Integrating short term variations of the power system into integrated energy system models: A methodological review

TLDR: In this paper, the authors present a review of integrated energy system models and power systems models and identify the strengths, limitations, and applicability of these different methodologies, and the analysis identifies remaining gaps and shortcomings.

Abstract: It is anticipated that the decarbonisation of the entire energy system will require the introduction of large shares of variable renewable electricity generation into the power system. Long term integrated energy systems models are useful in improving our understanding of decarbonisation but they struggle to take account of short term variations in the power system associated with increased variable renewable energy penetration. This can oversimplify the ability of power systems to accommodate variable renewables and result in mistaken signals regarding the levels of flexibility required in power systems. Capturing power system impacts of variability within integrated energy system models is challenging due to temporal and technical simplifying assumptions needed to make such models computationally manageable. This paper addresses a gap in the literature by reviewing prominent methodologies that have been applied to address this challenge and the advantages & limitations of each. The methods include soft linking between integrated energy systems models and power systems models and improving the temporal and technical representation of power systems within integrated energy systems models. Each methodology covered approaches the integration of short term variations and assesses the flexibility of the system differently. The strengths, limitations, and applicability of these different methodologies are analysed. This review allows users of integrated energy systems models to select a methodology (or combination of methodologies) to suit their needs. In addition, the analysis identifies remaining gaps and shortcomings.

Figures

Figure 2 -Flow chart of soft-linking methodology. [11]

Figure 3: Example of a time-slice division used in energy system optimization models [74]

Figure 8: Chronological (left) and duration curves (right). The upper black line represents the load, while the lower grey line represents the residual load that needs to be covered by dispatchable power plants after adding 25% generation from wind and 25% generation from PV. To calculate the duration curves on the right, both load and residual load are reordered from highest to lowest value. The RLDC on the right shows three main challenges arising from including wind and solar: They do not fully contribute to the reduction of peak load, they lead to lower utilization of dispatchable power plants, and they can produce more than load, leading to curtailments.

Figure 9: Left: Representation of RLDCs in REMIND in a discretized form with the help of four load bands. Left: Black line represents the RLDC at 0% VRE; the load band heights of the four load bands are fitted to best represent the RLDC. Right: At a wind share of 40%, the RLDC is decreased (black curve). According to the changing slope of the RLDC, the reduction of load band heights (as shown by the blue arrows) is very different across the different load bands. The height of the base load band is reduced much stronger than the height of the mid and peak load bands.

Figure 1 - Comparison of the level of detail of the model types considered in this analysis

Figure 6: A stylized approach for introducing ancillary markets in ESOMs.

Full text

KULeuven Energy Institute
TME Branch
WP EN2016-09
Integrating short term variations of the power
system into integrated energy system
models: A methodological review
Seán Collins, Paul Deane, Kris Poncelet, Evangelos Panos,
Robert Pietzcker, Erik Delarue and Bryan Ó Gallachóir
TME WORKING PAPER -
Energy and Environment
Last update: April 2017
An electronic version of the paper may be downloaded from the TME website:
http://www.mech.kuleuven.be/tme/research/

1
Integrating short term variations of the power system into integrated
energy system models: A methodological review
Seán Collins
a, b ,c *
, Paul Deane
a,b , c
, Kris Poncelet
d,e , f
, Evangelos Panos
g
, Robert
Pietzcker
h
, Erik Delarue
d,e
, Brian Ó Gallachóir
a,b , c
a) Environmental Research Institute, University College Cork, Cork, Ireland
b) School of Engineering, University College Cork, Cork, Ireland
c) MaREI Centre, Environmental Research Institute, University College Cork, Cork, Ireland
d) Department of Mechanical Engineering, KU Leuven, Leuven, Belgium
e) EnergyVille, Genk, Belgium
f) VITO, Mol, Belgium
g) Paul Scherrer Institut, Villigen, Switzerland
h) Potsdam Institute for Climate Impact Research (PIK), Potsdam, Germany
*Corresponding Author
Contact Information: Telephone: +353 (0)21 490 1959
Fax: + 353 (0)21 4901970
Email: sean.collins@umail.ucc.ie
Address: Environmental Research Institute, Lee Road, Cork, Ireland
Abstract
It is anticipated that the decarbonisation of the entire energy system will require the introduction of
large shares of variable renewable electricity generation into the power system. Long term
integrated energy systems models are useful in improving our understanding of decarbonisation but
they struggle to take account of short term variations in the power system associated with increased
variable renewable energy penetration. This can oversimplify the ability of power systems to
accommodate variable renewables and result in mistaken signals regarding the levels of flexibility
required in power systems. Capturing power system impacts of variability within integrated energy
system models is challenging due to temporal and technical simplifying assumptions needed to make
such models computationally manageable. This paper addresses a gap in the literature by reviewing
prominent methodologies that have been applied to address this challenge and the advantages &
limitations of each. The methods include soft linking between integrated energy systems models and
power systems models and improving the temporal and technical representation of power systems
within integrated energy systems models. Each methodology covered approaches the integration of
short term variations and assesses the flexibility of the system differently. The strengths, limitations,
and applicability of these different methodologies are analysed. This review allows users of
integrated energy systems models to select a methodology (or combination of methodologies) to
suit their needs. In addition, the analysis identifies remaining gaps and shortcomings.
Keywords
Energy system planning; Integration of renewable energy sources; Technical Resolution; Temporal
resolution; Power systems

