DPU-ALDOSKI V2
Description
The fundamental hypothesis of this research is that educational buildings, particularly departments with high laboratory density, exhibit a "large idle baseline," where laboratory equipment and systems remain powered to stay in standby mode even during periods of inactivity. Traditional Energy Management Systems (EMS), based on fixed timers or simple rule-based control, lack the proactive thinking necessary to address irregular academic schedules and fluctuating occupancy. Therefore, this study proposes a binary intelligent scheduler based on the Chimpanzee Optimization Algorithm (ChOA) to optimize on/off schedules. The main objective is to demonstrate that AI-driven optimization can reduce avoidable energy waste without compromising established laboratory processes or equipment safety. The DPU-ALDOSKI dataset provides sub-minute telemetry measurements (at a rate of 30 seconds) capturing voltage, current, active power, and frequency from the Duhok Polytechnic University (DPU) laboratories. Raw fluxes include electrical logs from two institutes: Duhok and Sheikhan. The results demonstrate that the scheduler based on the ChOA algorithm achieved a significant positive reduction in energy consumption across all time horizons of the test. Specifically: Device 1 (Sheikhan): showed a weekly decrease of 17.21%, bringing the total decrease to 24.01% over an 11-month monitoring period. Device 3 (Duhok): Achieved a weekly decrease of 13.16% and a similar long-term decrease of 24.01%. Operational efficiency: The system successfully reduced "working hours" (power supply time) by 26% for device 1 and by 33% for device 3 by converting idle periods into planned downtime. Statistical significance: Verification using t-tests and the Wilcoxon signed-rank test confirmed that these savings are highly statistically significant, with p-values significantly lower than 10-4. The data highlights that the savings are time-driven, not load-driven, meaning they result from intelligent scheduling rather than reduced equipment performance or impact on power quality. The 3-sigma control scheme verified that voltage and frequency remained stable throughout the study period, ensuring that the energy savings were not due to external fluctuations. This derived dataset (DPU-ALDOSKI-After) records optimized hourly on/off paths aligned with the original timestamps. It serves as a reusable evaluation framework for building control research, enabling other researchers to benchmark algorithms such as PSO or GA, or test short-term load prediction (STLF) models under the same policy (8:00–21:00) and institutional constraints.
Files
Steps to reproduce
1. Data Sources and Infrastructure: This research relied on field measurements obtained from the monitoring project at Duhok Polytechnic University (DPU). Data was collected from two main laboratory nodes: Device 1 (Sheikhan Institute) and Device 3 (Duhok Institute). The data collection infrastructure is based on a distributed Internet of Things (IoT) architecture, utilizing nodes based on ESP32 microcontrollers that communicate with back-end systems via MQTT and HTTP protocols. These nodes provide high-resolution telemetry data with a high sampling rate (approximately every 30 seconds). 2. Electrical Instruments and Measurements: Energy analyzers and smart meters were used at representative points within the educational laboratories. The dataset collected and available in the DPU-ALDOSKI archive includes the following measurements. o Voltage (volts) o Current (amps) o Active Power (kW) o Frequency (Hz) o Power Factor and Timestamps 3. Data Preprocessing and Cleaning Protocols: To ensure the accuracy and comparability of the results, the raw data underwent a series of rigorous processing steps: o Temporal Alignment: The time zone of all timestamps was standardized and precisely reordered, with duplicate records necessarily merged. o Cleaning and Quality Assurance: Physically illogical values were excluded, and robust statistical techniques (such as absolute deviation from the median - MAD) were used to identify anomalies caused by measurement noise o Hourly Aggregation: The data was converted from 30-second resolution to hourly frames to align with the daily planning horizon (24 steps), where average power, maximum current, and number of activity times per hour are stored. . 4. Policy Masking and Constraints: The institutional rules of Duhok Polytechnic University were encoded in a binary "policy mask" that defines the authorized working hours:
Institutions
- Duhok Polytechnic UniversityDuhok