Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems: A Case Study of the ELB Building

arXiv:2610.20777 · eess.SY, cs.SY · Submitted 2026-10-14 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: I'm Rosa.

Dev: And I'm Dev, here with Taro.

Rosa: Today's paper: "Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems".

Dev: The gist The proposed framework presents an optimal energy management model that integrates demand response and battery energy storage systems to minimize electricity costs in campus buildings by scheduling building…

Rosa: First, who's behind it and why it matters.

Title and authors: Rosa: So we're looking at this paper today, "Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems: A Case Study of the ELB Building." It’s pretty straightforward: they’ve built an optimization model that ties together demand response and battery storage to lower electricity bills on a campus.

Dev: Exactly. Think of it like this, we're scheduling exactly when to turn off or adjust loads in a building, and when to charge or discharge the battery, all based on how much electricity costs at different times of the month.

Taro: What I find interesting is that they aren't just looking at total energy use; they are focused on minimizing operational cost by scheduling things based on time-varying prices and demand profiles.

Rosa: Right, so the core idea is scheduling building loads and battery operations according to those fluctuating electricity prices and demand patterns. It’s a mixed-integer linear program that figures out the best hourly schedule for a whole billing period.

Dev: And they model it in two layers: an optimization layer where they decide what to do, and then the physical layer representing the actual building energy system, including the existing grid connection and rooftop solar generation.

Taro: That means you have inputs like measured loads at fifteen-minute intervals and hourly PV generation feeding into this program, which dictates the hourly operating schedule over a whole month.

Rosa: And they’re using a specific tariff structure—the DTE six point two Tariff—which includes both energy rates and those demand charges applied to the monthly peak.

Dev: The problem formulation is set up as a mixed-integer linear program, or MILP, over a monthly billing cycle defined by intervals T = zero one N-one. The objective is strictly to minimize the total operational cost of the building.

Taro: That cost includes the energy charge for power imported from the grid and those degradation costs associated with charging and discharging the battery. Plus they handle demand charges applied to that monthly peak and a fixed monthly service charge.

Rosa: And demand response is handled through two things: load shedding, where you just reduce demand without recovering it, and load shifting, which moves that demand to a different time slot during lower peak hours.

Dev: Load shifting is modeled by using two variables for each hour: one for the amount of demand removed and one for the amount added to that hour. They enforce a constraint that whatever you shift out over the whole cycle must equal what you shift in, which is a key part of how they keep the schedule balanced.

Taro: Speaking of constraints, they model the battery energy storage system using equations eleven and twelve to cap charging and discharging power, making sure it only operates when it's allowed.

Title and authors: Rosa: They also have equation thirteen which prevents charging and discharging the battery at the same time, which is a physical necessity in any real system. Then there are equations fourteen and fifteen that track how stored energy carries over from one hour to the next.

Dev: And equation sixteen requires that the battery ends exactly where it started at the end of the billing cycle so that the schedule is self-contained, while equation seventeen ensures you don't exceed any physical limits on battery storage capacity.

Taro: So, when we look at their simulation results for May two thousand twenty-six demand response alone cuts the monthly electricity bill by about twelve percent compared to just running everything conventionally <ref:2610.20777#pg1>.

Rosa: But the real payoff is when they coordinate that demand response with battery energy storage; in that scenario, the total electricity cost reduction jumps to approximately twenty-six percent. That shows a significant economic gain from coordinating those resources.

Dev: This highlights how integrating battery storage with demand response really boosts the operational flexibility of campus buildings and gives them a way to manage costs more effectively and sustainably.

Taro: The main contributions they point to are formulating this coordinated DR–BESS energy management framework as a MILP problem for an institutional building, jointly scheduling flexible loads and BESS operation under practical operating constraints.

Rosa: They also made sure to incorporate the actual commercial electricity tariff into the cost model, including those energy and demand-related billing components, so that peak-demand reduction is directly reflected in the optimization objective.

