Energy Usage Dataset

Predictive model and dataset for energy management optimisation

Image 1: Snapshot from the SunSync Dataset

Asset Description

This asset provides a predictive model and dataset aimed at optimising energy management processes. Built using a decision tree classifier and trained on historical energy usage data, the model forecasts upcoming power needs and identifies optimal decision states. This implementation helps organizations streamline their energy usage, improve resource planning, and enhance the efficiency of their energy systems, whether through preventive maintenance or better alignment of generation with consumption.

Dataset and Model Details

Dataset Overview

The dataset comprises 3,466 entries each represented by 17 features that capture a variety of system metrics and states related to energy usage. These features include measurements of power (input and output), wattage predictions, battery status, and operational flags. The data format is primarily numerical with timestamps, facilitating time series analysis and predictive modelling.

Key attributes include:

Feature	Description
Id	A unique identifier for a specific device or data entry in the system.
Timestamp	A record of the date and time when an event, measurement, or data entry took place.
rootNodeIsRun_value	A status flag indicating whether the recycling module (energy consumer) is currently running.
Upcoming Watt Hours	A forecast of the amount of energy (in watt-hours) expected to be generated in the upcoming hour.
Daily Wattage (Watt Hours)	The total amount of energy predicted to be generated in a day, measured in watt-hours.
Remaining Watt Hours	An indication of the amount of energy (in watt-hours) left to produce for the remainder of the day.
DC Input Power	The power (in watts) entering the system as Direct Current (DC), typically from solar panels or other DC sources.
AC Input Power	The power (in watts) entering the system as Alternating Current (AC), potentially from the solar panel or other AC sources.
AC Output Power	The power (in watts) being supplied by the system in AC form, to power devices.
DC Output Power	The power (in watts) being supplied by the system in DC form, possibly for charging batteries or powering DC devices.
Power Generation	The total power (in watts) being generated by solar panels at the moment.
Total Battery Percent	The current charge level of the battery system, expressed as a percentage of its total capacity.
AC Output On	A status flag indicating whether the AC output of the system is currently active or turned on.
DC Output On	A similar status flag indicating whether the DC output of the system is active or turned on.
AC Input Voltage	The voltage level (in volts) of the AC power entering the system.
Internal DC Input Voltage	The voltage level (in volts) of the internal DC power within the system.
decision	A binary target variable representing a decision state.

Model Details

The predictive model is a Decision Tree Classifier built using scikit-learn, a popular machine learning library in Python. The model was trained to predict outcomes based on the dataset's features, focusing on optimizing decisions related to energy management.

Usage

Prerequisites

Before using this asset, ensure you have:

• Python installed on your system (recommended version 3.9 or newer)
• Required Python libraries such as scikit-learn and pandas installed.
  • You can install them with pip: pip install scikit-learn pandas

Loading the Asset

• Download/Access Dataset: Ensure you have access to the dataset file (Output_data.csv).
• Load the Data: Use pandas to load the CSV file into a DataFrame:
  • import pandas as pd
  • data = pd.read_csv('Output_data.csv')

Model Usage

• Loading the Model: Ensure that the model file (model.joblib) is in the same directory or provide the correct path. Load the model using joblib:
  • import joblib
  • model = joblib.load('model.joblib')
• Prepare the Data: Make sure the dataset columns match the model's expected input format. This might involve feature engineering or scaling

input_features = data[[

       'rootNodeIsRun_value', 
       'upcoming_watt_hours', 
       'dailyWattage_watt_hours', 
       'remaining_watt_hours', 
       'dc_input_power', 
       'ac_input_power', 
       'ac_output_power', 
       'dc_output_power', 
       'power_generation', 
       'total_battery_percent', 
       'ac_output_on', 
       'dc_output_on', 
       'ac_input_voltage', 
       'internal_dc_input_voltage' ] ] 

• Make Predictions: Use the loaded model to make predictions on new data
  • predictions = model.predict(input_features)

Licence

Open source

Resources

• Input data, output data and data model available in Github
• Created by PBN / am-LAB Digital Innovation Hub
• Contact gyula.gal@pbn.hu / martin.dan@pbn.hu

Acknowledgement

The dataset was created in the framework of the AI REDGIO 5.0 project as DF XIII PBN am-LAB (HUNGARY): SUNSYNC: AI SOLUTION FOR OPTIMIZING RECYCLING IN INDUSTRY AT THE LEVEL OF AM-LAB’S DF.