Smart HVAC Automation: How Occupancy Sensing & AI Scheduling Cut Cooling Costs in Dhaka

No other operating cost hits Dhaka commercial building owners harder than air conditioning. Cooling a mid-rise office building through Bangladesh's 8-plus month summer — with outdoor temperatures regularly exceeding 36°C and humidity making it feel several degrees hotter — demands enormous energy. For many commercial buildings in Dhaka, HVAC accounts for 55–65% of total electricity consumption, significantly above the global average of 50%.

The good news is that smart HVAC automation can cut these costs by 25–40% without compromising occupant comfort. The technology is mature, the Bangladesh payback periods are short, and the BPDB tariff structure actually amplifies the financial return for buildings that reduce peak demand. This article explains exactly how occupancy sensing, AI scheduling, and demand response work — and what the financial impact looks like for a real Dhaka office building.

Dhaka's Cooling Challenge

Before exploring the solutions, it helps to understand the specific characteristics of Dhaka's cooling challenge. They are distinct from temperate climate markets where much building automation research originates.

Duration: Dhaka's cooling season runs from approximately March through October — roughly 8 months. During this period, outdoor temperatures routinely reach 34–40°C during the day. Even at night, temperatures rarely drop below 26–28°C from May through September. This means air conditioning must run nearly continuously, with minimal benefit from economiser cooling (using outdoor air to reduce mechanical cooling). Every hour of HVAC operation contributes meaningfully to the energy bill.

Humidity: Bangladesh's tropical monsoon climate combines high temperatures with high humidity. Relative humidity regularly exceeds 80% during the monsoon months (June–September). HVAC systems must dehumidify supply air as well as cool it — a process that requires additional energy beyond simple temperature control. Smart humidity management, alongside temperature setpoint control, is therefore an important element of Dhaka HVAC optimisation.

Occupancy mismatch: Despite the long cooling season, Dhaka commercial buildings share the same occupancy patterns as offices globally — they are substantially occupied for only 6–8 hours per working day. Weekends, public holidays, and evenings see near-zero occupancy. Running full-capacity HVAC during these unoccupied periods is expensive and entirely unnecessary.

Peak demand charges: Bangladesh Power Development Board (BPDB) and its distribution entities charge commercial consumers not only for energy (kWh) consumed but also for peak demand (kVA). The demand charge component can represent 20–30% of a large commercial electricity bill. Reducing peak HVAC load — even briefly — produces financial savings that compound monthly.

Occupancy Sensing: The Foundation of Smart HVAC

Effective HVAC automation starts with knowing where people are. Occupancy sensing technology has advanced considerably and there are now three main approaches, each with different accuracy, cost, and privacy characteristics.

PIR (Passive Infrared) Sensors

PIR sensors detect infrared radiation emitted by human bodies as people move within the sensor's field of view. They are the most commonly installed occupancy sensors in commercial buildings — low cost, reliable, and well-proven. PIR sensors in HVAC applications are typically ceiling-mounted and cover a detection zone of 10–15 metres diameter.

PIR sensors work well for detecting motion — someone walking through a space is detected immediately. Their limitation is with sedentary occupants. A person sitting still at a desk for more than a few minutes may not be detected by a basic PIR sensor. For HVAC control (where setback delays of 15–30 minutes are acceptable), this limitation is less critical than for lighting control (where a 5-minute delay causes noticeable disruption).

CO2 Sensors

CO2 (carbon dioxide) sensors measure the concentration of CO2 in the room air. Human occupants continuously exhale CO2, raising the room CO2 level above the outdoor baseline of approximately 420 parts per million (ppm). The rate of rise depends on the number of occupants and the ventilation rate.

CO2 sensors do not detect individual presence — they detect aggregate occupancy load. A CO2 level of 600 ppm suggests light occupancy. 900 ppm suggests moderate occupancy. Above 1,000 ppm, ventilation should be increased for comfort and cognitive performance. Above 1,500 ppm, the space is over-occupied for its ventilation rate.

For HVAC demand-controlled ventilation (DCV), CO2 sensors are ideal. The HVAC controller adjusts fresh air ventilation rates in real time based on CO2 measurements — providing exactly the ventilation needed for the actual occupancy level, no more and no less. ASHRAE Standard 62.1 includes CO2-based DCV as an approved ventilation design method. The energy saving from DCV compared to fixed-rate ventilation is typically 10–20% on the ventilation energy component.

