Project
| # | Title | Team Members | TA | Documents | Sponsor |
|---|---|---|---|---|---|
| 28 | Electric Load Forecasting (ELF) System |
Ao Zhao Liyang Qian Yihong Jin Ziwen Wang |
Xiaoyue Li | design_document3.pdf final_paper2.pdf proposal1.pdf |
Ruisheng Diao |
| # Electric Load Forecasting (ELF) System # Team members: Ao Zhao, aozhao2 Ziwen Wang, ziwenw5 Liyang Qian, liyangq2 Yihong Jin, yihongj3 # Problem Electric load forecasting (ELF) is a method that takes into account unstable factors, such as weather conditions and electricity prices, to predict the demand for electricity. Many utility companies rely on manual forecasting techniques based on specific datasets, but these methods may lack accuracy when fine-grained time particle forecasting is required. To accurately predict expenses on electricity and construct reliable infrastructures that can withstand a certain electrical load, utility companies need more advanced and reliable forecasting methods. # Solution Overview The electric load forecasting system is a powerful tool for predicting future electric load usage based on dedicated hardware and AWS services. By combining the data collection subsystem, data storage subsystem, prediction subsystem, query API subsystem, and web page subsystem, customers can easily retrieve and visualize the predicted electric load usage and use it for planning and optimization purposes. The system is designed to be accurate, effective, reliable, and easy to use, providing customers with a complete solution for electric load forecasting. # Solution Components [Data Collection Subsystem] - This subsystem is responsible for collecting real-time data on electric load usage. The data collection hardware is designed to be reliable, scalable, and capable of handling large volumes of data. The collected data is then sent to the data storage subsystem for further processing through AWS IoT Core. Hardware I. Smart meters to collect voltage, current, power and other data which further improve the ability to collect information Hardware Ⅱ. A transmission communication device that connects a smart meter to software or concentrator Hardware Ⅲ. Sensors that collect some relevant external factor data data (ex. Temperature sensor) [Data Storage Subsystem] - This subsystem is responsible for storing the collected data in a secure, scalable, and durable storage system. The data is stored in a format that is compatible with the Forecast DeepAR+ algorithm. AWS S3 provides a highly available and cost-effective storage solution that is suitable for storing large volumes of data. [Prediction Subsystem] - This subsystem is responsible for generating accurate predictions of future electric load usage based on the collected data. The Forecast DeepAR+ algorithm is a state-of-the-art machine learning algorithm that is designed for time-series forecasting. The AWS Forecast service makes it easy to generate accurate predictions at scale. The output of this subsystem is a forecast of future electric load usage that can be used for planning and optimization purposes. [Query API Subsystem] - This subsystem provides a RESTful API that allows customers to retrieve the predicted electric load usage for a specified time period. The API is designed to be secure, scalable, and easy to use. Customers can send requests to the API with the necessary parameters, and the API will return the predicted electric load usage in a format that is easy to understand and use. [Web Page Subsystem] (Optional) - This subsystem provides a user-friendly web interface for accessing the predicted electric load usage. The web page is built on top of the query API and allows customers to easily select the time period they are interested in and view the predicted electric load usage in a graphical format. The web page is designed to be responsive, easy to use, and accessible from any device with a web browser. # Criterion for Success Accuracy: The system should generate accurate predictions of future electric load usage. The accuracy of the predictions should be high enough to enable effective planning and optimization of electric power usage. Scalability: The system should be capable of handling large volumes of data and generating predictions for a large number of electric load customers. The system should be able to scale up or down as the demand for electric power changes. Reliability: The system should be designed to be highly reliable and available. It should be able to handle failures gracefully and recover quickly from any disruptions in service. Security: The system should be designed to be secure and protect customer data from unauthorized access or disclosure. The system should use industry-standard encryption and access controls to protect customer data. Ease of Use: The system should be designed to be easy to use and accessible to a wide range of customers. The query API should be easy to understand and use, and the web page interface should be intuitive and user-friendly. Cost-Effectiveness: The system should be designed to be cost-effective and provide good value for money. The cost of running the system should be reasonable and should not be a significant barrier to adoption. # Distribution of Work Yihong Jin, Computer Engineering: As a [AWS Certified Solutions Architect - Professional](https://www.credly.com/badges/1e4aa7a1-3ee6-4dd8-94c5-c015a85c3b84/linked_in_profile), design and implement the software architecture of this solution based on AWS services. Responsible for building the data pipeline which ingest raw data from by dedicated hardwares and prepare it for Machine Learning model training. Liyang Qian, Computer Engineering: Train the deepAR+ model with data stored in AWS S3 and build the API to enable customers to take advantage of forecasting results. Ao Zhao, Ziwen Wang, Electrical Engineering : Design the hardware used to collect the data and connect smart meters and software through transmission devices or specific communication methods to realize data interaction between each other. |
|||||