✅ Keywords:
- ➡️ #EnergyConsumptionPrediction #Energy #ESP32 #PZMT004 #OpenSource #OpenHardware #RegressionLearner #SustainableEnergyInformation #DataCenter #NodeRed #EmbeddedSystems #MIcrocontroller #Phyton #MQTT #MySQL #Telegram
✅ Introduction:
- In this edition, we explore a groundbreaking research study titled "Learning-based Energy Consumption Prediction" conducted by our team of experts. We delve into the challenges posed by the increasing influx of data into cloud-fog infrastructures and the pressing need for sustainable energy consumption management. Our researchers have developed an energy consumption prediction model, focusing on hardware design, data pre-processing, and machine learning techniques. Join us on this journey as we discuss the methodology, findings, and future prospects of this pioneering study.
✅ Content:
- In our study, titled "Learning-based Energy Consumption Prediction," we address the critical issue of managing energy consumption in servers, fog devices, and cloud computing platforms. The growing demand for cloud-fog infrastructure has intensified the challenges associated with Green IT. To tackle this, we propose an innovative energy consumption prediction model. This model comprises hardware design, data pre-processing, and characteristic extraction, aiming to create a non-invasive meter utilizing a network of sensors. These sensors, equipped with microcontrollers and the MQTT communication protocol, measure various parameters such as voltage, current, power, frequency, energy, and power factor. The real-time energy measurements are then displayed on a dashboard, empowering users to optimize their IT equipment and reduce maintenance costs.
- During our research, we evaluated different linear regression models to select the most suitable one. These models enable us to predict energy consumption on an hourly and daily basis, providing valuable insights for effective energy management strategies. Our supervised machine learning algorithms have paved the way for accurate predictions, guiding businesses and individuals in optimizing their energy usage patterns.
✅ Conclusions:
- In our study, we introduced a hardware-based prototype capable of recording extensive measurements from workstations. We employed a Robust Linear Regression Model, selected based on the Root Mean Square Error (RMSE) of predicted values. Our hourly predictions exhibited a low RMSE of 0.025712 [kWh], signifying the model's effectiveness in short-term predictions. However, the daily predictions revealed a higher RMSE of 3.25029 [kWh], indicating underfitting within this time window.
- As a part of our future endeavors, we plan to conduct additional experiments with extended data acquisition periods, allowing us to train the prediction model for more extended time windows, such as weeks and months. Furthermore, we aim to design an FPGA-based gateway to aggregate data from multiple energy consumption sensors connected to data center servers. This approach will enhance the accuracy of our dataset and validation of the energy consumption prediction model, contributing to more reliable and efficient energy management solutions.
✅ References:
- Estrada, R., Torres, D., Bazurto, A., & Valeriano, I. (2022). Learning-based Energy Consumption Prediction. Procedia Computer Science, 203, 272-279.
- ✅The Matlab code used for data processing are available in: https://github.com/vasanza/Matlab_Code/tree/EnergyConsumptionPredictionDatacenter
- ✅The dataset used for data processing are available in: https://ieee-dataport.org/open-access/data-server-energy-consumption-dataset
- Read More
Comments
Post a Comment