Dev: They used measured electricity-demand and rooftop PV data from a real campus building for their evaluation, rather than relying only on synthetic or benchmark profiles. This makes the results much more applicable to actual campus operations.

Taro: And they compared several scenarios: baseline, DR-only, BESS-only, and coordinated DR–BESS cases under the same data and tariff to show exactly what each flexibility mechanism contributes economically.

Rosa: So, looking at their conclusion for this paper on "Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems: A Case Study of the ELB Building," they emphasize that peak power management, not just total energy consumption, is the main driver of cost for institutional buildings billed with a demand charge.

Dev: They found that demand response alone reduced the monthly bill by twelve percent, but when you coordinate it with battery storage, they got a twenty-five point eight percent reduction. Plus, they managed to lower the billed peak from one hundred thirty-four point five kilowatt to seventy-six point four kilowatt.

Taro: It's important to note that demand response alone was better than the battery alone in terms of cost savings, twelve percent against nine point seven percent, because shed load is removed entirely. They weren't redundant; they were complementary resources working together.

Rosa: Based on these results, it suggests that demand response, which doesn't require any upfront investment, is the logical first step for institutional buildings to take. Battery storage is then best deployed as a complement to what’s already there.

Title and authors: Dev: That makes sense from an engineering standpoint; you get immediate flexibility with DR without needing the capital outlay for storage right away. But the battery is clearly needed to tackle that peak demand charge effectively.

Taro: I want to add something about capacity scaling, and they looked at how increasing battery capacity affects things. They found that while it reduces the peak and total cost, it isn't as effective as demand response on its own, supporting the idea that DR should come first.

Rosa: And they also examined seasonal variations by looking at a December two thousand twenty-five case against the May two thousand twenty-six case <ref:2610.20777#pg1>. In that December scenario, they found the savings were driven almost entirely by that demand charge.

Dev: So it shows how load profiles matter; a flatter profile in December leaves less peak power available for the battery to shave once demand response has already been applied. The performance changes based on the load characteristics of that specific month.

Taro: And they mentioned a limitation: their framework is formulated as an MILP, which means it's mathematically rigorous but might not capture every single unpredictable physical nuance in real-time perfectly.

Rosa: That's fair; they admit the model has these practical operating constraints, like the limits on HVAC and lighting systems that can be shifted or shed without equipment damage.

Dev: And another point is that they used a specific tariff structure for their cost model, which means if you switch to a completely different utility rate structure, you'd need to re-run the whole optimization.

Taro: So what this paper means for someone just listening is that institutional buildings have a clear path: start with demand response because it’s free flexibility, and then add battery storage on top for that extra savings when demand charges are high.

Rosa: That's the main point of this work on "Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems: A Case Study of the ELB Building." It gives a concrete roadmap for how to use flexible loads and storage together to manage energy costs in large buildings.

Dev: It shows that coordinating these resources isn't just theoretical; it’s mathematically solvable and yields measurable savings when you use real data like that from the ELB Building.

Taro: I think the real implication is that demand response, as a cost-saving measure, should be prioritized for buildings before we invest in adding new battery infrastructure to maximize return on investment.

Rosa: It’s a practical approach to making campus energy management more cost-effective and sustainable by focusing on peak demand reduction through smart scheduling.

Dev: That’s where we’ll leave it for this paper. Next up, we look at how other systems handle real-time control loops in these complex environments.

The paper's summary: Rosa: So, to wrap up what we just talked about, this paper takes everything we discussed—the scheduling of loads and battery storage based on real energy prices—and boils it down to a practical solution for a campus building like the ELB.

Dev: Basically, they’ve built this mixed-integer linear program that acts as the brain for the whole system over an entire month, figuring out exactly when to use demand response and when to cycle that battery.

Taro: What’s really interesting is how they handled those constraints—like making sure you don't try to charge and discharge the battery at the exact same time—it keeps it realistic for a real building.

