Peak Shaving with Edge Device Control

This tutorial walks you through setting up automated peak shaving for a commercial building. By the end, your edge device will continuously monitor grid power and automatically shed or curtail loads when consumption approaches the peak limit — saving you money on demand charges every month.

What you will build

A complete peak shaving loop:

The edge device runs a Conductor — a two-point (bang-bang) controller with hysteresis — that evaluates grid power against the current peak limit and orchestrates load controllers to keep consumption below the threshold.

Requirements

• Edge device registered to the site (Base Mini, Base Safe, or Base Virtual)
• Energy metering module subscription
• Technical systems module subscription
• Grid power meter connected to the edge device (Modbus, MQTT, OBiX, or other supported protocol) reporting total site consumption in kW
• At least one controllable load connected to the same edge device (battery inverter, HVAC, lighting, etc.)

Architecture overview

The system has three layers, each with a distinct role:

Step 1 — Set up the energy metering technical system

The grid power meter is the conductor's primary input. It must be a connected technical system with a datapoint that reports total site consumption in kW.

Create the technical system

1. Open your site and navigate to Technical Systems
2. Click New Technical System
3. Configure:
   • Name — e.g., "Main Grid Meter" or "Utility Incoming Feed"
   • Type — Energy or Metering
4. Save

Connect the grid meter datapoint

The edge device must have a technical connection (Modbus TCP, MQTT, OBiX, etc.) to the physical meter. Within that connection, configure a datapoint for grid power:

This datapoint serves two purposes:

• Real-time — The edge conductor reads it every evaluation cycle to compare against the limit
• Analytics — The cloud Peak Shaving Calculator reads the aggregated hourly values to compute cost factors and derive optimal limits

Step 2 — Set up controllable load technical systems

The conductor needs loads it can shed (reduce) when the grid exceeds the limit and recover (restore) when there is headroom. Each controllable load is a separate technical system connected via the same edge device.

There are three types of controllable loads:

Battery storage (BatteryLoadController)

For battery inverter systems that can discharge to offset grid consumption.

Technical system setup:

• Type — Battery or Energy
• Connection — to the battery inverter (e.g., Modbus TCP to a Huawei, SMA, or Victron inverter)

Required datapoints:

Conductor parameters:

Curtailable loads (CurtailableLoadController)

For variable-power loads such as VFD-driven pumps, dimmable lighting banks, or EV chargers with adjustable power.

Technical system setup:

• Type — appropriate for the equipment (e.g., Heating, Cooling, Lighting, Electrical)
• Connection — to the controller or VFD

Required datapoints:

Conductor parameters:

Switchable loads (SwitchableLoadController)

For binary on/off loads such as HVAC compressors, water heaters, or non-essential lighting circuits.

Technical system setup:

• Type — appropriate for the equipment
• Connection — to the relay, contactor, or controller

Required datapoints:

Conductor parameters:

Timing flow: shed → wait minOffTimeMin → recover → wait cooldownPeriodMin → shed → ...

Step 3 — Configure the Peak Shaving Calculator

The Peak Shaving Calculator is an AI cloud meter that runs in the cloud and generates hourly power limits based on historical consumption and your tariff structure. The edge conductor fetches these limits and uses them as the real-time target.

For a detailed explanation of how the calculator works — cost factors, thresholds, top-N ranking, and self-learning — see the Peak Shaving Calculator reference.

Create the calculator

1. Navigate to your site → Automation → AI Cloud Meters
2. Click New Cloud Meter → select type Peak Shaving

3. Configure:

What the calculator produces

The calculator creates two output measurement devices:

• Peak Limit — Hourly kW limit values for the forecast window. This is what the edge conductor fetches and enforces.
• Cost Factor — Hourly cost factor values (kW x price weight). Useful for diagnostics.

The edge conductor polls the cloud hourly for updated limits using getPeakShavingLimits. Each limit specifies a time window (from, to) and a power value in kW. The conductor matches the current time to the active limit.

Step 4 — Configure the Peak Shaving Conductor on the edge

The conductor is a Bench node (lar-peak-shaving) that runs on the edge device. It ties everything together: reading the grid meter, fetching limits, and orchestrating load controllers.

How the conductor works

The conductor uses a two-point controller with hysteresis to make shed/recover decisions:

   SHED zone          │  Dead band  │  RECOVER zone
   ───────────────────┼─────────────┼───────────────────
              Peak Limit    Limit × Hysteresis

• When grid power exceeds the peak limit → start shedding loads (highest-priority first)
• When grid power drops below limit x hysteresis → start recovering loads (lowest-priority first, reverse order)
• When grid power is between the two thresholds (dead band) → hold current state, no action

The hysteresis dead band prevents rapid on/off oscillation. With a limit of 100 kW and hysteresis of 0.85, the conductor sheds at 100 kW and doesn't recover until power drops below 85 kW.

Node configuration

Open the edge device's Bench editor and add the Peak Shaving node.

Load priority strategy

The conductor processes loads in the order you define. Design your priority list based on:

Reporting

Every shed and recover decision is reported to the cloud as an OPTIMIZATION_PEAK_SHAVING event with full context:

These events appear in the site's Logs and are linked to the relevant technical system, giving you a complete audit trail of every peak shaving action.

Example: office building with battery and HVAC

Consider a 500 kW office building with:

• Grid meter — Modbus TCP to the main utility meter
• 100 kWh / 50 kW battery — Modbus TCP to a Huawei inverter
• 2 x rooftop AHUs — BACnet/IP, 30 kW each, curtailable to 10 kW
• Water heater — Relay control, 15 kW, switchable

Calculator output

The Peak Shaving Calculator determines that the optimal daytime limit is 380 kW based on the top-3 billing model with a 10% reduction target.

What happens when a spike hits

1. 09:15:00 — Grid power spikes to 395 kW (morning startup + EV charging). The conductor detects 395 > 380 (above limit).
2. 09:15:00 — Conductor walks the priority list. Battery controller proposes discharging 15 kW (the excess). Decision applied — battery writes -15 kW to the inverter.
3. 09:15:10 — Next evaluation. Grid reads 380 kW. In the dead band (between 380 and 323 kW). Hold.
4. 09:45:00 — Morning startup subsides. Grid drops to 310 kW. Below 380 x 0.85 = 323 kW. Conductor recovers the battery — writes 0 kW (idle), unlocks the datapoint.

If the spike had been larger (e.g., 420 kW), the battery alone would not be enough. The conductor would continue down the priority list: battery discharges 50 kW (max), then AHU-1 curtails from 30 kW to 20 kW (frees 10 kW). Total reduction: 60 kW → grid drops from 420 to 360 kW.

Monitoring and troubleshooting

Node status display

The conductor node shows real-time status in the Bench editor:

Common issues

Best practices

What's next