2
Highlights
 Long term energy system modelling challenges identified for the power sector
 State-of-the-art methodologies for integrating the challenges related to the integration of
variable renewables are presented
 Comparison of methodologies succinctly exposes the strengths & limitations of each
respective methodology
1. Introduction
The transition to a low-carbon energy system is expected to require the electricity sector to
integrate large amounts of variable renewable energy sources (VRES) [1-4]. The instantaneous
electricity generation by VRES is highly intermittent, location specific and only predictable to a
limited extent. A massive penetration of VRES, therefore, has a strong impact on the operation of
the power system [5-9]. Capturing the economic and technical challenges related to a large-scale
penetration of VRES, therefore, requires modelling the variability in system load and renewable
generation, the limited flexibility of thermal units and the spatial smoothing of the variability. This
requires models with a high level of temporal, technical and spatial detail.
Long-term planning models have been applied frequently to analyse scenarios for the evolution of
the energy system over multiple decades. Due to computational restrictions, the level of temporal,
technical and spatial detail in these models is typically low. In contrast, operational power system
models focus on the operations of the power system using a high level of detail but do not consider
its long-term evolution.
Multiple authors have recently analysed the impact of temporal detail [10-16], technical detail [10,
11, 17-20] and spatial detail [21-23] employed in long-term planning models. Depending on the
representation of integration challenges, low levels of detail can either favour or disfavour VRES: For
high penetrations of VRES, If electricity is treated as a homogeneous good or only a low number of
averaged time-slices is used, the low level of detail leads to an overestimation of the value of
baseload technologies and VRES, while the value of flexible generation technologies with higher
generation costs is underestimated [10]. In contrast, if a model uses rather crude representations of
integration challenges such as upper limits on VRES shares or fix backup requirements, the low level
of detail can overly restrict the deployment of VRES compared to more detailed representations
[24]. As a result, the cost of achieving ambitious greenhouse gas emission reduction targets can be
either significantly under- or overestimated.
Moreover, the importance of capturing critical elements of power system operation for planning a
reliable and adequate power system is analysed in [25-29], making clear that a reliable operation of
the power system cannot be guaranteed for the scenarios generated by current long-term planning
models. As such, Pfenninger et al [30] consider ‘resolving time and space’ to be the main challenge
for energy system optimization models. For such long term modelling analyses it is also critical from
an operational perspective to capture the current state of play and development of technologies so
as to ensure a realistic trajectory of future technology development is considered [31-35].
Bridging the gap between highly-detailed operational power system models and long-term planning
models has become an active field of research, in view of the challenge of the transition to a less

