Seedlabs

TrendSplit: Month-Ahead Utility Load Forecasts with a Visible Trend–Seasonal Split

A service for power companies that predicts electricity demand by splitting base load from seasonal spikes.

A forecasting service that gives electricity utilities week-to-month-ahead regional load forecasts from years of hourly data, with each forecast broken into underlying demand trend versus recurring seasonal peaks so planners can justify outage and procurement decisions.

Computer ScienceTime Series Analysis and Forecasting
Hybrid Graph-Based Memory and SchedulingEnergy & utilities — long-horizon regional load forecasting for system operators, retailers, and trading desks

Concept

A hosted forecasting engine for electricity utilities and energy traders. It ingests 2–5 years of hourly substation- or region-level load history (with temperature as a covariate), runs a seasonal–trend decomposition front end that extracts the global demand trend and recurring seasonal shape, and lets a frequency-enhanced Transformer model the local residual detail. Planners get two things per region: a week-to-month-ahead load forecast, and the decomposition behind it — how much of next month's expected demand is structural trend growth versus seasonal peak. That split is the wedge: outage scheduling, reserve procurement, and regulatory reporting all need defensible reasoning, not just a number. Because compute scales linearly with lookback length, one deployment covers a full utility's set of regional series on commodity hardware rather than a GPU cluster.

Why now

FEDformer [0] demonstrates exactly this architecture: pairing seasonal–trend decomposition with a frequency-enhanced Transformer cut multivariate forecasting error by 14.8% versus state-of-the-art baselines across six widely used benchmarks, while achieving linear complexity in sequence length — the long-horizon accuracy and multi-series efficiency a utility-wide deployment needs. The transformer time-series survey [1] identifies energy as a leading application domain for transformer forecasters but names computational cost as a key barrier to adoption, which this linear-complexity result directly removes.

Written by GLM · 8 Oct 2026

AI assessment

Backed by 2 papers89

A high-utility forecasting tool that leverages FEDformer's linear complexity to provide the 'why' behind load forecasts, creating a strong value proposition for regulatory and procurement compliance in utilities.

Evidence strength
5/5
The idea directly implements the FEDformer architecture's core strengths—seasonal-trend decomposition and linear complexity—to solve the specific computational and structural problems identified in the provided research.
Market pull
4/5
Utility planners and trading desks have high urgency for defensible forecasts to justify expensive procurement and outage decisions to regulators.
Novelty & moat
4/5
While other catalog ideas focus on EV-specific load or short-term occupancy, this targets regional utility-scale long-horizon forecasting with a specific focus on the trend-seasonal split.
Feasibility
4/5
The FEDformer model is a known architecture that can be implemented on commodity hardware, making a prototype highly achievable given available historical load data.
Wedge clarity
5/5
The 'trend vs. seasonal' split is a sharp, specific wedge that transforms a generic forecast into a decision-support tool for regulatory justification.
Simplicity / focus
5/5
The product is focused on a single, clear output: a decomposed long-term load forecast for a specific set of users.

Scored by Gemma 4 · 8 Oct 2026. Rated against a fixed rubric (evidence, market, novelty, feasibility, wedge, simplicity), relative to the closest ideas already in the catalog. A prior estimate to compare ideas before real-world signal arrives.

Persona discussion

AI personas trained on real people's expertise debate this idea as it evolves.

View the discussion →

Act on this idea

Ideas only matter if someone runs with them. Your message goes straight to the founder's inbox — nothing is stored on our servers.

Who benefits

  • Amperoncompany

    This energy-analytics firm sells load forecasts to retailers and utilities; a more accurate, linear-complexity engine lets it cover more customer series without proportionally more compute spend.

  • National Gridcompany

    Its UK transmission and system-operation planning teams need month-ahead regional demand views to schedule maintenance outages and procure reserves; the trend-vs-seasonal split gives auditable justification for those decisions.

  • EDFcompany

    One of Europe's largest generators and retailers needs long-horizon load forecasts for generation scheduling and hedging retail supply; a 14.8% multivariate error reduction translates directly into lower balancing costs.

  • Statkraftcompany

    Europe's largest renewable generator trades across multiple power markets and needs month-ahead demand signals to time bids and plant maintenance; interpretable forecasts support its market-operations desk.

Research it builds on

  1. Fedformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
    Tian Zhou, Ziqing Ma, Xue Wang et al. · 2026 · 213 citations
    All ideas from this paper →
  2. A survey of transformer networks for time series forecasting
    Jingyuan Zhao, Fulin Chu, Lili Xie et al. · 2025 · 45 citations
    All ideas from this paper →
Leave feedback
feasibility