Rosa: Right, so it’s not just a theoretical idea; they put the actual commercial electricity tariff in there, including that demand charge applied to the peak, which means the optimization is directly minimizing what you actually pay on your bill.

Dev: And they used data from a real campus building instead of just making up some numbers for testing, so their results are grounded in actual load profiles and PV generation patterns.

Taro: What this means for us right now is that we can see how much savings you get just from shifting demand versus what you get when you actually have the battery storage working alongside it.

Rosa: And the big picture here is that this coordinated approach—demand response plus battery—cuts your costs significantly more than either one alone, which points to a very clear strategy for institutional buildings.

Dev: They found that the demand charge component is actually what drives most of the savings in those scenarios, showing that managing the peak power demand is way more impactful than just trying to save energy overall.

Taro: It also showed how increasing battery size helps shave off that peak, but they confirmed that demand response alone can be better for cutting costs because you’re not paying for the storage degradation every time you shift a load.

Rosa: So, if you're looking at an existing building, this framework gives you a roadmap: start with demand response since it costs nothing extra to implement, and then look at battery storage as the next step to maximize those peak-demand savings.

Dev: That’s the practical application—it moves from just knowing *how* to optimize to having a model that can actually be fed real-world data for deployment.

Taro: It also showed how different seasons affect the results; for instance, in a December month with a flatter load profile, the savings are almost entirely dictated by that demand charge component.

Rosa: That’s it for this deep dive into their findings on the ELB Building study. Now, if you want to know how these kinds of scheduling models actually handle real-time control loops when things go wrong outside of a controlled test environment, we'll be looking at some other papers next.

The paper's improvements: Rosa: So, we're talking about what the authors suggest they could do next to improve this energy management framework for campus buildings.

Dev: They focus on making it more robust when the real world throws something unexpected at it, like unpredictable load changes or sudden grid fluctuations.

Taro: What’s interesting is that they look at scenarios where demand response and battery storage aren't working perfectly together, but how the system handles that failure.

Rosa: They suggest incorporating a more sophisticated way to handle those constraints on the battery—not just simple caps, but maybe modeling it with more complex state-of-charge management during rapid transitions.

Dev: And from my side, I think they should explore how the optimization layer can be updated in real time instead of running a full monthly schedule every time the prices change drastically.

Taro: I agree with that; if you're dealing with a fast-changing environment, a static monthly plan isn't going to cut it for graceful recovery.

Rosa: They also pointed out that they should look into how demand response can be made more intelligent, maybe by integrating predictions about future building occupancy or weather directly into the load shifting variables.

Dev: That makes sense; if you know a big meeting is starting in an hour, you can pre-emptively shift loads before the peak hits, which reduces latency in your control loop.

Taro: And they mentioned that their current model might not fully capture all the physical limitations of the HVAC systems—like how quickly those systems can actually react to a sudden temperature change.

Rosa: So, future work involves building a simulation that incorporates more realistic physics for those physical assets so the optimization doesn't prescribe something impossible for a building to do.

Dev: And they should also look at making the entire framework more adaptive, perhaps using machine learning components to learn from past performance and adjust the MILP parameters on the fly.

Taro: That’s where it gets interesting—moving from a fixed schedule to an AI-driven system that learns how to be graceful when things misbehave.

Rosa: So they are pushing for a system that doesn't just follow a plan, but one that can adapt its strategy based on what the building and the grid are actually doing right now.

Dev: It moves the safety case from just checking if the initial schedule is feasible to ensuring continuous, safe operation under dynamic conditions.

Taro: That’s it for what they suggest for future work—moving toward a truly adaptive, learning system that handles real-time uncertainty better.

Rosa: If you want to see how these kinds of real-time control loops actually handle failures in complex systems, we'll be looking at some other papers next.

Conclusion: Rosa: So we’re wrapping up on this paper, "Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems: A Case Study of the ELB Building." Basically, they showed how to use a mixed-integer linear program to perfectly schedule demand response and battery storage based on actual electricity prices for a whole month.