3
carbon-intensive energy system. Numerous methodologies to bridge this gap have recently been
developed [10, 24, 30, 36, 37].
This paper presents a review of prominent methodologies developed to better capture the
economic and technical challenges related to the integration of VRES in two families of long-term
planning models, namely long-term energy system optimization models (ESOMs) usually focusing on
country-level (or group of countries, e.g. EU-level) scenarios for the next decades, and Integrated
Assessment Models (IAMs), which focus on global long-term scenarios for the full 21
st
century. The
strengths, limitations, and applicability of these different methodologies described in the literature
are analysed. This analysis allows users of long-term planning models to select a methodology (or
combination of methodologies) to suit their needs. In addition, the analysis exposes the needs for
further research.
The remainder of this paper is organized as follows. First, Section 2 identifies the problem space by
presenting a comprehensive overview of the different types of models and the level of temporal,
technical and spatial detail typically employed in these models. Second, Section 3 presents the
different methodologies developed in the literature for improved capturing of the economic and
technical challenges related to the integration of VRES in planning models. The strengths and
limitations of each approach are discussed in detail. Finally, main conclusions are formulated in
Section 4.
2. Overview of energy modelling tools
This section first presents a brief description of the models considered in this paper, i.e., operational
power system models, energy system optimization models and integrated assessment models.
Subsequently, the level of temporal, technical and spatial detail typically used in each of these
models is discussed.
2. 1. Operational power system models
Operational power system models analyse the operations of a given power system, i.e., investment
decisions are not considered. While there are large differences in the focus and applications of
operational power system models [38], the focus of this work is on unit commitment and economic
dispatch (UCED) models. UCED models determine for every time step within a certain time horizon
which units should be online and how much each unit should be generating in order to minimize the
cost of supplying a given demand for electricity. Detailed technical constraints, such as the minimal
operating level, restricted ramping rates, minimum up and down times, start-up costs and efficiency
losses during part-load operation are accounted for on a unit by unit level. Properly accounting for
the minimal operating level requires tracking the commitment status of individual units. As such,
most current UCED models rely on mixed-integer linear programming (MILP). Due to a large amount
of integer variables, solving UCED models can be computationally challenging. The time horizon of
UCED models is typically restricted to one day up to one year. This time horizon is disaggregated into
different time steps with a resolution in the range of 5 minutes up to one hour. Prominent examples
of UCED models include PLEXOS [39], LUSYM [40], GTMax [41], ORCED [42] and EnergyPLAN [43].
While UCED models allow analysing the operation of the power system in detail, these models do
not allow to consider the (cost-optimal) evolution of the installed generation capacity. Moreover,
the scope of these models is restricted to the power system. Interactions with other energy sectors

4
such as the heating and transport sector are generally modelled by exogenously specifying the
demand for electricity.
2. 2. Long-term energy system optimization models (ESOMs)
ESOMs are used mainly to generate scenarios for the long-term evolution of the energy system. As
such, ESOMS compute the investments and operation of the energy system that result in a partial
equilibrium of the energy system, i.e., ESOMs simultaneously compute the production and
consumption of different commodities (fuels, materials, energy services) and their prices in such a
way that at the computed price, production exactly equals consumption. This equilibrium is referred
to as a partial equilibrium since the scope of ESOMs is restricted to the energy system (comprising
the power sector, transport sector, heating sector, etc.), being merely a part of the overall economic
system. To compute this partial equilibrium, ESOMs rely on the fact that this equilibrium is
established when the total surplus is maximized (or when total cost is minimized in case of an
inflexible demand). Optimization techniques, such as linear programming, are applied to retrieve the
investments, production and consumption patterns as well as trade flows yielding a maximal surplus.
In contrast to some of the IAMs discussed below, partial equilibrium models are bottom-up models,
meaning that each specific sector is composed of multiple explicitly defined technologies which are
interlinked by their input and output commodities. Regarding the geographical scope, ESOMs are
generally applied to countries or regions, but can also be applied on a city level. The time horizon
spanned is generally multiple decades. The main strength of ESOMs is that these models provide a
comprehensive description of possible scenarios for the transition of the energy system by
considering the inter-temporal, inter-regional and inter-sectoral relationships. A limitation of ESOMs
that are applied to only one country is that they ignore the potential benefit of international
cooperation for the integration of VRES via expanded transmission grids. Well-known examples of
ESOMS are MARKAL/TIMES [44], MESSAGE [45] and REMIX [46].
2. 3. Integrated assessment models
IAMs and ESOMs share many characteristics and can consist of the same modelling frameworks
1
.
The main difference is their aim and scope: ESOMs typically focus on near-term energy system
transformations in individual countries or regions, whereas IAMs complement socio-economic
modelling with natural sciences to analyse long-term interdisciplinary questions, typically of a global
scope, such as assessing policies to mitigate climate change [50, 51]. To address these questions,
IAMs need to represent not only the different energy demand sectors such as transport, residential,
and industrial energy use, but also topics like economic growth, resource availability, and land-use-
related emissions. These differences in temporal, spatial and topical coverage imply that IAMs
require higher temporal and geographical aggregation compared to ESOMs in order to keep
computational complexity at a manageable level.
1
The IAMs ETSAP-TIAM and TIAM-UCL use the TIMES modelling framework, while IIASA's MESSAGE IAM
model is built on a MESSAGE modelling framework with additional non-energy sector modules. MESSAGE
modelling framework is distributed by the IAEA for national and regional planning purposes. [47] ETSAP.
http://www.iea-etsap.org/web/applicationGlobal.asp 2016, [48] UCL. https://www.ucl.ac.uk/energy-
models/models/tiam-ucl/#etsap-tiam. 2016, [49] Messner S, Schrattenholzer L. MESSAGE–MACRO: linking an
energy supply model with a macroeconomic module and solving it iteratively. Energy. 2000;25:267-82.

