Automated Peak Demand Control & Load Shedding Systems for Industrial Plants

1. Peak Demand and Its Hidden Cost

Thai utility tariffs for medium businesses and above have two main components: an energy charge for actual units consumed, and a demand charge based on the highest average power over any 15-minute window in the billing period. A single short episode of heavy machines running simultaneously can therefore set the demand charge for the entire month, and under TOU rates, the time at which the peak occurs also affects the applicable price.

Many plants unknowingly pay excessive demand charges because peaks arise from timing coincidences, such as several chillers restarting together after the lunch break, or an oven cycle overlapping a fully loaded air compressor. A peak demand controller solves this by tracking accumulated power within the live 15-minute window and shedding or deferring pre-approved interruptible loads before the average crosses the configured target.

2. Controller Principles and Load Prioritization

The system's core is an algorithm forecasting the end-of-window 15-minute average from the current accumulation rate. If the trend will exceed the target, loads are shed in a pre-defined priority order, starting with those least affected by a brief stop, such as ice makers, selected air conditioning zones, transfer pumps, or standby compressors, and are restored automatically when a new window begins or the trend recedes. Loads directly tied to production or safety are never enrolled for shedding.

Load prioritization is the step that most needs production team involvement. Engineers should run workshops with each system's owner to define how long and how often each load may be shed and under what conditions shedding is forbidden, for example a cold room only while its temperature stays in range. The outcome is a shedding table everyone accepts, which matters more to the system's longevity than the hardware itself, because a controller that disrupts operations will eventually be switched off by its users.

• Real-time forecasting of the 15-minute window average
• Shed loads by a priority order agreed with production
• Safety and core production loads are never enrolled
• Automatic restoration when the peak trend recedes

3. Analyzing the Load Profile before System Design

Design begins with at least 12 months of 15-minute load profile data from the utility meter or internal metering, establishing when each month's peak occurred, what caused it, and how far it stood above the normal operating baseline. The gap between peak and baseline is the size of the saving opportunity. Temporary sub-metering on major machines then identifies which loads actually create the peaks and which can genuinely be shed from a process standpoint.

The analysis leads to a demand target balancing savings against shedding frequency: too low and the system sheds so often it disrupts operations, too high and little is saved. Good practice starts with a target that removes only the coincidence-driven excess, then lowers it step by step as real behavioral data accumulates month by month, with periodic reviews alongside the production team.

4. Success Factors and Common Pitfalls

Common pitfalls include setting the demand target by intuition rather than profile data; shedding the same load so repeatedly that it degrades, such as a compressor short-cycled too often; failing to synchronize the controller's demand window with the utility meter; and neglecting to revise the shedding table when processes change. Restoration behavior also deserves care: loads returning simultaneously after shedding can create a fresh peak unless restoration is staggered with delays.

Successful systems share common elements: good profile data from the start, production team ownership of the priority table, a dashboard showing status and savings visible to all parties, and seasonal review of settings. Where a plant has its own generation, such as rooftop solar or battery storage, coordinating the peak controller with those resources unlocks a further level of demand reduction, provided the combined control logic is engineered carefully.

• Synchronize the 15-minute window with the utility meter
• Stagger load restoration to prevent a rebound peak
• Show savings on a dashboard visible to all stakeholders
• Revise the shedding table whenever processes change