Energy Demo

This is a script for Energy Forecast Demo

Contents

• Import Data
• Quick Visualization
• Basic Statistics
• Intermediate Statistics
• Automating Tasks
• Cell Execution and Publish
• Reusable functions
• GUI

Import Data

Copy and paste data from spreadsheet. Do this interactively. Later, I can explain about the programmatic methods (seen below).

• energyData - from SOP sheet
• DayType, HDD, Temp

    energyData= xlsread('January.xls', 'Liability', 'B2:Y32');
    DayType   = xlsread('January.xls', 'Weather'  , 'B2:B32');
    HDD       = xlsread('January.xls', 'Weather'  , 'C2:C32'); %Heating Degree Days
    Temp      = xlsread('January.xls', 'Weather'  , 'D2:D32');

Quick Visualization

Point-n-click visualizations

• Select energyData and click on plot icon - first plot, but not useful
• plot(energyData') to plot on a hourly basis
• Open Plot Catalog to do other visualizations - ribbon, waterfall

Customize using plottools

• Create initial plot using mesh
• Open plottools
• Add a second axes
• Use contourf for second axes
• Add colorbar
• Add annotations (labels, titles)
• Close plottools and generate m-file
• Run auto-generated m-file

The contour plot gives us more insights to the data. I can see the two peaks during the day. I can also deduce cyclical behavior during the month.

    myCreateFigure(energyData);

Basic Statistics

Start with simple statistics (some command line, some interactive)

• Let's determine the mean hourly profile:

    meanDay = mean(energyData);

• Plot all data and mean (thick line)

    figure;
    plot(energyData', 'green');
    hold;
    plot(meanDay, 'linewidth', 4);

Current plot held

• Examine the distribution of the variance:

    res = energyData - repmat(meanDay, 31, 1);

• Distribution Fitting Tool:

  dfittool(res(:))

• Fit a normal distribution.
• Generate m-file. Run m-file.

    myDistFit(res);

Intermediate Statistics

Now that we know the distribution is normal, we can carry on with more statistics assuming a normally distributed data set.

Let's see if we can determine the 95% confidence interval of the mean profile.

• Go to the doc and search for "confidence intervals normal distribution"
• The first entry is a page on "confidence intervals"
• Go down half a page where they have an example for normfit
• Highlight the 2 lines and "Evaluate"

    data = normrnd(10,2,100,2);
    [mu,sigma,muci,sigmaci] = normfit(data)

mu =
       9.8437       9.9803
sigma =
       1.9138       1.9955
muci =
       9.4639       9.5843
       10.223       10.376
sigmaci =
       1.6803        1.752
       2.2232       2.3181

• Explain what it does. Now apply to data.

    [mu,sigma,muci,sigmaci] = normfit(energyData);
    [mu2,sigma2,muci2,sigmaci2] = normfit(energyData');
    plot(mu);
    hold on;
    plot(muci', 'r');

• I have a function that does this for both hour and day dimensions and visualizes them. If appropriate, open plotCI.m and explain.

    plotCI(energyData,mu,sigma,muci,mu2,sigma2,muci2);

Automating Tasks

We have done a lot of interactive exploration of our data, and now I would like to start capturing my workflow so that the same processes can be applied to new data sets.

• Go to command history and capture appropriate commands and create M-file.
• Run m-file.
• Create cells by partitioning with cell headings. Here's an example:

    type myscript_template

%% Import Data
energyData= xlsread('January.xls', 'Liability', 'B2:Y32');
DayType   = xlsread('January.xls', 'Weather'  , 'B2:B32');
HDD       = xlsread('January.xls', 'Weather'  , 'C2:C32');
Temp      = xlsread('January.xls', 'Weather'  , 'D2:D32');

%% Visualize
createfigure(energyData)

%% Examine Distribution
averageProfile = mean(energyData);
residual = energyData - repmat(averageProfile, 31, 1);
mydist(residual)

%% Determine Confidence Interval
% search for 'confidence intervals normal distribution'
[mu,sigma,muci,sigmaci] = normfit(energyData);
plot([mu;muci]')

Cell Execution and Publish

Step through more detailed analysis using cell mode. Open EnergyForecastAnalysis.m. Publish the document.

Reusable functions

• First make EnergyForecastAnalysis.m a function with data file name as an input argument.
• Add a simple user interface:

     if nargin == 0
       dataFileName = uigetfile('*.xls', 'Select data file');
     end

GUI

• Open GUIDE and create a single button GUI and call EnergyForecastAnalysis.m

• Open final GUI energyForecastGUI.m and run

energyForecastGUI([], {'January.xls', 9, 1, 1, '10, 15'});

Published with MATLAB® 7.3