⚡ electrical engineering

Small HVAC Control Demonstrations in Larger Buildings Often Overestimate Savings

This paper demonstrates that estimating energy savings from advanced HVAC controls based solely on a subset of zones in a larger building often leads to significant overestimation due to neglected inter-zone heat transfer, and proposes a more accurate alternative method using temperature measurements of both controlled and adjacent zones that eliminates the need for baseline energy estimation.

Original authors: Arash J. Khabbazi, Kevin J. Kircher

Published 2026-02-27

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Arash J. Khabbazi, Kevin J. Kircher

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "Leaky Bucket" Problem

Imagine you are trying to save money on your heating bill. You decide to turn down the thermostat in just one room of your house (let's say the living room) to save energy, while leaving the rest of the house at the normal temperature.

You check your meter and see that the living room used less heat. You celebrate, thinking, "I saved 20% on heating!"

But here is the catch: Heat doesn't stay put. It flows like water. When you make the living room colder, heat from the warm kitchen, the hallway, and the bedroom starts rushing into the living room to fill the gap. To keep the kitchen and bedroom warm, your furnace has to work harder to replace that lost heat.

The Paper's Main Point:
Researchers often test new, fancy HVAC controls by only adjusting a few rooms in a big building. They report the savings for just those rooms. The problem? They are ignoring the "leak." The energy saved in the test rooms is often just borrowed from the rest of the building. If you only look at the test rooms, you think you saved a fortune. If you look at the whole building, the savings might be tiny, or even zero.

The Core Analogy: The "Fictitious Savings"

The authors use a great example to explain this: The Interior Room.

Imagine a conference room in the middle of a huge office building. It has no windows and no walls touching the outside world. It is surrounded entirely by other offices.

The Experiment: You turn down the heat in this middle conference room to save energy.
The Result: The room gets cold. But where does the cold come from? It sucks heat out of the surrounding offices.
The Consequence: The offices next door get colder, so their heaters kick on harder to stay warm.

The Verdict: If you only measure the conference room, you see a massive drop in energy use. But if you measure the whole building, you haven't saved a single penny. The energy didn't disappear; it just moved from the neighbors to the conference room. The "savings" are fictitious (fake).

The paper proves mathematically that the more "interior" a room is (the more walls it shares with other rooms and the fewer walls it has with the outside), the more likely these savings are to be fake.

The Solution: A New Way to Measure

Usually, to prove you saved energy, you have to guess what would have happened if you didn't do anything (the "baseline"). This is hard to get right.

The authors propose a clever new method that doesn't require guessing. Instead of guessing the past, they look at the temperature differences between the rooms.

Think of it like this:
Instead of trying to calculate how much water you saved by plugging a hole in a bucket, you just measure how much water is flowing between the buckets.

If the "Test Room" gets colder, and the "Neighbor Room" gets slightly warmer (or has to work harder), the math accounts for that flow.
By measuring the temperature of the controlled room and the adjacent rooms, you can calculate the true savings for the whole building without needing to guess what the building would have done otherwise.

Why Does This Matter?

Trust: Right now, many companies and governments are skeptical about "smart" building controls because the real-world results don't match the computer simulations. This paper explains why: the tests were measuring the wrong thing.
Better Decisions: If we use this new method to measure savings, we can tell the truth. If a new control system actually saves money, we'll know. If it doesn't, we'll know that too.
Climate Impact: Heating and cooling buildings create a huge amount of pollution. If we can trust the data, we can adopt better technologies faster, saving billions of dollars and reducing emissions.

Summary in One Sentence

Don't just look at the room you changed; look at the whole house, because heat flows like water, and if you ignore the neighbors, your "savings" are just an illusion.

1. Problem Statement

Advanced HVAC control strategies, such as Model Predictive Control (MPC) and Reinforcement Learning (RL), show significant promise in simulations for reducing energy costs, emissions, and peak demand. However, their real-world adoption is hindered by skepticism regarding whether they can deliver persistent savings comparable to simulation results.

A critical issue identified in field demonstrations is a systematic error in Measurement and Verification (M&V). Researchers often control only a subset of thermal zones (e.g., a single conference room) within a larger multi-zone building and report energy savings based solely on those controlled zones. This approach neglects cross-zone heat transfer.

The Mechanism of Error: If an advanced control system lowers the temperature in a controlled zone to save energy, heat naturally flows from adjacent, uncontrolled zones into the controlled zone. To maintain their setpoints, the HVAC systems in the adjacent zones must work harder, consuming additional energy.
The Consequence: By ignoring the increased energy load in adjacent zones, researchers systematically overestimate the net building-wide savings. In extreme cases (e.g., controlling a purely interior zone with no direct contact to the outdoors), all perceived savings are fictitious.

