A Predictive & Recommendative Model For PV Power Output

Inspiration

What motivated us to participate to this hackathon are our passion, determination, and curiosity towards science and technology to invent and develop intelligent systems and applications. These applications are to be used for the aid and ease of people all over the globe. We also want to use and devote our knowledge and skills to come up with significant contributions for Neom to achieve the vaunted 2030 vision of the kingdom.

What it does

It is a predictive model for residential solar power production. The end user (the prosumer) will be provided with the necessary information that is a main concern which is the power production per specific date and time. This predicted PV power output is based on real-time collected data. It is furthermore used to compute the aggregate monthly bill that they will pay to the service provider in case the solar production doesn't cover the load intake.

How we built it

The project has been built mainly by using Python Jupyter Nootbook. The system has a dataset containing information about various weather conditions that affect the production of the solar panel. The dataset is preprocessed with main features selected after correlation analysis and visualization of data. The main features are further standardized and trivial null data entries are removed for smaller bias in the model. The system decision depends on a machine learning technique called linear regression which is used for continuous output, in our case the power. The dataset was divided into 80% for training set and 20% for testing set as in standard Pareto form. Plots of the residual error is given, along with performance measure in the form of variance and root mean square error to check on system capabilities. The model takes into account date and time as well as temperature and other weather conditions and the PV module efficiency as inputs, and computes the resultant expected power output and the expected cost. The user interface is build using python through pycharm.

Challenges we ran into

There were two main challenges encountered during the developing stage of the project. The first challenge is with the provided data-set interpretation and understanding due to the lack of description provided over the data parameters. The second challenge is the preprocessing stage in which we need the dataset to transform before inputting it in the model so that we can get more accuracy. Finally, the third challenge is related to the limited background of the team about the programming language used for the developing the code. However, with patience, cooperation, passion, online tutorial and article and mentors' guides, the team has successfully overcame these challenges and achieve their desired goals and outcomes.

Accomplishments that we're proud of

As a team, it was our honor to participate in such an event that will contribute greatly in the development of the kingdom and the achievement of 2030 vision. We are extremely proud to be chosen and provided with this opportunity to share our valuable ideas with Neom.

What we learned

The PentaX team has learned a lot from this experience. We have gained a lot of knowledge about solar energy in general and solar panel production in specific as well knowledge in machine learning programming. We have met and talked to plenty of experts and professional mentors in and out of our main fields who helped us perceive a different horizon of the world.

What's next for Energy:A Predictive & Recommendative Model For PV Power Output

The future work and development of this project will mainly focus on training the system with real-time dataset from NEOM region to acquire more logical and reliable estimations and outputs. The system can be expanded to use dust sensor readings along with weather prediction information to determine the optimal-time for the autonomous cleaning system.The system also can send notification emails monthly for the subscribers to provide them statistical information about their power consumptions and production for the next month. In addition, the system suggests to the customers a list and numbers of appliances when the power production drops off.

Built With

• jupyter
• pycharm
• python