For most commercial buildings, the most painful number on the electricity bill is not the kilowatt-hours, but the "contract demand." Taipower's basic electricity charges are based on contract demand, which is determined by the highest 15-minute average demand throughout the year -- often that moment on a summer afternoon when all HVAC systems are running at full capacity. If that peak can be shaved, contract demand can be reduced, basic charges decrease accordingly, and energy charges also drop significantly by avoiding peak time periods. This article systematically analyzes the engineering and economic logic of HVAC peak shaving, from ice thermal storage and battery energy storage to demand response and AI smart scheduling.

1. The Peak Demand Problem of HVAC Electricity and Its Economic Impact

1.1 HVAC Dominates Building Peak Load

According to the Bureau of Energy statistics, HVAC electricity consumption in Taiwan's commercial buildings accounts for 50-60% of total consumption, but during summer afternoon peak periods, this proportion surges to 60-70%[1]. Chillers, chilled water pumps, cooling tower fans, and air handling unit fans all operate at full speed simultaneously, forming steep electricity peaks. This peak typically occurs between 1:00 PM and 3:00 PM -- precisely when solar radiation is strongest and indoor cooling loads are highest.

1.2 Contract Demand and Basic Charge Cost Pressure

Taipower's electricity rate structure consists of basic charges and energy charges. Basic charges are based on "contract demand" (kW), at approximately NTD 236-256 per kW per month (for high-voltage customers)[2]. If actual peak demand exceeds contract demand, the excess is billed at double the rate. This means that even if peak demand exceeds the threshold on only a few days per year, building owners must sign a higher contract demand to avoid penalties, resulting in paying for "unused" capacity during the remaining 11 months.

1.3 Time-of-Use Rate Peak-Valley Price Differential

Taipower's time-of-use rate system divides the day into peak, semi-peak, and off-peak periods, with the price difference between peak and off-peak rates reaching 2-3 times[2]. For high-voltage summer customers, peak energy charges are approximately NTD 5.9 per kWh, while off-peak is only about NTD 2.1 per kWh. If the HVAC system can shift some electricity consumption from peak to off-peak, nearly NTD 3.8 per kWh can be saved -- this is the economic incentive for "peak shaving and valley filling."

1.4 Additional Impact of Summer Electricity Rates

Taipower's summer rates (June-September) increase energy charges by approximately 15-25% compared to non-summer months, and summer happens to coincide with the highest HVAC load period. For large commercial offices, electricity costs during the four summer months often account for over 40% of annual electricity costs, making peak shaving strategies particularly effective during summer.

2. Ice Thermal Storage: The Classic Solution for Shifting HVAC Peaks

2.1 Ice Storage Principle: Making Ice at Night, Melting During the Day

The core concept of ice thermal storage is to use off-peak low-cost electricity to make ice at night and store cooling energy, then melt the ice during daytime peak periods to supply chilled water for HVAC, avoiding or reducing chiller operation during peak hours[3]. During the phase change from water to ice, approximately 334 kJ of latent heat can be stored per kilogram. Compared to chilled water storage (which only utilizes sensible heat), ice thermal storage has 5-6 times higher energy density, significantly reducing the required storage tank volume.

2.2 Full Storage vs. Partial Storage Design Strategies

Ice storage system design has two basic strategies:

  • Full Storage: Nighttime ice production is sufficient to handle all daytime HVAC loads, with chillers completely shut down during peak periods. The advantage is that contract demand can be significantly reduced; the disadvantage is large ice storage tank volume and high initial investment. Suitable for locations with very large peak-off-peak rate differentials or ample space.
  • Partial Storage: Nighttime ice production only supplies part of daytime loads; chillers still operate during peak periods but at reduced load. Smaller storage tanks and lower investment make this the practical choice for most Taiwan projects. A common design has ice storage handling 40-60% of the peak cooling load.

