Load, Demand and Capacity Forecasting
An electricity network or energy-intensive facility must maintain enough capacity to serve demand when it occurs. However, recorded load changes with weather, tariffs, economic activity, customer behaviour, outages and generation behind the meter. A forecast that extends historical load without explaining these drivers can fail precisely when management must commit reserves, maintenance or capital.
Marketways develops conditional forecasts that connect demand uncertainty with operating choices. The analysis distinguishes consumption from net load, reconciles forecasts across locations and time horizons, and shows how much capacity each plausible condition requires.
Define the capacity decision and its time horizon
A forecast has value only when it changes a decision. System operators may need a day-ahead reserve, while infrastructure planners may need a ten-year investment range. A facility manager may instead need to schedule equipment, maintenance or contracted demand.
Marketways begins by defining the decision, forecast horizon, geographic level and acceptable service risk. We also identify which responses remain available when actual load departs from the forecast.
This framing determines the required frequency, uncertainty range and lead time. It prevents one forecast from being presented as suitable for operational dispatch, annual budgeting and long-term capacity investment.
Separate customer demand from recorded net load
Metered load records electricity drawn from the network after embedded generation, storage, outages and demand response have changed the amount. Therefore, a fall in recorded load may reflect lower consumption, more rooftop generation or customers shifting use to another period.
Marketways combines meter data with distributed generation, outages, tariffs and programme participation. Where the evidence permits, we estimate customer demand and the contribution of each intervening factor.
The distinction matters because each explanation has a different capacity implication. Lower underlying demand may reduce the need for investment, while greater embedded solar can deepen the evening ramp even if midday network load falls.
Model weather as a local and nonlinear driver
Temperature often has little effect within a comfortable range and a much larger effect beyond it. Humidity, wind, building stock and appliance use can change that relationship. Consequently, a national average can conceal a local peak that constrains one feeder or facility.
Marketways tests nonlinear weather relationships at the level where capacity is managed. We use forecast weather rather than observed weather when evaluating a decision that must be made in advance.
This discipline avoids overstating performance with information that operators would not have possessed at the decision time. It also reveals the locations and conditions that create the greatest load sensitivity.
Reconcile forecasts across the system hierarchy
Capacity decisions occur at several levels. A system forecast can be accurate even when regional errors cancel one another, while independently produced local forecasts may not add to the system total.
Marketways develops forecasts for the system, region, feeder, facility or customer group and reconciles them into a coherent hierarchy. The method preserves useful local evidence while ensuring that aggregated forecasts agree.
Management can then see where demand is likely to arise, not only how much the entire system may serve. This location matters when transmission, distribution or site capacity cannot move freely between constrained areas.
Represent the demand range rather than one precise value
A point forecast conceals the range of outcomes that capacity may need to serve. The same expected peak can carry very different risk when uncertainty is narrow or when extreme weather, electrification or customer response creates a wide upper tail.
Marketways develops calibrated probabilistic forecasts. Prediction intervals and scenarios show the likelihood of different demand levels at the relevant time and location.
Management can use these ranges to compare reserve, flexibility and investment choices. The forecast does not dictate one risk tolerance; it makes the consequence of that tolerance explicit.
Treat structural change as a forecast assumption
Electrification, data centres, distributed generation, storage and tariff reform can change the relationship between historical load and future demand. A model may remain accurate during ordinary periods yet fail after one of these changes alters customer behaviour or system shape.
Marketways represents material changes as explicit assumptions or adoption scenarios. We estimate which historical relationships may continue, which require adjustment and where evidence remains weak.
This approach is more defensible than hiding a strategic assumption inside a single growth rate. Management can revise the forecast when adoption, policy or investment evidence changes.
Test how tariffs and demand response move load
A tariff or demand-response programme may reduce one peak and create another. Customers also differ in their ability and willingness to shift consumption. A post-programme load reduction cannot be attributed to price if weather, activity or technology changed at the same time.
Marketways uses experiments or credible comparison groups where possible. We estimate total consumption, timing, persistence and differences across customer segments.
The analysis shows how much response the capacity plan can rely upon. It also prevents a temporary or selective response from being treated as firm capacity.
Connect forecast uncertainty with available capacity options
Management must decide which demand range the system should serve and what reliability standard applies. The available responses may include generation, storage, contracted imports, demand response, maintenance rescheduling, network reinforcement or capital investment.
Marketways compares these options across the same demand scenarios. We include lead time, availability, operating constraints, cost and the consequence of shortage or overbuilding.
The result identifies when additional fixed capacity is necessary and when flexible action can manage uncertainty more economically. Forecast quality becomes an input to a decision rather than an isolated modelling score.
Use Agentic AI to monitor exceptions and prepare approved responses
Actual load can depart from the expected range because weather changed, a large customer altered consumption, distributed generation failed or data quality deteriorated. Operators need to distinguish these explanations before selecting a response.
An Agentic AI workflow can retrieve permitted evidence, compare the current condition with approved scenarios and prepare the relevant response options for review. It can also route a forecast exception to the responsible operating or planning team.
The workflow must expose its sources, uncertainty and model version. It must not change safety limits, dispatch material capacity or commit investment outside its authority.
Evaluate the forecast through the decision it supports
Average percentage error can reward good performance during ordinary periods while concealing missed peaks. It can also obscure systematic bias in one region or customer group.
Marketways evaluates calibration, peak error, regional bias, missed constraint events and performance after structural change. We compare the forecast with a credible baseline and estimate the operating or investment consequence of its errors.
A more complex model must improve the decision enough to justify its data, maintenance and governance cost. Statistical accuracy alone does not establish business value.
A hot-weather capacity decision illustrates the method
Suppose a utility expects a summer peak above its current firm capacity. A single high-growth forecast appears to justify immediate reinforcement.
Marketways would separate underlying consumption from rooftop generation and test local weather sensitivity. Reconciled probabilistic forecasts may show that one constrained area drives the upper range. Scenario analysis may then compare reinforcement with a targeted demand-response contract, storage or a staged capacity addition.
Management can select a response that meets its reliability criterion and preserve an option to expand if demand materialises. The decision no longer depends on one apparently precise peak.
How Marketways develops load and capacity forecasts
Marketways defines the capacity decision, identifies the drivers of demand and reconciles conditional forecasts across locations and horizons. We quantify uncertainty, test structural change and compare operating and investment options against the same service criterion.
The engagement may combine Sales Forecasting and Demand Planning, Risk and Uncertainty Modelling and Forecasting and Time-Series Analysis. Management receives demand ranges, stated assumptions, capacity scenarios and review triggers tied to the decisions it can make.