Matrix Modeling of Energy Hub with Variable Energy Efficiencies

Wujing Huang, Ning Zhang, Yi Wang, Tomislav Capuder, Igor Kuzle, Chongqing Kang

Abstract

The modeling of multi-energy systems (MES) is a fundamental task in analyzing energy systems integration. Variable energy efficiencies in conversion and storage components introduce nonlinearities that complicate the analysis and optimization of these systems. This paper proposes a standardized matrix modeling approach to automatically model MES with variable energy efficiencies based on the energy hub (EH) framework. By using piecewise linearization, a component with variable efficiency is made equivalent to several parallel components with constant efficiencies. This allows the nonlinear relationships in an EH to be modeled under a linear matrix framework, making the modeling process highly automated. Case studies demonstrate that the proposed model balances accuracy and computational efficiency effectively in optimal EH operation.

Introduction: The Challenge of Variable Efficiency

As we transition toward integrated multi-energy systems (MES)—where electricity, gas, and heating/cooling networks interact—the Energy Hub (EH) has emerged as a powerful conceptual tool for modeling energy conversion and storage. However, a significant limitation in existing EH models is the assumption of constant energy conversion efficiencies.

In reality, efficiencies for components like combined heat and power (CHP) units or boilers vary significantly based on the operating point, a phenomenon known as part-load performance. Ignoring these variations leads to inaccurate optimization results, while modeling them directly creates complex nonlinear programming problems that are difficult to solve reliably. There is a clear need for a method that is standardized, automated, and linearized.

Methodology: Linearizing the Nonlinear

The authors propose a standardized matrix modeling approach based on graph theory, where components are nodes and energy flows are branches. The core innovation involves transforming nonlinear components into a linear framework.

Piecewise Linearization (PWL)

The first step involves approximating nonlinear efficiency curves using multiple linear segments. This transformation effectively replaces a single variable-efficiency component with a set of parallel components, each operating at a constant efficiency within its specific segment.

Piecewise linear approximation of a nonlinear efficiency curve.

The Splitter-Concentrator Structure

To maintain a rigid linear matrix framework, the authors introduce a "splitter-concentrator" abstraction. A nonlinear component is modeled as a three-layer structure: a splitter divides the input flow into segments, linear components process them at constant efficiencies, and a concentrator merges the outputs.

Transformation of a nonlinear component into a linear splitter-concentrator structure.

This "Aha!" moment allows the model to handle the branching logic required by piecewise linearization while staying within a matrix-based linear framework. This enables the use of robust Mixed-Integer Linear Programming (MILP) tools.

Case Study and Key Results

The framework was validated using a simulated hospital site in a Mediterranean area, with data including hourly electricity, heat, and cooling demand, alongside real-world market prices. The researchers compared their proposed matrix model against an exact nonlinear model and a traditional constant-efficiency model:

  • Accuracy: The constant-efficiency model had a significant error of 13.7% in optimal operating costs. The proposed matrix model reduced this error to below 1.0% with only 12 segments, and to 0.06% with 36 segments.
  • Computational Speed: While the exact nonlinear model took nearly 300 seconds to solve, the proposed MILP model solved in less than 1 second for most practical segment counts.
  • Operational Insights: The study found that assuming constant efficiency can lead to entirely "wrong and uneconomical" operation strategies, such as running a CHP unit at inefficient low-load levels during the night.

Conclusion

The proposed matrix modeling approach enables the "plug-and-play" modeling of complex multi-energy systems. By automating the generation of energy flow equations, it removes the need for manual formulation for each new hub layout. While currently focused on steady-state hourly dispatch, future work will extend the framework to consider the dynamics of energy conversion and incorporate stochastic variables like renewable generation.

Summary updated

Abstract

The modeling of multi-energy systems (MES) is the basic task of analyzing energy systems integration. The variable energy efficiencies of the energy conversion and storage components in MES introduce nonlinearity to the model and thus complicate the analysis and optimization of MES. In this paper, we propose a standardized matrix modeling approach to automatically model MES with variable energy efficiencies based on the energy hub (EH) modeling framework. We use piecewise linearization to approximate the variable energy efficiencies; as a result, a component with variable efficiency is equivalent to several parallel components with constant efficiencies. The nonlinear energy conversion and storage relationship in EH can thus be further modeled under a linear modeling framework using matrices. Such matrix modeling approach makes the modeling of an arbitrary EH with nonlinear energy components highly automated by computers. The proposed modeling approach can further facilitate the operation and planning optimization of EH with variable efficiencies. Case studies are presented to show how the nonlinear approximation accuracy and calculation efficiency can be balanced using the proposed model in the optimal operation of EH.
BibTeX
@article{huang2020matrix,
  author = {Wujing Huang and Ning Zhang and Yi Wang and Tomislav Capuder and Igor Kuzle and Chongqing Kang},
  title = {{Matrix Modeling of Energy Hub with Variable Energy Efficiencies}},
  journal = {International Journal of Electrical Power \& Energy Systems},
  year = {2020},
  doi = {10.1016/j.ijepes.2020.105876},
}
Figure