Can AI Write an HVAC Sequence of Operations?

Can AI Write an HVAC Sequence of Operations?

Let's cut to the chase. The brain of any modern commercial HVAC system is its sequence of operations (SOO). It’s the detailed, step-by-step logic that tells multi-million dollar pieces of equipment how to work together. For decades, writing a good SOO required years of experience, a deep understanding of thermodynamics, and a lot of painstaking manual work. Now, AI is on the job site. The question is, can you trust it with something this critical?

The short answer is yes, but with a big 'but'. AI can be an incredibly powerful assistant for drafting an SOO. It can save you hours of typing and help you think through complex logic. But it is not a replacement for your experience. It's a tool, like a digital multi-meter or a recovery machine. You are still the one in charge, and you are still responsible for the final product.

What is a Sequence of Operations Anyway?

Before we let a robot write one, let's get on the same page. A sequence of operations is the written instruction manual for a Building Automation System (BAS). It dictates exactly how the HVAC equipment should operate under all conditions.

A solid SOO tells the system:

• When to turn on and off (occupancy schedules).
• What temperature and humidity to maintain (setpoints).
• How to respond to changing loads (e.g., a full conference room vs. an empty one).
• What to do when something goes wrong (alarms and safety shutdowns).

This isn't just about comfort. A well-written SOO is crucial for energy efficiency, equipment longevity, and occupant safety. Standards like ASHRAE Guideline 36 provide a solid framework for high-performance sequences, but every building is unique. That's where your expertise comes in, and where AI can help you work faster.

Where AI Fits in Your Workflow

Think of AI as a smart apprentice. It's read every textbook and manual ever published, but it's never actually been on a roof or in a mechanical room. It can give you a fantastic starting point, but it needs your guidance and real-world knowledge to get the job done right.

Here’s how you can use AI to write an SOO:

1. Initial Drafting: The most time-consuming part of writing an SOO is getting the basic structure and language down. AI can generate a comprehensive draft in seconds based on your inputs. This saves you from starting with a blank page.

2. Translating Ideas: You know what you want the system to do. You can describe it in plain English, and AI can translate that into the formal, structured language required for an SOO document.

3. Exploring Options: Not sure about the best way to stage your chillers? Want to compare a simple lead-lag sequence with a more complex load-based one? You can ask AI to generate both versions, letting you quickly compare the pros and cons.

4. Standardizing Language: If you work for a company that wants all its SOOs to follow a specific format, you can give AI your template and have it generate new sequences that match it perfectly. This improves consistency across all your projects.

Getting Started: Your First AI-Generated Sequence

Let's try a common scenario: a single-duct, pressure-independent VAV terminal unit serving an office space. We need it to handle heating and cooling based on a zone sensor and an occupancy schedule.

Instead of writing it from scratch, you can give a large language model (like ChatGPT or Claude) a detailed prompt. The key is to be specific.

Act as a senior mechanical controls engineer. Write a detailed sequence of operations for a pressure-independent VAV terminal unit with a series fan and a three-stage electric reheat coil.

System Details:
- The VAV box serves a single office zone.
- It has a modulating cooling damper.
- The reheat coil has three stages of electric heat.
- The series fan runs continuously during occupied hours.
- The zone has a temperature sensor and an occupancy sensor.

Operating Parameters:
- Occupied Hours: Monday-Friday, 7:00 AM to 6:00 PM.
- Occupied Cooling Setpoint: 74°F
- Occupied Heating Setpoint: 70°F
- Unoccupied Setpoints: 85°F (cooling), 60°F (heating)
- Minimum Damper Position (occupied): 30% of max flow.

Write the sequence covering Occupied Mode, Unoccupied Mode, and Morning Warm-up/Cool-down modes. Be clear about the logic for damper modulation and staging the electric heat.

The AI will take this information and generate a well-structured SOO. It will have sections for each mode of operation and define the logic for how the damper and heat stages respond to the zone temperature relative to the setpoints. It's a solid 80% solution that you can then refine.

Leveling Up: Complex Systems

AI isn't just for simple VAV boxes. You can use it for much larger and more complex systems, like a central chilled water plant. The principle is the same: the more detail you provide, the better the output will be.