2.3 Comparison of Ice Ball, Coil, and Ice Harvester Storage Types

Ice storage tanks can be classified into three main types by ice formation method[3]:

  • Encapsulated Ice (Ice Ball): Water is sealed in plastic spheres or containers and immersed in ethylene glycol solution. During ice-making, low-temperature glycol circulates to freeze the water in the spheres; during melting, warm water circulates to melt the ice. Advantages include simple structure and easy maintenance; disadvantages include lower storage density and periodic glycol replacement.
  • Ice-on-Coil (External/Internal Melt): Refrigerant or glycol circulates inside coils, with ice forming on the outer surface. External melt uses water to directly wash the ice layer off the coils, providing chilled water close to 0°C; internal melt uses warm fluid inside the tubes to melt ice. Coil-type systems have high storage density and mature technology, making them the most common choice for commercial buildings.
  • Ice Harvester: Thin ice sheets are formed on evaporator plates and then scraped into the storage tank. High ice-making efficiency and low chilled water outlet temperatures, but higher equipment complexity and maintenance costs, mostly used in large industrial or district cooling systems.

2.4 Storage Capacity Calculation and System Configuration

Ice storage capacity is calculated in refrigeration ton-hours (RTh). The calculation process is as follows:

  • Analyze the building's hourly cooling load curve (Design Day Load Profile)
  • Determine the storage strategy (full or partial storage) and the proportion of peak loads to be handled by ice storage
  • Calculate required storage capacity (RTh) = Peak period ice-handled load x Peak hours
  • Select ice storage tank type and calculate required tank volume based on storage density
  • Configure nighttime ice-making chillers -- in ice-making mode, the chiller leaving water temperature must be reduced to -5°C to -7°C, reducing COP by approximately 20-30% compared to normal HVAC mode (7°C leaving water)[4]

2.5 Investment Payback Analysis: Equipment Premium vs. Electricity Savings

The additional investment for ice thermal storage systems primarily comes from storage tanks, piping, and control systems. For a commercial office building with 1,000 RT of cooling load, a partial storage solution (ice storage handling 50% of peak load) requires an additional investment of approximately 20-35% of the total HVAC system investment. However, electricity savings come from three sources: reduced contract demand (15-25% basic charge savings), electricity shifted from peak to off-peak (20-30% energy charge savings), and higher chiller efficiency when making ice at night in cooler ambient temperatures. Combined, the payback period is typically 5-8 years.

3. Battery Energy Storage (BESS) and HVAC Integration

3.1 Behind-the-Meter Energy Storage System Architecture and Capacity Sizing

Battery Energy Storage Systems (BESS) are installed behind the meter (Behind-the-Meter), using lithium iron phosphate (LFP) batteries to store off-peak electricity and discharge during peak periods for peak shaving[5]. Behind-the-meter storage capacity sizing must consider: daily peak period shaving amount (kW) x peak hours (h) = required battery capacity (kWh), plus charge-discharge efficiency losses (approximately 10-15%) and battery depth of discharge limitations (DoD typically set at 80-90%).

3.2 Energy Storage + HVAC Collaborative Peak Shaving Strategies

Battery energy storage and HVAC system collaborative peak shaving has multiple modes:

  • Demand Management Mode: Real-time monitoring of total power demand; when approaching the contract demand limit, the energy storage system automatically discharges to supplement, preventing demand threshold exceedance. This is the most direct peak shaving application.
  • Time-of-Use Rate Arbitrage: Charging during off-peak and discharging during peak periods, leveraging the rate differential for economic benefits. The logic is consistent with ice thermal storage, but batteries offer greater flexibility -- they can simultaneously serve loads beyond HVAC.
  • Maximizing Solar Self-Consumption: Combined with rooftop solar panels, daytime generation is first stored in batteries and discharged during HVAC peak periods, increasing the self-consumption ratio of renewable energy.