Frequently asked questions

Q1. What have the authors contributed in "Integrating short term variations of the power system into integrated energy system models: a methodological review" ?

This paper addresses a gap in the literature by reviewing prominent methodologies that have been applied to address this challenge and the advantages & limitations of each. This review allows users of integrated energy systems models to select a methodology ( or combination of methodologies ) to suit their needs. 

Q2. What future works have the authors mentioned in the paper "Integrating short term variations of the power system into integrated energy system models: a methodological review" ?

A key motivator in this was to aid future research by presenting and contrasting these methodologies so that, in future, energy system modellers can select and apply methodologies best suited to their situation. Future work is required to effectively compare strengths and weaknesses of the different approaches, this is a key hotspot for future research in this area. These suggestions for future work would also benefit from uni-directional or bi-directional soft-linking which could operationally analyse under high resolution the various power sectors projected and give insights into their operational realisation. Reliable operation of the modelled power system in the short term ( hourly ) is difficult to assess Endogenous determination of the value of flexibility requires to include additional constraints, which further increase computational cost Computational complexity increases with an increasing number of timeslices Integral balancing based on approximating the joint probability distribution of the load and VRES Allows for increased optimality of the solution 

Q3. What is the main advantage of the stochastic programming based methodology?

The stochastic programming based methodology has benefits in that it makes the need for back-up capacity endogenous, allows for the hedging of flexible generation and allows for detailed quantification of uncertainty. 

Q4. What is the main strength of direct integration methodologies for ESOMs and IAMs?

A key strength of direct integration methodologies for ESOMs and IAMs discussed in this work is that are directly integrated into the model optimisation thus eliminating the need for an iterative approach as is required in the bidirectional soft-link approach. 

Q5. What is the main challenge for energy system optimization models?

As such, Pfenninger et al [30] consider ‘resolving time and space’ to be the main challenge for energy system optimization models. 

Q6. Why is the value of storage systems and other sources of flexibility determined?

Due to the fact that chronology is retained within each day, the value of storage systems and other sources of flexibility can be endogenously determined. 

Q7. What is the way to represent the effect of variability in a stylized way?

To offset the low temporal detail and still represent the variability of load and VRES, most IAMs have introduced additional equations and constraints that try to mimic the effect of variability in a stylized way. 

Q8. What is the key benefit of the stylized integration of operational constraints?

The stylized integration of operational constraints has a key benefit in that it allows easy integration of different operational constraints the model that directly increase the optimality of the solution. 

Q9. What is the final limitation of the RLDC approach?

A final limitation is that it requires an assumption on the evolution of the accuracy of the forecasting techniques regarding wind, solar and electricity load profiles. 

Q10. What are the principals that have been identified as guides for addressing flexibility in energy models?

There are certain principals that have been identified as guides for addressing flexibility in energy models such as careful consideration of model simplifications, definition of appropriate temporal and geographic resolution, definition of system flexibility constraints and model validation [36]. 

Q11. What are the main assumptions that are needed to improve the quality of the forecasting techniques?

they are also needed assumptions about the evolution of the quality of the forecasting techniques in the long-term and to the extent that different technologies can contribute to these reserves [119]. 

Q12. What is the difference between ESOMs and IAMs?

These differences in temporal, spatial and topical coverage imply that IAMs require higher temporal and geographical aggregation compared to ESOMs in order to keep computational complexity at a manageable level. 

Q13. What is the main difference between ESOMs and IAMs?

The main difference is their aim and scope: ESOMs typically focus on near-term energy system transformations in individual countries or regions, whereas IAMs complement socio-economic modelling with natural sciences to analyse long-term interdisciplinary questions, typically of a global scope, such as assessing policies to mitigate climate change [50, 51].