Sense, Model and Identify the Load Signatures of HVAC Systems in Metro Stations

Technical Summary: Sense, Model and Identify the Load Signatures of HVAC Systems in Metro Stations

Problem Statement
Subway systems are major energy consumers, with Heating, Ventilation, and Air Conditioning (HVAC) subsystems accounting for over 40% of total energy usage in summer seasons. Current HVAC control strategies in established subway lines often rely on simple time-driven rules or passive responses due to the difficulty in estimating dynamic heating and cooling loads. These loads are highly variable, driven by fluctuating outdoor temperatures, passenger flows, piston winds from trains, and complex underground structures. Existing building load estimation models, which rely on construction materials and empirical heat transfer equations, fail to capture the unique dynamics of subway stations, particularly the impact of passenger flows and tunnel aerodynamics. This mismatch between load and supply leads to significant energy waste. The core problem addressed is the lack of an effective method to quantify and identify the "load signatures"—the specific quantities of heat imported from the outdoor environment and by passengers—in real-time to enable optimal control policies.

Methodology
The authors propose a novel sensing and learning approach to identify load signatures using real-world data from the HaiDianHuangZuang subway station (Beijing Line 4). The methodology involves three main components:

1. Data Acquisition and Sensing:

   • Environmental Sensors: Deployed to monitor indoor and outdoor temperatures ($T(t)$, $T_o(t)$), humidity, and CO2 levels.
   • Passenger Flow: Recorded via the ticket checking system ($n(t)$), calculated as the sum of checked-in and checked-out passengers.
   • HVAC Monitoring: Smart meters and embedded sensors track the energy consumption and working states of refrigerators, ventilators, cooling towers, and pumps. Temperatures of return air, cooling air, new air, and mixed air are measured, along with cooling water flow rates and temperature differentials.

2. Mathematical Modeling:

   • Load Model ($L(t)$): Defined as the total heat imported into the station per unit time. It is decomposed into two components:
     • Passenger Introduced Load (PIL): Heat generated by passengers, modeled as $c_p n(t) (T_p - T(t))$.
     • Environment Introduced Load (EIL): Heat from conduction through walls/roofs and air infiltration (piston wind), modeled as $(c M_{air} + \frac{1}{R_{eq}}) (T_o(t) - T(t))$.
   • Supply Model ($S(t)$): Defined as the cooling capacity provided by the HVAC system. The model adapts to three operating modes: New Air mode (outdoor air cooling), Refrigerator mode (mechanical cooling), and Mixed mode. The cooling supply is inferred from sensor readings (water temperature differentials, flow volumes) and fan power consumption (using fan affinity laws to estimate air volume).
   • Thermal Dynamics: The variation in indoor temperature is modeled as the difference between load and supply: $L(t) - S(t) = c M_z \Delta(t)$, where $\Delta(t)$ is the temperature change over time.

3. Parameter Identification Algorithm:

   • The problem is formulated as a linear regression to identify unknown coefficients ($c_p$, $\alpha$, $\beta_{ac}$) in the load and supply equations.
   • To address sensor noise and the non-negativity constraints of physical parameters, the authors propose a search algorithm. Instead of minimizing absolute error, the algorithm minimizes the relative difference between the integrated load and integrated supply over a time period.
   • To further mitigate sensor noise (specifically in temperature difference calculations), the method utilizes integrated values of load and supply rather than instantaneous point values.

Key Contributions

• Novel Load Signature Identification: The paper presents a data-driven approach to separate and quantify the specific thermal loads contributed by outdoor environments versus passenger flows in a subway station, a factor often overlooked in standard building models.
• Integrated Sensing and Modeling Framework: The authors successfully integrated heterogeneous data sources (ticketing systems, environmental sensors, and HVAC energy logs) to construct a comprehensive thermal model.
• Robust Parameter Estimation: A specific search algorithm is proposed to solve the constrained optimization problem, handling sensor noise by minimizing relative errors of integrated values rather than instantaneous values.

Experimental Results
The model was validated using real data collected from August 21 to August 23, 2013, at the HaiDianHuangZuang station.

• Load Signatures: The derived signatures revealed that environmental loads (outdoor temperature differences) constitute the major portion of the total load, while passenger loads, though significant during rush hours, represent a smaller portion.
• Model Accuracy: The calculated cooling supply closely matched the derived load signatures. The relative error between the integrated load and integrated supply was found to be small, indicating the algorithm provides a confident estimation of the load parameters.
• Temporal Patterns: The results confirmed distinct load patterns corresponding to rush hours (morning and evening peaks driven by passenger flow) and daily temperature cycles.

Significance and Claims
The paper claims that identifying these load signatures provides critical context for designing smart, optimal control policies. By understanding the specific contributions of passengers and the environment to the thermal load, HVAC systems can be controlled to supply only the necessary cooling efforts to meet predicted demands. This approach aims to maintain indoor thermal comfort while significantly optimizing energy consumption. The authors state that this work lays the foundation for future research into optimal control strategies for subway HVAC systems, moving beyond simple time-driven rules to data-driven, predictive control.