3.3 Charge-Discharge Scheduling and HVAC Load Forecasting

The effectiveness of energy storage systems depends on the accuracy of charge-discharge scheduling. Scheduling must integrate the following information: next-day weather forecasts (affecting HVAC loads and solar generation), building usage schedules (meetings, events, etc.), Taipower time-of-use rate periods, and battery state of charge (SOC). Advanced Energy Management Systems (EMS) can automatically generate optimal charge-discharge schedules based on these variables and make real-time adjustments during execution.

3.4 Behind-the-Meter Storage Subsidy Program (2026-2029)

The Ministry of Economic Affairs announced the "Behind-the-Meter Energy Storage Installation Subsidy Guidelines" at the end of 2025, allocating a four-year budget of NTD 5 billion to promote enterprise behind-the-meter energy storage installation[6]. Key subsidy details include:

  • Subsidy Amount: Up to NTD 5 million per MWh of storage capacity, with a maximum of NTD 25 million per application
  • Subsidy Ratio: Not exceeding 50% of total system installation cost
  • Eligibility: Large electricity users with contract demand of 100 kW or above, installing capacity of 100 kWh or more
  • Conditions: Must participate in Taipower demand response programs for at least 3 years and provide usage data for power dispatch research
  • Application Timeline: The first round of 2026 is expected to open for applications in June; early planning is recommended to secure first-round allocations

Evaluating the feasibility of implementing energy storage or thermal storage systems? Contact our engineering team -- we can assist with load analysis, solution comparison, and subsidy application planning.

4. Demand Response: From Passive Consumption to Active Dispatch

4.1 Taipower Demand Response Program Types

Demand Response (DR) is a mechanism through which Taipower uses economic incentives to encourage users to reduce electricity consumption during power system peak periods[7]. Current main programs include:

  • Planned Load Reduction: Users pre-commit to reduce consumption during specific periods; Taipower provides capacity payments and energy payment rebates. Each execution lasts 2-4 hours.
  • Temporary Load Reduction: Taipower notifies users on the previous day or same day; users cooperate to reduce consumption. Rebate amounts are higher but with greater uncertainty.
  • Demand Bidding: Users self-bid the capacity and price they are willing to reduce; Taipower awards based on system needs. The pricing mechanism is more market-oriented.

4.2 HVAC System Demand Response Participation Strategies

HVAC system demand response operating strategies include:

  • Pre-cooling: 1-2 hours before demand response execution, pre-cool indoor temperatures 1-2°C below the setpoint, utilizing the building's thermal mass to maintain comfort during the execution period.
  • Load Rotation: Multiple chillers or air handling units take turns shutting down, each offline for 15-30 minutes before restarting, rotating to reduce total consumption without affecting comfort across all zones.
  • Temperature Reset: During the execution period, raise the cooling setpoint by 1-2°C, reducing HVAC electricity consumption by approximately 6-12%.
  • Energy Storage Discharge: Activate ice storage or battery energy storage during demand response periods to replace partial chiller operation.

4.3 Demand Response Rebate Calculation

Using the planned program as an example, rebates include capacity payments (approximately NTD 60-80 per kW per month) and energy payments (approximately NTD 2-10 per kWh, depending on time period)[7]. A commercial office building with 2,000 kW contract demand that can reduce 400 kW (20%) through demand response can earn NTD 500,000 to 1,200,000 annually in rebates, while also saving on basic charges through reduced contract demand. Demand response requires no additional hardware investment (if BMS and monitoring systems are already in place), making it one of the most cost-effective peak shaving strategies.

5. Smart Power Management: AI Prediction and Optimal Scheduling

5.1 Load Forecasting: Weather, Schedules, and Historical Patterns

Accurate load forecasting is the foundation of all peak shaving strategies[8]. Modern AI load forecasting systems integrate multi-dimensional data sources:

  • Weather Data: Next-day hourly temperature, humidity, and solar radiation forecasts directly affecting HVAC cooling loads
  • Building Usage Schedules: Meeting room reservations, event calendars, overtime requests, etc., predicting occupancy density and internal heat loads
  • Historical Usage Patterns: Using 1-3 years of hourly electricity data to train machine learning models to identify cyclical patterns
  • Special Events: Handling atypical scenarios such as national holidays, typhoon days, and equipment maintenance shutdowns

