Synconn_build: A python based synthetic dataset generator for testing and validating control-oriented neural networks...

Area of Science:

• Building energy systems
• Computational modeling
• Data science for built environments

Background:

• Model-based predictive control (MPC) in buildings requires robust control-oriented models.
• A significant gap exists in high-quality, varied datasets for benchmarking these models.
• Existing synthetic data generation methods may lack flexibility or produce excessive data.

Purpose of the Study:

• To introduce synconn_build, a novel open-source Python tool for generating synthetic building operation data.
• To address the limitations of current datasets for training and evaluating building control models.
• To provide a flexible and user-friendly platform for creating diverse building simulation datasets.

Main Methods:

• Utilizes EnergyPlus as the core building energy simulation engine.
• Automates simulation setup, control input signal generation, and co-simulation for occupancy.
• Employs a text-based configuration file for user-guided simulation parameterization.
• Acquires external weather data to enhance simulation realism.

Main Results:

• Synconn_build automates complex EnergyPlus configuration and simulation workflows.
• The tool generates unique random control signals and occupancy schedules for diverse operational scenarios.
• It selectively produces data based on user inputs, preventing data overproduction.
• Generated data exhibits more frequent variations than typical operational schedules, covering broader conditions.

Conclusions:

• Synconn_build effectively addresses the need for synthetic datasets in building control research.
• The tool's automation and user-centric design simplify the creation of valuable benchmarking data.
• It offers a flexible and efficient solution for advancing control-oriented building dynamics prediction models.