VlietBuilding

Published: 23 August 2021| Version 1 | DOI: 10.17632/xzdy23nzvj.1
Contributors:
,
,

Description

These are historical operational data from a thermally activated test building located in Levuen, Belgium, from 2020-10-21 until 2021-04-20. The data were obtained when controlling the building with a model predictive controller that alternated different controller models during its operation. Additionally, three Python scripts are provided to visualize the experiments, the key performance indicators, and the control prediction deviations. This is a standalone package with no other dependencies than Numpy, Pandas and Matplotlib. The scripts can be run directly to generate the plots. A brief description of the data is provided below: - plant.csv : time-series data for the measurements in the test building during operation. - disturbances.csv : time-series data for the disturbances during operation. This includes weather conditions, pricing, and comfort setpoints. - historical_optimizations.csv : history of all optimization trajectories. Every time step there is a full optimization trajectory during a prediction horizon of two days from that time step. These data are used to compute the controller prediction deviations. - /pred_error : prediction errors obtained as the "a priori" values of the state estimator when using the controller models to predict the temperature with inputs from "plant.csv" and "disturbances.csv". These data are used to compute the n-step-ahead prediction errors. bb, gb, and wb indicate grey-box model, black-box model, and white-box model. The next integer indicates the experiment number. the final integer indicates the prediction horizon in seconds. E.g.: "pred_error_gb_2_86400.csv" contains the prediction errors when using the grey-box model to estimate the temperature with a prediction horizon of one day. - /kpis : Summary of obtained KPIs for each experiment and controller model.

Files

Steps to reproduce

This is a standalone package with no other dependencies than Numpy, Pandas, and Matplotlib. The scripts can be run directly to generate the plots.