Project LERA

Project LERA, or Luzon Electricity Reliability Analysis, aims to assess electricity outage frequencies across Luzon from 2022 to 2023, comparing the performance of public and private electric cooperatives.

Luzon, home to gorgeous cities, bright lights, and cities that never sleep.

Have we taken it all for granted?

In order to sustain these cities, Luzon averages a consumption of 11,033 megawatts per day ^[1] . Last 2022, Luzon took up a share of 81% of the Philippines’ total electricity consumption ^[2] . This huge energy requirement highlights the critical need for robust and sustainable energy infrastructure to support it. As the region continues to expand, addressing the challenges of energy efficiency, renewable energy integration, and grid reliability becomes increasingly important to ensure a stable and sustainable future.

In many parts of the Philippines, access to reliable electricity poses a significant challenge. Frequent power outages disrupt daily life, impacting productivity, essential services, and overall quality of life for Filipino communities.

PIDS: LGUs losing P10 billion from single power outage

Louella Desiderio

January 21, 2024

Economy loses P556 million from power outage

Louella Desiderio

May 21, 2023

This is why we decided to look into this further.

Researching this enables communities to better prepare for and mitigate the impact of outages, while also allowing for the development of predictive models to prevent future disruptions. Ultimately, studying power outages contributes to ensuring reliable access to electricity, driving economic development, and improving the quality of life across the Philippines

Problem

Unplanned and planned power outages significantly disrupt the daily lives of Filipinos, particularly in areas where access to generators is uncommon.

Our Solution

We want to leverage data science to analyze patterns and trends. This may assist local operatives improve the reliability of power supply at critical times while also empowering citizens prepare strategies to lessen the impact of outages.

Primary

Is there any observable power outage trend shared by the eight electric cooperatives?

Null Hypotheses

There is no observable power outage trend shared by the eight electric cooperatives

Alternative Hypotheses

There is a observable power outage trend shared by the eight electric cooperatives

Secondary

Is there a difference with the power outage trends between the types of cooperatives, whether public or private?

Null Hypotheses

The power outage trend is not different between public and private cooperatives.

Alternative Hypotheses

The power outage trend is different between public and private cooperatives.

Our Action Plan

Analyze Luzon electric cooperatives' Facebook posts from 2022 to 2023 to depict power outage mentions through visualizations, aiding in situating the problem and identifying patterns.

Data Collection

We mined the internet for data.

Due to the limitation of language, we collected data from electric cooperatives in Luzon exclusively.

We were able to scrape the Facebook pages of 8 different electric cooperatives from Luzon to produce a dataset with over 12,000 entries.

Each with the following attributes:

Date

Exact date of when announcement was posted

Location

Place of power outage

Type of Cooperative

Either public or private

Name of Cooperative

Cooperative that posted

Analysis

This is what we found.

In order to answer our research questions, we created graphs to help visualize our data. Take a look at these!

The area graph illustrates the trends in electrical shortages for the eight electric cooperatives from Luzon over the period from 2022 to 2023.

IEC, BATANELCO, MASELCO, ALECO, and OEDC have the most stable scheduled power interruption which can be seen by the lack of bumps in the graph which is a sign of low variance, while ISELCOII, PANELCO3, and MERALCO have the highest variance in scheduled interruption. In particular, ISELCOII has large increases during 2023, and especially during August 2023.

There is no significant associations between the peaks of any two pairs of electric cooperative, except for MERALCO and MASELCO. However, the months with the most number of cooperatives having it as peaks are November 2023, November 2022, June 2022, February 2023, and July 2023. Furthermore, no electric cooperatives have significant associations between these peaks.

Finding #1

The only observable power outage trend shared by the electric cooperatives are from MERALCO and MASELCO. Otherwise, there is no observable power outage trend shared by the eight electric cooperatives.

The graph illustrates the peaks in electrical shortages for Private and Public categories over the period from 2022 to 2023.

The top five months with the most peaks are the March, June, July, September, and November of 2022. Chi-square tests reveal no significant difference in the distribution of peaks between Private and Public cooperatives, and only a weak evidence of an association between Private peaks and the top five peak months , which is not significant at the 0.05 level but is at the 0.10 level.

However, as can be observed in the graph above, Public cooperatives seem to have significant peaks which are much higher than their average peak (e.g. June 2023 an August 2023). Studying the drastic peaks of Public cooperatives can be a topic for another research.

No significant association was found between the peaks of public cooperatives and the top five peak months. Thus, the peaks distribution in Private and Public categories appears statistically similar, with weak evidence of temporal clustering in the Private category.

Finding #2

Null Hypothesis:

The power outage trend is not different between public and private cooperatives.

Modelling

We tried predicting the outages

With the help of our data, we trained a machine learning model that attempts to predict the amount of outages there will be given a month in the future.

We used a random forest regression as our Machine Learning Model of choice. The goal of using a Random Forest Regressor is to predict the number of power outages in a given month based on historical data, including features such as month, type of outage (IEC, etc.) and class (public or private).