Dev: Right, it boils down to minimizing the total operational cost by balancing energy charges against those demand charges applied during peak hours.

Taro: What this really changes for autonomous systems is seeing how coordinated scheduling works in a complex, real-world physical environment like a campus building.

Rosa: They proved that combining demand response and battery storage cuts costs by about twenty-six percent compared to just using one or the other alone, which highlights the value of having both resources available.

Dev: The implication is pretty clear: for institutional buildings, demand response is a no-cost way to start saving money immediately, and battery storage becomes a necessary complement to tackle that peak demand charge.

Taro: It’s also important because they used real data from the ELB Building, so we know the model works on actual load profiles and PV generation, not just some synthetic stuff.

Rosa: That’s right; it shows that you need real-world data to make these models useful for actual deployment outside of a lab setting.

Dev: For controls engineers, it's a good example of how you can use optimization to find the best sequence for time-varying inputs, but you always have that safety case to check the loop rate and failure modes.

Taro: I just want to push on what happens when things go wrong; they looked at how the system recovers when demand response isn't perfectly executed during a peak event.

Rosa: They did look at those constraints, and it showed that while the model is mathematically rigorous, there are still practical physical limits on how fast those building systems can actually react.

Dev: So, the caveat is that this MILP framework gives you an optimal schedule for a known set of inputs, but adapting it in real time to completely unknown events is a different kind of challenge.

Taro: That’s the next step—moving from a fixed plan to something that learns and adapts dynamically when the world misbehaves.

Rosa: So, this paper gives us a solid foundation on how to use demand response and storage together for cost savings in buildings, even if it’s not perfectly real-time yet.

Dev: It’s a practical roadmap for energy management that shows the synergy between flexibility and storage.

Taro: I think the next big thing will be integrating those learning components they mentioned to make it truly resilient against unpredictable events.

Rosa: And that’s our time on this paper, "Optimal Demand Response-Based Energy Management for Campus Buildings with Battery Energy Storage Systems: A Case Study of the ELB Building."

Dev: It's a solid piece of work showing how optimization handles real-world energy economics.

Taro: It sets a good baseline for how we can model these complex scheduling interactions.

Ali Beidoun, Quang Manh Hoang, Guilherme Vieira Hollweg, Van-Hai Bui

University of Michigan-Dearborn

eess.SY, cs.SY

Submitted: 2026-10-14

Updated: 2026-10-14

Comments: 1-12

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 79/100

The gist: The gist The proposed framework presents an optimal energy management model that integrates demand response and battery energy storage systems to minimize electricity costs in campus buildings by

Key concepts

Demand Response (DR): DR involves changing how a building uses energy by either reducing its load or shifting when it uses power. This paper models this as load shedding (reducing usage) and load shifting (moving usage to off-peak times) to lower electricity bills.
Battery Energy Storage System (BESS): A BESS is a battery installed in the building that stores excess electricity and releases it when needed. The model manages when the battery charges or discharges, constrained by physical limits and operational rules to minimize costs.
Mixed-Integer Linear Program (MILP): This is a mathematical optimization technique used to find the best possible solution for a complex problem. In this paper, it's used to determine the optimal hourly schedule for building loads and battery operations over a monthly billing cycle.
Demand Charge: A demand charge is a fee billed by utilities based on the highest power demand reached during a billing period, not just total energy consumed. The optimization focuses heavily on reducing this peak demand to achieve the largest cost savings.

Summary

The gist The proposed framework presents an optimal energy management model that integrates demand response and battery energy storage systems to minimize electricity costs in campus buildings by scheduling building loads and battery operations based on time-varying electricity prices and demand profiles.