5.2 Optimization Algorithms: Joint Dispatch of Ice Storage + Battery + Rates

When a building is equipped with both ice thermal storage and battery energy storage, optimal scheduling becomes a multi-objective mathematical problem: minimizing total electricity costs (basic charges + energy charges - demand response rebates) while meeting cooling requirements. Optimization variables include: chiller hourly output, ice storage charge/discharge schedule, battery charge/discharge schedule, and grid electricity purchase amounts for each time period. Common solving methods include Mixed Integer Linear Programming (MILP) and Deep Reinforcement Learning (Deep RL)[9].

5.3 BMS Integration and Automated Execution

After optimal schedules are generated, they must be executed automatically through the Building Management System (BMS). The BMS receives dispatch commands from the EMS, controlling chiller start/stop, ice storage valve switching, battery charge/discharge power, air handling unit temperature setpoints, and more. Simultaneously, the BMS feeds back real-time operational data to the EMS, forming a closed-loop control system. Key integration interfaces include BACnet, Modbus TCP, and MQTT communication protocols, which should be incorporated into the integration architecture design during the system planning phase.

6. Engineering Planning Recommendations: Strategy Differences for New and Existing Buildings

6.1 New Construction: Reserving Space for Thermal Storage and Energy Storage

Incorporating peak shaving strategies during the planning phase of new buildings can significantly reduce the difficulty and cost of future implementation:

  • Space Reservation: Reserve ice storage tank space in basements or rooftops (approximately 15-20 m² of floor area per 100 RTh), as well as battery energy storage room space (including ventilation, fire protection, and maintenance access)
  • Structural Load: A fully loaded ice storage tank can weigh 2-3 tons per square meter; structural design must account for this in advance
  • Piping Reservation: Pre-install chilled water system piping interfaces and valve spaces for the ice storage circuit
  • Electrical System: Reserve distribution panel grid-tie interfaces and protective relay positions for the energy storage system
  • Communication Infrastructure: Pre-install BACnet/Modbus communication wiring to each equipment control point

6.2 Existing Buildings: Phased Implementation Strategy

Existing buildings have more space and equipment constraints; a phased implementation approach is recommended[10]:

  • Phase 1 -- Data Inventory (1-2 months): Install smart meters and sub-meters, establish an hourly electricity data baseline. Analyze peak occurrence periods, duration, and contract demand utilization to evaluate peak shaving potential.
  • Phase 2 -- Software Peak Shaving (3-6 months): Implement demand controllers and BMS scheduling optimization, using zero-investment strategies like load rotation and pre-cooling to first achieve 10-15% peak reduction. Simultaneously enroll in Taipower demand response programs to begin earning rebates.
  • Phase 3 -- Hardware Implementation (6-18 months): Based on data analysis results, evaluate the investment benefits of ice thermal storage or battery energy storage. Prioritize solutions with shorter payback periods, combined with government subsidies to reduce initial capital expenditure.
  • Phase 4 -- AI Optimization (Continuous Improvement): After accumulating more than one year of operational data, implement AI prediction and optimal scheduling, evolving from "rule-based control" to "data-driven smart dispatch."

Conclusion

HVAC peak shaving is not a single-technology choice but a systems integration engineering effort combining thermal storage, battery storage, demand response, and smart management. Ice thermal storage directly shifts cooling loads using mature phase-change energy storage technology; battery energy storage provides more flexible power dispatch capabilities; demand response transforms passive electricity consumption into active participation in the power market for revenue; and AI scheduling ensures these systems operate collaboratively at optimal efficiency. Under the dual trends of continuously rising electricity prices and energy storage subsidy policies, HVAC peak shaving strategies have evolved from an "advanced option" to a "necessary engineering requirement" for commercial building electricity cost management. The key is finding the optimal combination suited to your building's conditions -- and that starts with accurate load analysis.