The limitations of our model include:

Limited historical outage data. This impacts model’s accuracy and performance.
Data granularity (e.g., monthly vs. daily). This affects the model’s ability to capture finer details.

Why Random Forest Regression?

It combines the predictions of multiple decision trees to improve accuracy. By averaging multiple trees, it mitigates the risk of overfitting that a single decision tree might have. Provides insights into which features are most influential in making predictions. Captures complex relationships between features and the target variable.

Training Model Process

Prepared data by parsing dates, extracting features like month and year, and aggregating the number of outages.
Used One-hot encoding to convert categorical variables into numerical format.
Split data into training and testing sets to train the model.
The Random Forest Regressor was trained on the training data and evaluated using certain metrics

Results & Discussion

Mean Absolute Error (MAE)

13.8

Mean Squared Error (MSE)

398.87

Root Mean Squared Error (RMSE)

19.97

Discussion of Results

MAE: On average, the model’s predictions are off by about 14 outages.

RMSE: The model’s predictions have an average error of approximately 20 outages, indicating larger errors are slightly more penalized.

In conclusion,

With the data we gathered, the only correlation found between the power outage trends of the eight electrical cooperatives was between MASELCO, who services Masbate, and MERALCO, who services Metro Manila. All the other cooperatives have their own unique power outage trend.

Although there is no significant difference between the frequency of outages between public and private cooperatives, it was observed that Public cooperatives have peaks much higher than their average.

It’s important.

By leveraging data science to understand the trends of power outages, local government units (LGUs), electric cooperatives and Filipino citizens can understand the supply of electricity better. Understanding this can aid in:

Making informed investments on alternative reliable energy sources and electrical grid upgrades
Building more climate-resilient electrical grids
Mitigating social and economic losses from the loss of power

As we start to improve the supply of power in the Philippines, we can then start to make major strides towards being a technologically advanced country. Stable power supply across the entire nation allows for reliable internet and connectivity which opens up many opportunities for different industries to flourish. It does not have to be exclusive in Metro Manila. This can help, not only the issue of decongesting Metro Manila, but also improve the lives of the ones outside it.

But until then, can we really consider the Philippines to be ready for the Digital Age if we cannot even provide stable electricity?

We have a few recommendations.

For further studies, we suggest the following:

Collecting data from the electric cooperatives themselves . This allows the data to span to longer periods and can also help in making sure that the data is completely reliable. Otherwise, due to some electric cooperatives releasing their announcements through images or other social media channels, it would be good to also scrape the information from these. A more complete dataset may uncover power outage trends or confirm that there is no correlation at all.
A comparative study which maps the climate/weather to the power outages. Considering that the Philippines is in the Pacific Ring of Fire and a vulnerable target to natural calamities, it would be good to see how this plays a role in affecting the reliability of power supply.
Starting with a smaller, but geographically tighter, scope. Since it is possible that geographically closer regions share the same power sources, it would be good to take a look at the similarities between trends there. Our study was limited by the scraper only working for Facebook posts in text format, hence the scatteredness of the electric cooperatives geographically.

In summary,

There is still a lot more to be studied before the Philippines can achieve SDG 7 which is to ensure access to affordable, reliable, sustainable and modern energy for all. As not many studies like this were performed, Project LERA , aims to serve as the groundwork for the many researches to come. The only power outage trend that showed significant correlation was between MASELCO and MERALCO which are 500 kilometers away from each other. As this could be by coincidence alone, the team urges other data scientists to consider our recommendations to allow for a more complete understanding of power outages.

About Us

Rae Gabriel Samonte

What's up everyone! Gab here, currently a graduating student from the University of the Philippines Diliman. I am a competitive programmer; I like solving problems through code.

Outside of computer science, I am a big sports fan - Table Tennis, Basketball, Roundnet, and Cornhole are my go-tos. I also like to read manhwa and Isekai's are my favorite.

resamonte@up.edu.ph

Marc Macaraeg

Hello! I'm Marc, a computer science student from the University of the Philippines Diliman who loves both software and hardware side in code.

In terms of software, I have developed visualization solutions for the web. For hardware, I love tinkering with microelectronics and the Internet of Things. I am an avid reader and you can ask me the last thing I read on my Kindle!

mcmacaraeg2@up.edu.ph

Miko Surara

Hey! I'm Miko, a computer science student from the University of the Philippines Diliman who loves to solve problems, and maybe code.

I'm a former competitive programmer who had a serious phase in studying algorithmic code and currently, I'm a curious spirit who is interested in things tech and roguelike.

rcsurara1@up.edu.ph

Medwin Devilleres

What’s good! I’m Medwin, a 4th year Computer Science student from UP Diliman. I’m a big fan of web development and data science.

I enjoy playing sports and video games in my free time. Those things satisfy my competitive nature. If I’m not doing any of those, I’m probably eating outwith friends or at the gym.

madevilleres@up.edu.ph