Early attempt at combining weather and kwh models

R/clusterviz.R

@@ -33,6 +33,8 @@ aggdf$cluster <- factor(aggdf$cluster)
 str(aggdf)
 clusters = levels(aggdf$cluster)
 
+mtempdf <- read.csv("../data/weatherharm.csv", stringsAsFactors = FALSE) %>% mutate(x = as.POSIXct(x, tz = "UTC"))
+
 ggplot(aggdf, aes(y = kwh_tot_mean, x = cluster)) + geom_boxplot()
 
 facall <- ggplot(aggdf, aes(x = read_time, y = kwh_tot_mean, color = cluster, fill = cluster)) + 
@@ -260,7 +262,7 @@ hist(resid(cmodi), breaks = 100)
 qqnorm(resid(cmodi))
 
 
-clust = "9"
+clust = "5"
 
 cts <- filter(aggdf, cluster == clust)$kwh_tot_mean
 ctsy <- ts(cts, frequency = floor(48 * 365.25))
@@ -276,3 +278,35 @@ ggplot(cldf, aes(x = x, y = y)) + geom_line(aes(y = f), color = "blue", size = 2
     geom_point(aes(y = r), color = "darkgreen") +
     coord_cartesian(xlim = c(as.POSIXct("2017-01-01", tz = "UTC"), as.POSIXct("2017-12-31", tz = "UTC")))
 
+cwcombdf <- left_join(cldf, mtempdf, by = "x")
+str(cwcombdf)
+
+
+# A variety of models
+plot(f.x ~ f.y, data = cwcombdf)
+plot(r.x ~ r.y, data = cwcombdf)
+plot(r.x ~ y.y, data = cwcombdf)
+plot(y.x ~ r.y, data = cwcombdf)
+plot(y.x ~ x, data = cwcombdf)
+plot(y.x ~ y.y, data = cwcombdf)
+plot(y.x ~ y.y, data = cwcombdf)
+plot(y.y ~ r.x, data = cwcombdf)
+plot(y.y ~ x, data = cwcombdf)
+
+
+summary(lm(f.x ~ f.y, data = cwcombdf))
+summary(lm(r.x ~ r.y, data = cwcombdf))
+summary(lm(r.x ~ y.y, data = cwcombdf))
+summary(lm(y.x ~ r.y, data = cwcombdf))
+summary(lm(y.x ~ x, data = cwcombdf))
+summary(lm(y.x ~ y.y, data = cwcombdf))
+summary(lm(y.x ~ y.y, data = cwcombdf))
+summary(lm(y.y ~ r.x, data = cwcombdf))
+summary(lm(y.y ~ x, data = cwcombdf))
+
+summary(lm(r.x ~ r.y * f.y, data = cwcombdf))
+imod <- lm(r.x ~ r.y * f.y, data = cwcombdf)
+plot(fitted(imod))
+plot(resid(imod))
+plot(cwcombdf$r.x)
+plot(cwcombdf$r.x ~ resid(imod))