colasgael/machine-learning-for-solar-energy-prediction

Predict the Power Production of a solar panel farm from Weather Measurements using Machine Learning

Machine-Learning-for-Solar-Energy-Prediction

by Adele Kuzmiakova, Gael Colas and Alex McKeehan, graduate students from Stanford University

This is our final project for the CS229: "Machine Learning" class in Stanford (2017). Our teachers were Pr. Andrew Ng and Pr. Dan Boneh.

Language: Python, Matlab, R

Goal: predict the hourly power production of a photovoltaic power station from the measurements of a set of weather features.

This project could be decomposed in 3 parts:

• Data Pre-processing: we processed the raw weather data files (input) from the National Oceanographic and Atmospheric Administration and the power production data files (output) from Urbana-Champaign solar farm to get meaningful numeric values on an hourly basis ;
• Feature Selection: we run correlation analysis between the weather features and the energy output to discard useless features, we also implemented Principal Component Analysis to reduce the dimension of our dataset ;
• Machine Learning : we compared the performances of our ML algorithms. Implemented models include Weighted Linear Regression with and without dimension reduction, Boosting Regression Trees, and artificial Neural Networks with and without vanishing temporal gradient

Our final report and poster are available at the root.

machine-learning-for-solar-energy-prediction's Issues

.fit attribute

'function' object has no attribute 'fit' . Please resolve this issue.

the data shape problem during adaboost

hi, When I run your code, I have an error, expected adaboost.fit(train_data, train_labels) problem.
I turn off the vanilla lstm model line and turn on adaboost model.
Then the problem occurs . Expected 3 dimensions but got array with shape (,)
I think the 3 dimension data shape put in the lstm layer, but the adaboost might need 2 dimension shape.

How can I do.?
I know your changed code to apply adaboost model.
plz.

Just simple question about time step in the LSTM model :)

Hi,

I have tested the LSTM model according to time-step.
To change the number of time-step from None to 2, LSTM input_shape would be changed from (None,12) to (2, 12) ? or (2,6)? ,. I think (2,6) is right,
and if I wanna set the time-step to 2, I think the shape of X_train and X_test should be changed.

SO, if the time step is 2, LSTM input_shape is changed (2,6) and X_train and X_test are changed to (train_data.shape[0], 2,6).

But,, it isnt working! :( what is the problem?

ps. I think you set the LSTM input_shape (None,1) and, this means full sequence is used to time step?

Request for instructions on deploying the photovoltaic power station prediction model

Hi there,

I am interested in using the photovoltaic power station prediction model from this project for a web application I am building on .NET. I would like to request instructions on how to deploy the model and integrate it with my server.

To give some context, my goal is to create a web page where I can input weather features and receive a prediction for the hourly power production of a photovoltaic power station. I am not familiar with Python and would appreciate assistance in getting this set up.

From the project's Readme.md file, I understand that the model is divided into three parts: data pre-processing, feature selection, and machine learning. Could you please provide guidance on how to deploy each of these parts and how to bind the model with my server?

Thank you for your time and assistance.

Best regards,
Oliver Valiente Oliva