Dear Thomas,

I have a research paper presently under review relating to your question. In my opinion, I would suggest that you have 3 inputs and one output to your Artificial neural network (ANN); the inputs being monthly temperature, monthly relative humidity and the corresponding months. Your output would be the monthly electricity consumption. One way of coding the months would be say Jan as ‘1’, Feb as ‘2’, March as ‘3’......Nov as ‘11’ and Dec as ‘12’. Note that including corresponding months as input attributes make much sense considering that the different months significantly influence electricity consumption. Going further, you will have to normalize the 3 inputs to the range 0 to 1 before feeding them as inputs to the ANN.

Input data coding

Note that you will have to normalize the 3 inputs using the range (or maximum) value for each input correspondingly. Never use a single value to normalize the 3 different inputs, as you create bias in the structure of the data. For example, to normalize the month inputs, you could use the maximum value '12' to divide all months input attributes. You find a suitable value for the temperature and relative humidity too, and then proceed with normalization just as for the months input data.

Output data coding

Method 1: You may normalize your desired output data (monthly electricity consumption) to the range 0 to 1, in which case you'd been using Log-Sigmoid activation function in your output layer. If you choose this method, note that you would have to rescale the outputs back to actual values when you test the trained network. I was absolutely fine with this option in my paper.

Method 2: You may use the monthly electricity consumption as they are (actual values) as desired outputs and then use a linear activation function in your output layer. Note that this option may make learning extremely difficult.

 Training and Testing ANN

After you have successfully prepared the data. i.e. normalization of inputs and may be output; you'd want to divide your data into training and testing data; optionally, you could have a validation data. Also, select a suitable ratio for training data: testing data. One more important point, in order to reduce the chance of the ANN only memorizing (over-fitting) your training data, you'd probably want to randomize the training data first before training.

Hidden layer

You will have to heuristically determine the suitable number of hidden neurons for properly learning the task. For example, increasing the number of hidden neurons gradually. Also, note that the activation function used in the hidden layer should allow the learning of non-linear functions; hence, you could use the Tangent-Sigmoid or Log-Sigmoid function, though I would strongly recommend the later in your situation.

Other network training heuristics

You still have to determine parameters such as learning rate, momentum rate, maximum number of epochs, required MSE (error).

 I hope this helps.

Oyebade Oyedotun

hi,

If I understand you correctly, you have 3-dimensional data (humidity, temperature, electricity load) for all month of the year from 2008 to 2015, so total raw data vector is 12*7.

one solution for your problem to use ANN, and build network structure, is to set input size to 2 and output size to 1, which each input vector is 2D vector of (temp, humidity) and output label is its corresponding electricity load

thus your network  structure should be like this:

[input layer size = 2 ; hidden layer size = * ; output layer size = 1]

you can randomly select 70% (as example ) of your data for train and other 30% for test to validate your method.

for best result its better to normalize your data before training.

Someone suggested I should use month (in numeric values ie January -1,feb-2,..,Dec-12),Average temperature,Relative humidity from 2008 to 2013 as input data (for training in nntools) and Electric load(2008 to 2013) as the target data,while months(numeric),year,Avearge temperature,Relative humidity and Electric load (2014 to 2015) as the test data.

Please I need help on what data do I use for Training using nntool?

How do I forecast 12 months (from January ....,Decemeber 2016)?

You actually need to divide the data (month, humidity and temperature) into two parts. Use one part for the training and the other part to test the ANN

yeah thomas,i thought about the idea which add a new dimension to input vector (month name) but i think it is redundant and does not useful for NN to modeling data,

after you done NN train, with nn tools, for prediction phase you must  firs save your trained network (for safety) then delete input data and target data from nntool window, at the end adds your prediction data for year 2016 as input data (which is a matrix with size of 2*12 and u dont have label for them),click on the network,select "simulate tab" then choose 2016 data mat as an inputs, when click  on simulate , you will give a vector by size 12 which is predicted electric loads form jan to dec (same as input order)

Hi,

If you expect that a given month influenence somehow the output i.e. electrical load it is better not to give numers for months, but to create the 14 element input layer (example for February data): 1 humidity; 2 temperaturę; 3 January (value 0); 4 Ferbruary (value 1); 5 March  (value 0);.... 14 December (value 0).

It will protect ANN from reading that December (twelve's month) influences output more than January (the first month).

Regards

Hubert

Dear Thomas,

I have a research paper presently under review relating to your question. In my opinion, I would suggest that you have 3 inputs and one output to your Artificial neural network (ANN); the inputs being monthly temperature, monthly relative humidity and the corresponding months. Your output would be the monthly electricity consumption. One way of coding the months would be say Jan as ‘1’, Feb as ‘2’, March as ‘3’......Nov as ‘11’ and Dec as ‘12’. Note that including corresponding months as input attributes make much sense considering that the different months significantly influence electricity consumption. Going further, you will have to normalize the 3 inputs to the range 0 to 1 before feeding them as inputs to the ANN.

Input data coding

Note that you will have to normalize the 3 inputs using the range (or maximum) value for each input correspondingly. Never use a single value to normalize the 3 different inputs, as you create bias in the structure of the data. For example, to normalize the month inputs, you could use the maximum value '12' to divide all months input attributes. You find a suitable value for the temperature and relative humidity too, and then proceed with normalization just as for the months input data.

Output data coding

Method 1: You may normalize your desired output data (monthly electricity consumption) to the range 0 to 1, in which case you'd been using Log-Sigmoid activation function in your output layer. If you choose this method, note that you would have to rescale the outputs back to actual values when you test the trained network. I was absolutely fine with this option in my paper.

Method 2: You may use the monthly electricity consumption as they are (actual values) as desired outputs and then use a linear activation function in your output layer. Note that this option may make learning extremely difficult.

 Training and Testing ANN

After you have successfully prepared the data. i.e. normalization of inputs and may be output; you'd want to divide your data into training and testing data; optionally, you could have a validation data. Also, select a suitable ratio for training data: testing data. One more important point, in order to reduce the chance of the ANN only memorizing (over-fitting) your training data, you'd probably want to randomize the training data first before training.

Hidden layer

You will have to heuristically determine the suitable number of hidden neurons for properly learning the task. For example, increasing the number of hidden neurons gradually. Also, note that the activation function used in the hidden layer should allow the learning of non-linear functions; hence, you could use the Tangent-Sigmoid or Log-Sigmoid function, though I would strongly recommend the later in your situation.

Other network training heuristics

You still have to determine parameters such as learning rate, momentum rate, maximum number of epochs, required MSE (error).

 I hope this helps.

Oyebade Oyedotun

Given the nature of your data, which seems to be a time-series, I would suggest to look at techniques to analyze time-series.

Hi, I think method suggested by Hubert should work better. Reason being, if you are assigning numerical values 1-12 to January-December, you are kind of saying that January (1) is more close to February (2) than December (12). So use one-hot vector representation for months.

I suggest to classify the months to 4 seasons, summer =4, spring=3, fall=2, and winter=0, as power consumption is related to those seasons. and add this to input layer with the temperature and humidity,  

Please note that classification of months to seasons is different from country to another.