People-Counting Cameras

People-counting cameras — typically installed at room or zone entry points — use computer vision to count the number of people entering and exiting a space. Integrated with an access control system or a standalone counting controller, they provide a real-time headcount for each zone. This headcount is far more accurate than either PIR or CO2 for understanding exact occupancy levels.

People-counting data enables the HVAC controller to scale cooling capacity directly proportional to occupancy. If a conference room designed for 20 people has 8 people in it, the HVAC system supplies 40% of design capacity — not 100%. This granularity produces significantly better energy performance than simple occupied/unoccupied binary control.

For privacy compliance under Bangladesh's PDPO 2025, people-counting systems should use body counting (shape recognition) rather than facial recognition, ensuring that no personally identifiable information is captured or stored.

AI Scheduling: Learning Patterns and Predictive Pre-Cooling

Static HVAC scheduling — "turn on at 7 AM, turn off at 7 PM, Monday to Friday" — is better than nothing but far from optimal. Real building occupancy is more complex. Mondays often see later arrival than Fridays. Ramadan changes working hours significantly. Conference days see full-building occupancy while holiday Fridays see almost no one. A fixed schedule cannot adapt to these patterns.

AI-based HVAC scheduling learns from historical occupancy and environmental data to produce dynamically optimised cooling schedules. The algorithms are typically trained on data from occupancy sensors, access control badge records, calendar systems, and weather forecasts. Over weeks and months of operation, they develop increasingly accurate predictions of occupancy patterns for each zone on each day type.

Predictive pre-cooling is one of the most valuable outputs of AI scheduling. The fundamental challenge of Dhaka HVAC is thermal mass: a building that has been at 30°C overnight cannot be cooled to 23°C in 10 minutes. Conventional fixed schedules compensate by starting HVAC very early — often 2 hours before occupancy — to ensure comfortable conditions by arrival time. This wastes considerable energy.

AI scheduling calculates the minimum start time needed to achieve comfort temperature by occupancy time, using a model that accounts for outdoor temperature (from weather forecast data), building thermal mass (from historical cool-down rate data), and chiller efficiency at different operating points. On a mild morning (30°C at 7 AM), HVAC may start 45 minutes before occupancy. On a hot morning (36°C at 7 AM after a warm night), it may start 90 minutes earlier. On days when building occupancy is predicted to be low, setpoints may be raised 1–2°C throughout the day.

The result: occupants arrive to comfortable conditions without the energy waste of running full HVAC for 2 hours against an empty building. AI scheduling typically reduces HVAC energy by 12–18% compared to conventional fixed-schedule BMS control.

Learning and adaptation: AI scheduling continuously updates its models based on observed outcomes. If the system predicted that a floor would be unoccupied on a particular afternoon but sensors recorded moderate occupancy, the model updates accordingly. Over a 3–6 month learning period, scheduling accuracy improves substantially.

Demand Response: Cutting Peak Load to Reduce Electricity Bills

BPDB's commercial tariff structure for large consumers creates a strong financial incentive for peak demand reduction. The demand charge is calculated based on the maximum demand recorded in any 15-minute interval during the billing month. Even a single peak event — driven by, say, all HVAC units starting simultaneously after a power interruption — can set the demand charge for the entire month.

Smart HVAC automation implements demand response in two ways:

Staggered equipment start (soft start sequencing): Rather than starting all air handling units simultaneously after a power restoration or morning startup, the BMS sequences equipment starts with delays between each unit. This prevents the current inrush peak that simultaneous starting would create. The full cooling capacity is available within 5–10 minutes, but the peak current draw is dramatically reduced.

Active demand limiting: When the BMS monitors real-time demand approaching a preset threshold — say, 85% of the building's contracted demand capacity — it automatically implements shedding actions. HVAC setpoints in non-critical zones (corridors, unoccupied meeting rooms) are raised by 1–2°C for 15–30 minutes. This small, temporary change in setpoint is imperceptible to occupants but meaningfully reduces the electrical load profile during the critical measurement window.

For a Dhaka commercial building with a contracted demand of 500 kVA and a demand charge rate of BDT 400 per kVA per month, reducing average peak demand from 480 kVA to 420 kVA (12.5% reduction) saves BDT 24,000 per month — BDT 288,000 per year in demand charges alone. Combined with energy consumption savings, the financial return is compelling.

The BPDB Demand Charge Structure

Bangladesh Power Development Board and its licensed distribution companies — Dhaka Power Distribution Company (DPDC) and Dhaka Electric Supply Company (DESCO) for the Dhaka metropolitan area — apply demand charges to commercial consumers above certain consumption thresholds.

