Project Peak

Inspiration

With the projected increase in electricity costs, Americans need to be more aware and conscious of their power expenditure. Keeping track of expenditure can lead to more efficient use of power, and also benefit the world globally as less electricity expenditure correlates to reduced pollution emitted from non-renewable sources of electricity, such as the fossil fuel required to generate our electricity. These days, with the use of appliances and devices, such as your energy sensors, lights, and thermostat, throughout your home, it can be hard to keep track of and maintain them all. In addition to bringing clients crystal clear insight into power usage, our project utilizes machine learning to streamline the monitoring and maintenance of your devices at home.

What it does

Our solution involves four main parts. We have a main dashboard that offers analytics for users regarding average overall energy usage per room, predictive usage over the year, as well as any updates to users such as maintenance updates or app updates. Our next main section is the energy page that illustrates the power usage for all your devices – sensors, outlets, utilities, and lights as well as the overall power usage. We then have a quality check page that relies on our ARIMA model to predict whether or not sensors or devices need to be inspected for repairs within the next 15 days. The prediction enables users to take preventative actions before issues arise. Lastly, our cost savings page shows users the cost over time, and key financial analytics such as cost over time, and savings per month. We even included a component that displays advice in order to reduce costs! We’ve even included settings and profile page in order to allow for push notifications and account settings, allowing users to customize their experience and make Peak as user-friendly as possible!

Key Features

–Real-time sensor data collection and monitoring –Predictive maintenance using the ARIMA model -Visually appealing frontend -Easily comprehensible graphs and analytics to offer excellent insight for users –Scalable for larger infrastructures as well as a larger market (for more and more clients) –Add feature to incorporate new devices –Easy integration with smart plugs, smart devices, etc. for remote access on the go -Settings and profile page for customizable push notification system

Our setup helps to prioritize our customers’ insights– offering easily comprehensible graphs and vital analytics that can be used to make better financial decisions. Our clean design offers easy access to all your devices remotely and streamlines maintenance.

How we built it

We developed the app’s frontend primarily using ReactJS. This approach enabled us to break down the page into individual components, allowing each team member to concentrate on creating different parts of the application. This resulted in an efficient and well-organized workflow. For the backend, we harnessed the power of Google Colab to build multiple models using three years’ worth of data from Arlington homes. These models enabled us to provide precise recommendations for energy consumption and effective preventive measures.

Challenges we ran into

Data Generation Limitations

We found it difficult to find large datasets online, and so we had to make the best out of what we had. Using NumPy on Python, we were able to generate ample data points using our original data set. This could lead to skewed or biased results in our mock dataset, but we chose to use it anyway. Next, we had to find data such that it would make sense in real time – for example, increased AC unit usage in the summer months, less overall power usage in the summer (as we are outside more), etc. We then utilized an ARIMA model to predict the quality of devices – based on our data regarding the cleanliness of device filters, its battery health, and date since last repaired, we are able to predict if devices need to be repaired soon.

Graph Components

Our next issue was creating graphs using react. We found it difficult to create a customized graph as it required a long time to understand it, and also whenever we tried to deploy the code, the graphs would work for a moment but then clear. Overall, we had some difficulty in deploying our components throughout the course of the project.

Another challenge we struggled with is deploying our project onto our domain– “sfpeak.tech.” SF Peak is short for State Farm PEAK, as we hope to offer our service to State Farm and its clients.

Accomplishments that we’re proud of

Our biggest accomplishment is easily submitting this finished product! We struggled a lot with frontend development and nearly called quits around 6 PM due to lack of progress frustration from being unable to produce any work. All we had done was an introduction slide and a “thank you” slide! However, we had to persevere. We believe that our idea needed to see the light of day and after taking a break, we used our newfound zeal to pursue our app development with tremendous progress.

Another major accomplishment is our dataset. Despite having limited data to work off of, the dataset generated is very accurate, as we can see trends in our graphs that logically make sense (in example, increased AC usage in summer due to hotter temperatures, increased light usage in the winter due to lower time frame of light outside, etc). This took a lot of time and effort and to see this work being utilized in our product is absolutely incredible!

We are proud of our capability to work together and create a fully operational product. We take great pride in the work we’ve submitted, and we really feel that our UI/UX for this app might be our best yet!

What we learned

We learned to utilize datasets to create a real time working model of our algorithms, we learned more about the implementation of the ARIMA model. We took a huge step in frontend development as we incorporated a lot of new react components that we’ve never worked with- new tables with conditional formatting, new graphs, and a pop-up modal to add new devices.

What’s next for PEAK

In the future, our company’s vision is to scale our services to a much larger scale to accommodate for not just larger homes, but even offices and enterprises. We hope to revolutionize industries with our technology to create much more efficient usage of power and keep track of power usage, in order to prevent paying such high fees. If we can save at least a few hundred dollars each year for a single home, the amount of money that can be saved when the client scales up can be immense. We want to be that difference in enterprises and even in local businesses. Next, we hope to get a larger data set to create better predictions and cost analysis. This can help us remove bias in our models and maximize cost savings. Lastly, we hope to utilize our ML model to actively control devices, turning devices on and off, turning down the thermostat, etc. such that cost efficiency is maximized, and automated. This automation will vary from user to user and will be customized off of routines users can set up, or even based off of habits tracked by sensors. We hope to personalize our software to our clients as much as possible, enhancing client experience.