System Configuration and Operation Strategy

The framework is split into two layers: the optimization layer and the physical layer Figure 1 illustrates the system configuration and operation strategy The upper layer contains the data and optimization components: measured building loads recorded at 15-minute intervals and aggregated to hourly intervals, measured hourly rooftop PV generation, and the DTE 6.2 Tariff These inputs feed a mixed-integer linear program implemented in Python, which determines the hourly operating schedule for the billing period The lower layer represents the physical building energy system The ELB is supplied from the utility grid and from the existing rooftop solar array, both of which connect to the building bus The battery energy storage system is modeled as a proposed addition rather than existing equipment and supplies power to the bus.

Problem Formulation

The framework is formulated as a mixed-integer linear program (MILP) that determines the optimal ELB schedule over a monthly billing cycle A billing cycle is defined by a set of hourly intervals T = 0,1,…, N-1 where N is the number of hours in a billing month The objective is to minimize the total operational cost of the building In Equation (1), starting from the left, the first term is the energy charge for power imported from the grid into the system The second and third terms are the charging and discharging degradation costs of the BESS The final terms are the demand charge that is applied to the monthly peak and the fixed monthly service charge, both of which are determined by the tariff rather than by hourly consumption.

Demand Response Mechanisms

Demand response is represented by two mechanisms: load shedding, in which demand is reduced and not recovered, and load shifting, where demand is relocated to a lower peak hour The maximum shedable load each hour is the sum of each hour’s flexible share, as expressed in (3) Load shifting is modeled using two variables representing demand removed and added to each hour The total shifted out over the cycle must be equal to the amount shifted in, which is enforced by (9).

Battery Energy Storage System Modeling

The BESS is modeled using equations (11)-(12) to constrain our BESS Charging and discharging power are capped and only permitted when they are active Equation (13) prevents a situation where we are charging and discharging the BESS at the same time The stored energy carries over from hour to hour in Equations (14) and (15) Equation (16) requires the battery to end the billing cycle at the level it began so that the schedule is self-contained The stored energy must hold within physical limits of the battery, as given by (17).

Simulation Results and Discussion

Simulation results demonstrate that demand alone reduces the building’s electricity cost by approximately 12% compared with conventional operation When battery energy storage is incorporated, the total electricity cost reduction increases to approximately 26%, highlighting the significant economic benefits of coordinated demand response and energy storage scheduling The results demonstrate that integrating battery energy storage with demand response substantially enhances the operational flexibility of campus buildings and provides an effective solution for costeffective and sustainable building energy management. The main contributions of this work are summarized as follows: 1. A coordinated DR–BESS energy-management framework is formulated as a MILP problem for an institutional building, jointly scheduling flexible loads and BESS operation under practical operating constraints 2. The actual commercial electricity tariff applied to the ELB is incorporated into the cost model, including its energy and demand-related billing components, so that peak-demand reduction is directly reflected in the optimization objective 3. The framework is evaluated using measured electricity-demand and rooftop-PV data from a real campus building rather than relying exclusively on synthetic or benchmark profiles 4. Baseline, DR-only, BESS-only, and coordinated DR–BESS cases are compared under the same building data and tariff to quantify the separate and combined economic contributions of load flexibility and storage.

Conclusions

This paper presented a MILP-based energy management framework for campus buildings that schedules demand response and battery energy storage under an accurate commercial electricity tariff Through a study of the Engineering Lab Building, we concluded that peak power management, rather than total energy consumption, is the primary driver of cost for institutional buildings billed with a demand charge Demand response alone reduced the monthly bill by 12.0%, while the coordinated operation of demand response and battery storage achieved a 25.8% reduction and lowered the billed peak from 134.5 kW to 76.4 kW The savings originate almost entirely from the demand charge of the bill However, it is also important to note that demand response alone outperformed the battery alone, 12.0% against 9.7%, because shed load is removed entirely The two were not redundant; rather, they were complementary. Based on the results, this suggests that demand response, which requires no investment, represents the logical first step for institutional buildings, with storage deployed as a complement to the existing system. The paper was supported by the Summer Undergraduate Research Experience (SURE) Program at University of Michigan-Dearborn.