2. Methodology

The authors develop a mathematical framework to characterize this error and propose a corrected estimation method.

A. Mathematical Model

Thermal Network: The building is modeled as an arbitrary network of $n$ thermal zones using thermal circuit analogies (where heat is charge, temperature is voltage, and thermal resistance/capacitance represent physical properties).
Dynamics: The temperature dynamics for each zone $i$ are governed by a linear differential equation:
$C_i \frac{dT_i(t)}{dt} = \sum_{j=0}^{n} \alpha_{ij}(T_j(t) - T_i(t)) + q_i(t) + w_i(t)$
Where $C_i$ is thermal capacitance, $\alpha_{ij}$ is thermal conductance between zones, $q_i$ is controlled heating/cooling power, and $w_i$ is exogenous heat gain.
Scenarios: Two scenarios are compared:
1. Baseline: A benchmark controller operates all $n$ zones.
2. Experiment: An advanced controller operates only $m$ zones ( $1 \dots m$ ), while the remaining $n-m$ zones remain under baseline control.

B. Theoretical Derivation (Theorem 1)

The paper derives Theorem 1, which mathematically characterizes the overestimation error.

The Error Term: The paper proves that reporting savings only for the controlled zones overestimates the true whole-building savings by:
$\text{Error} = \sum_{i=1}^{m} \sum_{j=m+1}^{n} \alpha_{ij} \int_{0}^{\tau} a_j(t) (T_i(t) - \tilde{T}_i(t)) dt$
Where $a_j(t)$ is the thermal price (cost/emissions per unit energy) and $(T_i - \tilde{T}_i)$ is the temperature deviation.
Corrected Estimation: The authors propose an exact formula for whole-building savings that relies on temperature measurements of the controlled zones and estimates of thermal conductances ( $\alpha_{ij}$ ). Crucially, this method does not require estimating a baseline energy consumption profile, which is a major source of uncertainty in traditional M&V.

3. Key Contributions

Mathematical Characterization of Bias: The paper provides the first rigorous mathematical proof quantifying how neglecting cross-zone heat transfer leads to savings overestimation, showing that the error scales with the ratio of inter-zone conductance to exterior conductance.
New M&V Methodology: It introduces a novel estimation method that uses readily available zone air temperature data and thermal conductance estimates. This eliminates the need for complex baseline energy modeling.
Identification of "Fictitious Savings": The authors demonstrate that for purely interior zones (no direct contact with the outdoors), the relative overestimation error approaches infinity, meaning any reported savings are entirely illusory.
Numerical Validation: A detailed simulation validates the theory, showing that a naive zone-level analysis can overestimate savings by nearly 200% compared to the true whole-building savings.

4. Results

The authors validated their theory using a numerical example of a two-zone building in West Lafayette, Indiana, during a cold snap.

Setup: Zone 1 (25% of area) was controlled by an advanced MPC system; Zone 2 (75% of area) remained under baseline control.
Naive Analysis: Looking only at Zone 1, the advanced control appeared to save $1.83 (an 18% reduction in cost for that zone).
Cross-Zone Effect: The temperature drop in Zone 1 caused heat to flow from Zone 2. Consequently, Zone 2's heating costs increased by $1.21 (a 4% increase).
True Outcome:
- Baseline Whole-Building Cost: $40.66
- Experiment Whole-Building Cost: $40.04
- True Savings: $0.62.
Conclusion: The naive zone-level analysis overestimated the true savings by 195% ($1.83 reported vs. $0.62 actual). This aligns closely with the theoretical prediction ( $\alpha_{12}/\alpha_{10} = 200\%$ ).

5. Significance and Implications

Restoring Trust: By providing a rigorous method to correct for cross-zone interactions, this paper addresses a major barrier to the adoption of advanced HVAC controls. It allows researchers and practitioners to generate trustworthy field demonstration data.
Policy and Adoption: Accurate M&V is essential for convincing government and business decision-makers to invest in advanced control systems. If savings are consistently overestimated, the economic case for these technologies is undermined when real-world results fail to match projections.
Practical Application: The proposed method is practical because it relies on standard temperature sensors and thermal conductance estimates (which can be derived from building blueprints or simple tests), rather than complex energy modeling.
Limitations & Future Work: The current model assumes linear heat transfer and linear cost objectives. Future work is needed to address non-linear objectives (e.g., peak demand shaving) and to develop strategies for handling uncertainty in thermal conductance estimates.

In summary, the paper argues that isolated zone control demonstrations are fundamentally flawed unless they account for thermal coupling with the rest of the building. The proposed correction method offers a path toward accurate, verifiable, and trustworthy validation of next-generation HVAC technologies.