This research, conducted as part of a broader project, focused on modeling and optimizing

energy systems, investigates the integration of renewable energy, industrial flexibility, and highresolution

forecasting. Our study centers on a coupled electricity and district heating system that

includes a large iron plant—an energy-intensive industrial consumer that also produces usable

waste heat. The models used are developed in-house, grounded in real operational data from the

plant and local networks, and implemented within a simulation-based framework.

At the core of this work is the challenge of uncertainty: renewable sources such as wind are

variable and difficult to predict. Similarly, the demand for electricity and heat is highly fluctuating.

Accurate forecasting is vital—not only in terms of predicting average values but also in capturing

the detailed temporal structure of the underlying processes. For example, operational decisions

are highly sensitive to when significant changes (jump times) in supply or demand occur, how

long errors persist (crossing times), and how current values relate to past and future observations

(autocorrelation). Our research emphasizes the development and use of sophisticated forecasting

models that accurately reflect these dynamics.

To operationalize these forecasts, we apply a parametric cost function approximation (CFA)

approach. This method tunes key parameters offline using rolling forecasts and simulation, allowing

real-time operations to proceed via a simple deterministic lookahead. Within this framework, the

iron plant serves a dual role: its electricity demand can be shifted in time to absorb surplus

wind energy, and its waste heat can be used to supply district heating. The combination reduces

curtailment of renewables and lowers heating system costs.

We demonstrate through simulation that this hybrid CFA approach performs well in practice.

It enables robust, forecast-informed decision-making that enhances system efficiency. Although

parameters must be carefully tuned for each context, the method is flexible and computationally

tractable. Depending on one’s disciplinary background, this work may be viewed as either a form

of model predictive control (MPC) or stochastic optimization. It represents a practical, scalable

solution for coordinating renewable energy with industrial heat recovery in complex, uncertain

environments.