The Case of the Curious Correlations

Post on 10-May-2015

92 views 2 download

Tags:

description

When it comes to energy business, and especially electricity, things can get a little odd sometimes. Higher temperatures mean people are going to need more power. Lower temperatures, less power. Right. Usually. But not always. Sometimes you might need an expert to make sense of your data.

Transcript of The Case of the Curious Correlations

The Case of the Curious Correlations

Is this what you would expect?

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

The miracle of air conditioning

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)# Higher temperatures lead

to higher HVAC load

Lower temperatures lead to lower HVAC load

But what about this ?

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

Why would electric demand start to rise again as the temperature continues to fall ?

And why the weaker correlation ?

Electric heating ? Probably not too much – this is Texas.

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)# R²#=#0.32024#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q1?2012)#

R²#=#0.93592#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q2?2012)#

R²#=#0.90115#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q3?2012#

R²#=#0.45417#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q4?2012#

Digging into the data

The tight, positively correlated data is concentrated in Q2 and Q3

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)# R²#=#0.32024#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q1?2012)#

R²#=#0.93592#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q2?2012)#

R²#=#0.90115#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q3?2012#

R²#=#0.45417#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#

Q4?2012#

Digging into the data

The weaker, negatively correlated data is concentrated in Q1 and Q4

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

R²#=#0.20403#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

January#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

February#

R²#=#0.39612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

March#

R²#=#0.83112#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

April#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

May#R²#=#0.90612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

June#R²#=#0.76431#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

July#R²#=#0.93766#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

August#

R²#=#0.96721#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

September#

R²#=#0.72922#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

October#

R²#=#0.25539#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

November#

R²#=#0.58202#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

December#

Digging deeper into the data

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

R²#=#0.20403#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

January#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

February#

R²#=#0.39612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

March#

R²#=#0.83112#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

April#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

May#R²#=#0.90612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

June#R²#=#0.76431#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

July#R²#=#0.93766#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

August#

R²#=#0.96721#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

September#

R²#=#0.72922#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

October#

R²#=#0.25539#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

November#

R²#=#0.58202#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

December#

Digging deeper into the data

April looks odd, compared to March and May. Investigate further by looking at 2011 and 2013 data.

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

R²#=#0.20403#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

January#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

February#

R²#=#0.39612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

March#

R²#=#0.83112#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

April#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

May#R²#=#0.90612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

June#R²#=#0.76431#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

July#R²#=#0.93766#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

August#

R²#=#0.96721#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

September#

R²#=#0.72922#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

October#

R²#=#0.25539#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

November#

R²#=#0.58202#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

December#

Digging deeper into the data

Temperature is dominant driver of electric load in some months . . .

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

R²#=#0.20403#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

January#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

February#

R²#=#0.39612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

March#

R²#=#0.83112#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

April#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

May#R²#=#0.90612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

June#R²#=#0.76431#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

July#R²#=#0.93766#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

August#

R²#=#0.96721#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

September#

R²#=#0.72922#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

October#

R²#=#0.25539#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

November#

R²#=#0.58202#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

December#

Digging deeper into the data

But understanding what drives loads in other months requires more sophisticated models . . .

R²#=#0.88355#

0#

5,000#

10,000#

15,000#

20,000#

25,000#

30,000#

20# 40# 60# 80# 100#Temperature)@)DFW)(degrees)F))

2012)Daily)Peak)Electric)Demand)(ERCOT#North#Central,#MW)#

R²#=#0.20403#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

January#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

February#

R²#=#0.39612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

March#

R²#=#0.83112#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

April#

R²#=#0.55973#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

May#R²#=#0.90612#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

June#R²#=#0.76431#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

July#R²#=#0.93766#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

August#

R²#=#0.96721#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

September#

R²#=#0.72922#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

October#

R²#=#0.25539#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

November#

R²#=#0.58202#

0#

10,000#

20,000#

30,000#

20# 40# 60# 80# 100#

December#

Digging deeper into the data

July correlation significantly weaker than other summer months. Could it be due to Independence day falling on a Wednesday in 2012 ?

Looking for answers about energy and markets ? info@eRiskAnalytics.com

AnalyticsEnergy Risk

Uncertainty: measured, modeled, managed

PHILIP DIPASTENA(972) 656-9720info@eRiskAnalytics.com