Power Consumption Prediction: Analyzing Trends for Sustainable Solutions

Power Consumption Prediction" explores data analysis techniques to forecast energy usage patterns. This presentation focuses on methodologies like time series analysis, machine learning models, and data visualization to predict future power demands accurately. Tailored for a data analysis course or project presentation, this topic illustrates how insights gained can inform sustainable energy planning, optimize resource allocation, and enhance operational efficiency in diverse sectors.https://bostoninstituteofanalytics.org/data-science-and-artificial-intelligence/

1. Power Consumption Prediction

2. Title Slide Title : Predicting Power Consumption in Zone 3 Subtitle : A Machine Learning Approach Presenter : B.Aaswin Robert Date : 20-07-2024

3. Agenda Objective : Develop a machine learning model to accurately predict power consumption in Zone 3. Goals : Improved forecasting for optimized resource allocation. Cost savings by anticipating peak demand periods. Promoting energy efficiency and integrating renewable energy sources.

4. Dataset Description : Overview of the power consumption dataset . Columns : Datetime Temperature Humidity WindSpeed GeneralDiffuseFlows DiffuseFlows PowerConsumption_Zone1 PowerConsumption_Zone2 PowerConsumption_Zone3. Size : 52416 records .

5. Exploratory Data Analysis (EDA) Visualizations : Pairplot to visualize relationships between variables . Correlation matrix to identify significant relationships. Insights : Key features correlated with power consumption . Interaction effects between temperature and humidity.

6. Sample EDA Output

7. Data Preprocessing Steps : Handle infinite values and convert to NaN . Exclude non-numeric columns . Fill missing values with mean. Feature Engineering : Created new feature : temperature_humidity_interaction

8. Correlation Matrix

9. Feature Selection Selected Features : Temperature Humidity Temp-Humidity Interaction Target Variable : PowerConsumption_Zone3

10. Data Splitting and Scaling Train-Test Split : 80% Training data 20% Testing data Standardization : Used StandardScaler to scale features.

11. Model Comparison and Evaluation Models Compared : Linear Regression Random Forest Gradient Boosting Evaluation Metrics : Mean Squared Error (MSE ) R^2 Score

12. Model Evaluation Results Linear Regression MSE: 32942393.15 R^2 Score: 0.2445 Random Forest MSE: 28895145.97 R^2 Score: 0.3373 Gradient Boosting MSE: 27435253.91 R^2 Score: 0.3708 Best Model Selection Chosen Model: Gradient Boosting Regressor

13. Actual vs Predicted

14. Predictions Sample Data : Temperature: 25 Humidity: 60 Temp-Humidity Interaction: 1500 Predicted Power Consumption : 21128.08 kWh

15. Conclusion Summary : Successfully developed a model to predict power consumption in Zone 3 . Optimized model ( Gradient Boosting Regressor . ) showed high accuracy and reliability. Future Work : Incorporate additional features for improved accuracy . Implement real-time prediction system.

16. Questions ?

17. Thank You!