paste0(year(st),"-",month(st),"-",day(st))
p_end_time <- paste0(year(et),"-",month(et),"-",day(et))



p <- ggplot(data = dt) + 
  geom_point(aes(rolling_mean_outdoor_temperature,indoor_temperature,color=is_acceptable_80)) + 
  geom_line(aes(rolling_mean_outdoor_temperature,upper_bound_200_fpm),color="black") + 
  geom_line(aes(rolling_mean_outdoor_temperature,upper_bound_120_fpm),color="black") + 
  geom_line(aes(rolling_mean_outdoor_temperature,upper_bound_0_fpm),color="black") + 
  geom_line(aes(rolling_mean_outdoor_temperature,upper_bound),color="black") + 
  geom_line(aes(rolling_mean_outdoor_temperature,lower_bound),color="black") + 
  
  geom_text(aes(max(dt$rolling_mean_outdoor_temperature)+0.2,max(dt$upper_bound_200_fpm), label = "fpm 200", vjust = -1), size = 3) + 
  geom_text(aes(max(dt$rolling_mean_outdoor_temperature)+0.2,max(dt$upper_bound_120_fpm), label = "fpm 120", vjust = -1), size = 3) + 
  geom_text(aes(max(dt$rolling_mean_outdoor_temperature)+0.2,max(dt$upper_bound_0_fpm), label = "fpm 0", vjust = -1), size = 3) + 
  geom_text(aes(max(dt$rolling_mean_outdoor_temperature)+0.2,max(dt$upper_bound)-0.6, label = "no adj", vjust = -1), size = 3) + 
  geom_text(aes(max(dt$rolling_mean_outdoor_temperature)+0.2,max(dt$lower_bound), label = "no adj", vjust = -1), size = 3) + 
  
  
  theme_minimal() + 
  scale_colour_manual(values=c("Unacceptable Cold"="#0072B2","Acceptable"="#009E73","Unacceptable Hot"="#D55E00")) +
  ggtitle("Adaptive Method - House 2 Living Room", 
          subtitle = paste0(p_start_time," to ",p_end_time," | Acceptable: ",n_acceptable," ; Unacceptable: ",n_unacceptable)) + 
  ylab("Indoor Temperature (F)") + 
  xlab("Rolling Mean Outdoor Temperature (F)") 

p

Adaptive Method

Carlos V. Paradis

April 19, 2018