For 2025/2026, the applicable tariff for large commercial consumers (LV supply, 33 kVA and above) includes:

• Energy charge: approximately BDT 9.5–11.5 per kWh depending on consumption band
• Demand charge: approximately BDT 350–500 per kVA per month based on maximum demand
• Power factor penalty: additional charges apply when power factor falls below 0.85

The demand charge component creates a clear financial reward for peak demand reduction. Unlike energy efficiency improvements (which reduce the kWh charge), demand response improvements reduce the kVA charge — a separate and significant cost element.

Smart HVAC automation provides real-time visibility of current demand through integration with smart energy meters. The BMS dashboard shows current kVA draw, the 15-minute trend, and the month-to-date peak demand. Facility managers can make informed decisions about load management, and the automated demand response logic handles routine peaks without human intervention.

Case Study Approach: Multi-Floor Dhaka Office Building

Consider a representative scenario: a 12-floor, 12,000 m² commercial office building in Dhaka's Gulshan-Banani corridor. The building has a central chiller plant (2 × 300 RT chillers) serving floor-by-floor air handling units (AHUs), with fan coil units (FCUs) for perimeter and cellular office areas. Prior to automation, the building ran on basic time-clock control with manual HVAC adjustments by the facility manager.

Pre-automation energy profile:

• Annual electricity consumption: approximately 1,560,000 kWh (130 kWh/m²/year)
• Average monthly demand peak: 550 kVA
• Annual electricity bill (energy + demand charges): approximately BDT 2.3 crore

Automation interventions implemented:

• CO2-based demand-controlled ventilation in AHUs (all floors)
• PIR occupancy sensors for zone HVAC setback (all floors)
• AI scheduling with weather forecast integration
• Active demand response with 15-minute meter monitoring
• BMS remote monitoring and management interface

Post-automation results (12-month post-commissioning):

• Annual electricity consumption: approximately 1,012,000 kWh (35% reduction)
• Average monthly demand peak: 440 kVA (20% reduction)
• Annual energy savings: approximately 548,000 kWh
• Annual demand charge savings: BDT 528,000 (at BDT 400/kVA/month × 12 months × 110 kVA reduction)
• Total annual financial saving: approximately BDT 77 lakh (energy) + BDT 5.3 lakh (demand) = approximately BDT 82 lakh
• Automation system investment: BDT 1.8 crore
• Simple payback: approximately 2.2 years

This scenario is representative of outcomes SSAL has documented across comparable commercial building projects. Results vary based on pre-automation baseline, building type, HVAC equipment efficiency, and occupancy patterns — but the 25–40% energy saving range is consistently achievable.

Implementation Pathway

Implementing smart HVAC automation is a phased process. The recommended approach for existing buildings:

Phase 1 — Baseline and audit (4–6 weeks): Install temporary energy and occupancy monitoring to establish a baseline. Identify the highest-waste patterns (overnight HVAC, unoccupied floors running at full capacity, demand peaks). Develop the business case with documented projected savings.

Phase 2 — BMS upgrade (8–16 weeks depending on building size): Install DDC (Direct Digital Control) controllers on all AHUs and chillers. Connect to a central BMS with graphical user interface. Implement time scheduling and basic occupancy setback.

Phase 3 — Sensors and DCV (4–8 weeks): Install CO2 sensors in all occupied zones. Connect to BMS for demand-controlled ventilation. Install PIR sensors for zone setback control. Commission and tune sensor thresholds and response parameters.

Phase 4 — AI scheduling and demand response (2–4 weeks): Activate AI scheduling module with historical data feeding. Configure demand response thresholds and shedding sequences. Integrate with smart energy meters for real-time demand monitoring.

Phase 5 — Optimisation (ongoing): Monthly energy reporting, quarterly setpoint review, seasonal schedule updates. Annual system performance audit comparing actual to projected savings.

Conclusion

Dhaka's cooling challenge is real and expensive. But it is also addressable — with smart HVAC automation that matches cooling capacity to actual occupancy, learns from historical patterns, and actively manages demand charges. The financial case is strong: payback periods of 2–4 years are consistently achievable, and the ongoing energy savings compound over the entire building life.

SSAL's building automation team has the expertise, the technology partnerships, and the Bangladesh market experience to design, install, and commission smart HVAC automation systems that deliver documented results.

Get Free Consultation — Contact our HVAC automation specialists for a building energy assessment. We will review your current consumption profile, identify the highest-impact automation opportunities, and give you a projected ROI specific to your building. The first step toward lower cooling costs starts with a conversation.