failureModelTest.Rmd

---
title: "failureModelTest"
author: "GeoffRussell"
date: "2023-11-24"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(tidyverse)
```

## Failure model test

This file is a small set of test data for testing failure model code

First we specify a set of PV panels or anything else installed over time

```{r data}
df<-tribble(
  ~panels, ~year,
  5, "2000",
  9, "2001",
  20, "2002",
  35, "2003",
  40, "2004",
  56, "2005",
  59, "2006",
  69, "2007",
)
```


```{r}
makeWeibullFail<-function(mtbf,k) {
  lambda<-1/mtbf
  function(n,p) {
    (1-exp(-(lambda*(n-1))^k))*p
  }
}
makeFail<-function(mtbf) {
  lambda<-1/mtbf
  function(n,p) {
    (1-exp(-lambda*(n-1)))*p
  }
}
#----------------------------------------------------------------------------------
# we make two functions, the normal exponential failure and the weibull failure
# with a parameter of 5.  
#----------------------------------------------------------------------------------
failFun<-makeFail(5)
failWeibullFun<-makeWeibullFail(5,5)
nyears<-length(df$panels)
failed<-rep(0,nyears)
wfailed<-rep(0,nyears)
print(nyears)
df <- df %>% mutate(produced=panels-lag(panels))
df$produced[1]=df$panels[1]
for(i in 1:nyears) {
    if (i<nyears) {
      for(n in (i+1):nyears) {
        failed[n]<-failed[n]+failFun(n,df$produced[i])
        wfailed[n]<-wfailed[n]+failWeibullFun(n,df$produced[i])
        cat(paste0("i,n,panels[i],failed[n]:",i,",",n,",P=",df$produced[i],",F=",failed[n],",WF=",wfailed[n],"\n"))
      }
    }
}
```

** Let's plot the failure function

```{r}
nyears<-100
failFun<-makeFail(40)
failFun<-makeWeibullFail(40,1)
failWeibull5Fun<-makeWeibullFail(40,5.3)
failWeibull3Fun<-makeWeibullFail(40,2.4)
failed<-rep(0,nyears)
w5failed<-rep(0,nyears)
w3failed<-rep(0,nyears)
df<-tibble(
  panels=c(100,rep(0,nyears-1)),
  produced=c(100,rep(0,nyears-1))
)
for(i in 1:nyears) {
    if (i<nyears) {
      for(n in (i+1):nyears) {
        failed[n]<-failed[n]+failFun(n,df$produced[i])
        w5failed[n]<-w5failed[n]+failWeibull5Fun(n,df$produced[i])
        w3failed[n]<-w3failed[n]+failWeibull3Fun(n,df$produced[i])
        #cat(paste0("i,n,panels[i],failed[n]:",i,",",n,",P=",df$produced[i],",F=",failed[n],",WF=",wfailed[n],"\n"))
      }
    }
}
dfout<-tibble(
  expfail=failed,
  w5fail=w5failed,
  w3fail=w3failed,
  x=seq(1,100)
)
dfout %>% ggplot()+
  geom_line(aes(x=x,y=expfail),color="blue") +
  geom_line(aes(x=x,y=w3fail),color="red")+
  geom_line(aes(x=x,y=w5fail),color="orange")+
  annotate('text',x=24,y=50,label="k=1",color="blue")+
  annotate('text',x=22,y=25,label="k=2.4",color="red")+
  annotate('text',x=32,y=12,label="k=5.3",color="orange")+
  labs(x="Years",y="Percentage Failed",title="Various failure models with average panel life of 40 years,
Exponential (k=1),Weibull quality (k=3),Weibull best quality (k=5)")
```

Add some testing code to predict future growth

```{r}
library(modelr)
dfgw<-read_csv("GWByCountry.csv") %>% mutate(Country=str_replace(Country,"Total ",""))
countries<-dfgw$Country %>% unique
regionGW<-dfgw %>% filter(Country=="Australia") %>% 
  mutate(cumGWinstalled=GW) %>% select(cumGWinstalled,Year) 
model<-lm(cumGWinstalled~poly(as.numeric(Year),2),regionGW)
modelGW<-bind_rows(tibble(Year=seq(2023,2050),cumGWinstalled=rep(0,2050-2023+1)),regionGW)
growthPrediction<-modelGW %>% add_predictions(model,var="Predicted") %>% arrange(Year)

growthPrediction %>% ggplot() + 
  geom_col(aes(x=Year,y=cumGWinstalled),fill="blue") + 
  geom_line(aes(x=Year,y=Predicted),color="red")

df2<-read_csv("SolarWaste/df2.csv") %>% pivot_wider(names_from=`State`,values_from=`GW`) 
model<-lm(cumGWinstalled~poly(as.numeric(Year),2),df2[1:18,])
df2<-read_csv("SolarWaste/df2.csv")
df3 <- df2 %>% add_predictions(model,var="Predicted") 
write_csv(df3,"df3.csv")
df3 %>% ggplot() + 
  geom_col(aes(x=Year,y=cumGWinstalled),fill="blue") + 
  geom_line(aes(x=Year,y=Predicted),color="red")



```