For a chilled water plant, you'd need to provide information like:

• Number and capacity of chillers.
• Number of primary and secondary pumps (and whether they're variable speed).
• Cooling tower details (number of cells, fan speed control).
• Key setpoints (chilled water supply temperature, differential pressure).
• Desired staging logic (e.g., lead-lag, rotation based on runtime, load-based staging).
• Safety and operational alarms.

Giving the AI this level of detail allows it to draft a sequence that accounts for pump rotation, chiller staging, and condenser water temperature control. It's a massive head start that frees you up to focus on the finer points of the system's performance. Better ops management starts with better documentation, and this is a fast track to getting it done.

Act as an expert in HVAC control systems. Draft a sequence of operations for a primary-secondary chilled water plant with two 500-ton water-cooled centrifugal chillers (CH-1, CH-2) and associated cooling towers.

Equipment:
- Two variable speed primary pumps (PCHWP-1, PCHWP-2).
- Two variable speed secondary pumps (SCHWP-1, SCHWP-2).
- Two two-cell cooling towers (CT-1, CT-2) with variable speed fans.

Logic Requirements:
- The plant is enabled when the outside air temperature is above 65°F and there is a call for cooling from the building.
- Use lead-lag control for the chillers, rotating the lead chiller weekly based on runtime.
- The secondary pumps shall modulate to maintain a differential pressure setpoint of 15 PSID at the furthest point in the loop.
- Stage the second chiller on if the lead chiller runs at 95% capacity for 10 minutes and the chilled water supply temperature rises 2°F above its setpoint of 44°F.
- Control cooling tower fans to maintain a condenser water supply temperature of 85°F or 10°F above the outdoor wet-bulb temperature, whichever is higher.
- Include all necessary pump and valve interlocks for safe startup and shutdown.

The Human Factor: You're Still the Boss

This is the most important part. An AI does not have a license. It has no liability. It has never had to answer a 2 AM service call because a system went down.

You are the professional. The final SOO is your responsibility. Always review every line of an AI-generated sequence with a critical eye.

Ask yourself:

• Does this make physical sense? AI can sometimes suggest things that are impossible, like opening two valves that would create a dead-end.
• Are the safety interlocks correct? Does the sequence ensure a pump has flow before enabling a chiller? Are freeze-stats and high-pressure cutouts properly accounted for?
• Is it compliant with local codes? AI doesn't know the specific building codes for your project in Austin, Texas versus Boston, Massachusetts.
• Is it practical for the specific equipment being installed? The sequence must match the capabilities of the actual controllers, sensors, and actuators on the job.

Blindly copying and pasting an AI-generated sequence is not just lazy; it's dangerous. It can lead to equipment damage, massive energy waste, or unsafe conditions. The AI gets you on the 20-yard line. You have to carry the ball into the end zone.

Troubleshooting with AI

Beyond writing new sequences, AI can also be a powerful diagnostic partner. If you have a system that isn't behaving correctly, you can feed the existing SOO and the current system status into an AI and ask for potential causes.

Act as a senior HVAC controls troubleshooter. I have an air handling unit (AHU-1) that is failing to meet its discharge air temperature setpoint of 55°F. Here is the relevant part of the sequence of operations and the current system status. What are the most likely causes?

Sequence Excerpt:
"The chilled water valve (CHW-V-1) shall modulate to maintain a discharge air temperature of 55°F. The valve is commanded by a PID loop. If the temperature is above setpoint, the valve shall modulate open. If it is below setpoint, it shall modulate closed."

System Status:
- Discharge Air Temperature: 62°F
- Chilled Water Valve Command: 100% open
- Mixed Air Temperature: 78°F
- Chilled Water Supply Temperature at AHU: 50°F
- Chilled Water Return Temperature at AHU: 54°F

Based on this, list the top 5 most likely causes in order of probability, from a stuck valve to a sensor issue.

This helps you think through the problem methodically. The AI might suggest things you hadn't considered, like a low delta-T across the coil pointing to low water flow, or a stratified mixed air temperature causing the sensor to read incorrectly.

AI is a new tool in the truck. Like any tool, it takes skill to use it well. It won't replace your experience, but it can amplify it, helping you get the boring work done faster so you can focus on what you do best: delivering efficient, reliable, and safe HVAC systems.