References

International Energy Agency. Buildings. In Energy Efficiency 2025; IEA: Paris, France, 2025
Oldewurtel, F.; Parisio, A.; Jones, C.N.; Gyalistras, D.; Gwerder, M.; Stauch, V.; Lehmann, B.; Morari, M. Use of model predictive control and weather forecasts for energy efficient building climate control. Energy Build. 2012
Killian, M.; Kozek, M. Ten questions concerning model predictive control for energy efficient buildings. Build. Environ. 2016
Drgoňa, J.; et al. All you need to know about model predictive control for buildings. Annu. Rev Control 2020
Palensky, P.; Dietrich, D. Demand side management: Demand response, intelligent energy systems, and smart loads. IEEE Trans Ind Inform 2011
Pedrasa, M.A.A.; Spooner, T.D.; MacGill, I.F. Coordinated scheduling of residential distributed energy resources to optimize smart home energy services
Wang, B.; Zhang, C.; Dong, Z.Y. Interval optimization based coordination of demand response and battery energy storage system considering SOC management in a microgrid
Sæ le, H.; Grande, O.S. Demand response from household customers: Experiences from a pilot study in Norway
Leadbetter, J.; Swan, L.

Improvements for AI systems

textbfIncreased Cost-Aware Scheduling in MILP Objective Function: Incorporate Explicit Demand Charge Modeling Directly into Optimization Objective (Equation 1). The paper states that the demand charge component can be represented directly by including it in Equation (1), ensuring the optimization objective minimizes total operational cost, which includes the demand charge that is applied to the monthly peak.
textbfQuantified Complementary Resource Value: Explicitly Model and Compare Load Flexibility vs. Storage Benefits. The results demonstrate that Demand response alone reduces the building’s electricity cost by approximately 12% compared with conventional operation while When battery energy storage is incorporated, the total electricity cost reduction increases to approximately 26%, showing the synergistic benefit of coordinated operation.
textbfContext-Specific Load Profile Adaptation: Utilize Real Building Data for Scenario Validation. The framework is validated using measured ELB electricity-demand data and measured rooftop-PV generation data from a real campus building rather than relying exclusively on synthetic or benchmark profiles, allowing the system to accurately reflect actual load characteristics like those in May versus December.
textbfDynamic Capacity Scaling Analysis: Analyze the Diminishing Marginal Benefit of Storage Capacity. The results show that increasing battery capacity produces a reduction in the peak and the total cost; however, it is not as effective as demand response, supporting the conclusion that demand response should be the primary costsaving measure, with storage deployed as a complement.
textbfSeasonal Load Profile Sensitivity: Assess Inter-Month Performance Variability. The comparison between May and December results shows that the savings in that case are driven almost entirely by the demand charge and highlights how capacity impacts savings differently based on load profile characteristics, such as the flatter December load profile leaves less peak available for the battery to shave once demand response has already been applied.

Abstract

Buildings account for a significant portion of global electricity consumption, making demand response (DR) an effective strategy for reducing energy costs and improving grid flexibility. This paper presents an optimal energy management framework for a campus building that integrates demand response and battery energy storage systems (BESS) to minimize electricity costs while satisfying operational constraints and occupant comfort requirements. The proposed optimization model schedules building loads and battery charging/discharging based on time-varying electricity prices and building demand profiles. The framework is validated using the Engineering Laboratory Building (ELB) at the University of Michigan-Dearborn as a real-world case study. Four operating scenarios are investigated: (i) baseline operation, (ii) demand response without battery energy storage, (iii) battery energy storage, and (iv) demand response with battery energy storage. Simulation results demonstrate that demand alone reduces the building's electricity cost by approximately 12% compared with conventional operation. When battery energy storage is incorporated, the total electricity cost reduction increases to approximately 26%, highlighting the significant economic benefits of coordinated demand response and energy storage scheduling. The results demonstrate that integrating battery energy storage with demand response substantially enhances the operational flexibility of campus buildings and provides an effective solution for cost-effective and sustainable building energy management.