R&D Wavelet Based Dispatch for Mixed Energy Fleets

Overview.

Project background.

Full Stack Energy developed and evaluated a novel control approach for coordinating mixed energy systems — generation, storage, and grid-connected assets operating as a single fleet.


These systems share a common failure mode. Individual assets work fine in isolation, but system-level behavior is poor: batteries cycle inefficiently, grid import is mistimed, backup generators fire when they shouldn’t, and operational decisions bear no relation to the cost signals that determine whether the site makes or loses money.


The standard engineering response is to formulate this as a unit commitment and economic dispatch problem, typically solved with mixed-integer linear programming (MILP) or dynamic programming. These methods can find near-optimal solutions, but they scale badly, require commercial solver licences, and assume access to infrastructure that many real-world sites simply don’t have. A 72-generator fleet has 2⁷² possible on/off states per time interval — more than the number of grains of sand on Earth.


This project asked a different question: rather than solving a combinatorial explosion at every timestep, can the structure of the demand signal itself be used to derive dispatch decisions?

Full Stack Energy's

Solution.

The approach is based on wavelet decomposition of the demand profile using the Discrete Wavelet Transform (DWT).

Electricity demand is not a flat number to be met — it is a time-varying signal with layers. A slow baseload that barely shifts from hour to hour. A daily rhythm of morning showers and evening cooking. Fast transients that arrive and vanish in minutes. The DWT separates these layers mathematically, in the same way an audio equalizer separates music into bass, mid-range, and treble:

  • Low-frequency approximation band — the structural baseload
  • Mid-frequency detail bands — diurnal and sub-daily variation
  • High-frequency detail bands — short-term transients and spikes

Each band is then assigned to the asset class whose response characteristics match that timescale:

  • Low-frequency demand → steady baseload generation (gas turbines, grid import)
  • Mid-frequency variation → flexible mid-merit plant (diesel, dispatchable renewables)
  • High-frequency transients → fast-response storage (batteries)

Dispatch decisions emerge from the signal structure rather than from iterating over every possible combination of generator states.


To bridge the gap between oscillatory wavelet components and the binary reality of starting and stopping physical generators, additional control logic handles the edges:

  • Commitment hysteresis prevents rapid cycling of thermal units when wavelet bands oscillate near their commitment threshold — the mechanism that closed the cost gap from 55% to 2% above optimal in testing
  • Merit-order gap-filling ensures that any residual demand not covered by band-mapped assets is served by the cheapest available unit
  • Constraint enforcement for capacity limits, ramp rates, minimum run/stop times, and battery state of charge.

Implementation:

Prototype

Industry:

Complex Energy Systems

Client:

R&D Demonstrator

Visual Exerpts

Demand Behavior Decomposed Across Different Operating Timescales

Image

Cost Comparison

Image

→ View Detailed Case Study: Wavelet-Based Dispatch for an Off-Grid Holiday Park

Key technical highlights:

  • Signal-aware dispatch: the DWT decomposes demand while preserving both frequency content and temporal locality — unlike Fourier methods, it knows when a transient happens, not just that one exists

  • Frequency-to-asset mapping: generators are assigned to wavelet bands by response time, replacing combinatorial optimization with structured signal reconstruction

  • Implicit unit commitment: thermal commitment decisions emerge from band signals plus hysteresis, rather than from solving an explicit binary optimization
  • O(N log N) complexity: the decomposition scales with signal length, not with the number of possible generator states — no exponential blowup
  • Embedded-ready: the complete implementation is a 90 KB binary with no external dependencies. Execution time is sub-millisecond. It runs on a Raspberry Pi
For a detailed technical treatment of the method, including formulation, validation, and benchmarking against dynamic programming and heuristic approaches:

→ Download Whitepaper: Wavelet-Based Generation Dispatch: A Frequency-Domain Approach to Energy Coordination

Outcome.

Impact & results.

Why it matters.

Many energy sites already have the physical assets, metering, and control interfaces to operate well. What they lack is coordination — a way to match what each asset is good at to the part of the demand it should serve. This project demonstrates that the structure of the demand signal itself provides that matching. The same principle applies wherever heterogeneous assets serve variable load:

  • commercial and industrial energy systems

  • hybrid renewable and storage portfolios

  • EV charging and flexible demand systems

  • distributed energy resource aggregation

  • multi-site or mixed-asset energy systems

The value is not marginal optimization improvement. It is the ability to deploy intelligent coordination on hardware that costs less than a commercial solver license, in environments where cloud connectivity cannot be guaranteed, at speeds that allow genuine real-time control rather than periodic re-optimization.

The method was evaluated across synthetic benchmarks, IEEE standard test systems, and real Irish grid data from EirGrid.


On a microgrid fleet (8 assets, 4 classes), wavelet dispatch achieved within 2.3% of the mathematically optimal solution on 7-day profiles — a result that would require exhaustive dynamic programming to match. It outperformed conventional merit-order scheduling by 36–42% on total cost.


On the IEEE RTS-GMLC benchmark (72 generators, 48 hours), where the optimal solution is computationally intractable (2⁷² states), the method dispatched the entire fleet in 0.2 milliseconds.


On real EirGrid demand data, the results were more nuanced. After subtracting volatile wind generation, the residual load loses its multi-scale structure — 96% of signal variance sits in the baseload band. The wavelet decomposition remains correct but adds less discriminating power over simpler methods. On gross demand profiles, which retain their diurnal structure, the method showed 18–20% improvement pathways. This finding — that method effectiveness correlates with signal spectral richness — is a central result, not a caveat.


A companion case study applied the approach to a real-world scenario: a 60-lodge Irish holiday park with micro-hydro, wind, distributed solar, seven independently-owned batteries, diesel backup, and a metered grid connection. The wavelet dispatch coordinated all assets as a single system, kept the diesel generator off throughout normal operation, and identified which asset class would reach capacity first as the park grows — translating directly into capital planning decisions.

Projects.

VIEW MORE

Contact us.

Let's talk about how Full Stack Energy can create a custom solution for your business.