Author: Juan Miguel Serrano Rodríguez
Institution: University of Almería
Degree: Computer Science PhD Thesis
Year: 2025

Supervisors:

• Dr. Lidia Roca Sobrino
• Dr. Patricia Palenzuela Ardila

Keywords: CSP, cooling, desalination, solar thermal energy, resource management, optimization

This research encompasses two complementary studies on two intrinsically linked resources: water and energy. The first part focuses on the efficient management of water resources for power generation, while the second explores the efficient use of solar energy for clean water production.

Concentrated Solar Power (CSP) is poised to be a crucial contributor to the energy transition away from fossil fuels. The first phase of this transition is well underway, driven by the massive deployment of low-cost and non-dispatchable renewable technologies such as wind and solar photovoltaics. However, the second and more challenging phase, which involves achieving large-scale dispatchable renewable generation, is still ahead. CSP stands out as a renewable and scalable dispatchable technology with the potential to outcompete combined-cycle and coal-fired power plants.

One of the key challenges in CSP systems lies in cooling the power block, which is typically associated with high water consumption. The first part of this research is therefore dedicated to the efficient management of water resources in CSP plants. An optimal water management strategy is proposed for CSP systems that integrate novel combined cooling configurations. To this end, the annual performance of different cooling alternatives is evaluated for a commercial 50 MWₑ CSP plant, Andasol-II, located in southern Spain. Specifically, three cooling systems, all modelled and validated, are compared: the plant’s existing wet cooling tower, and two variants of a novel combined cooling system with dry cooler capacities of 75% and 100% of the nominal thermal load of the wet cooling tower.

For each alternative, plant operation is optimized under the same water-scarcity scenario using a multi-stage optimization framework. This framework minimizes the cooling cost, encompassing both electricity and water expenses, while ensuring that the cooling demand is satisfied. The key challenge lies in effectively managing the limited water resource. Results show that the integration of the combined cooling system can reduce specific cooling costs by up to 80% and annual water consumption by about 48%, with 38% savings during the driest months. These benefits arise primarily from reduced reliance on costly alternative water sources. The combined cooling alternatives also provide more stable and cost-effective operation throughout the year compared to the wet cooling tower, which is highly sensitive to water availability.

Thermal desalination, particularly multi-effect distillation (MED), can play an important role in mitigating water scarcity. These systems use thermal and electrical energy to separate seawater or contaminated feedwater into fresh water and concentrated brine. Although it may not become the dominant desalination technology, it is well suited for specific applications. Its competitiveness can be improved by leveraging opportunities such as brine mining or industrial wastewater treatment, which enhance both economic feasibility and environmental benefits.

To expand their applicability, MED systems must improve in two directions: (i) by enhancing efficiency through wider operating ranges, or (ii) by adapting to low-temperature applications in which their heat demands can be partially or fully met using low-exergy sources such as waste heat or solar thermal energy. However, the true cost of thermal energy, and consequently the performance of such systems, is often difficult to quantify, particularly because there is no unified set of criteria for evaluating their performance.

To address this challenge, this research proposes a standardized methodology for evaluating the performance of MED processes, which can also be extended to other thermal separation technologies. The method covers key aspects such as instrumentation requirements, process control, and the suitability of performance metrics. It also includes uncertainty quantification and an algorithm for automatic steady-state detection. The proposed approach enhances the reliability and robustness of experimental evaluations under variable conditions. Experimental results confirm that the methodology is both reliable and consistent, enabling fair comparisons of MED systems across different operating scenarios.

The experimental campaign includes evaluations at high top brine temperatures. Results analyzed using several performance metrics and scale formation risks demonstrate that the MED process can operate at high temperatures without significant scaling and can achieve higher concentrations. However, no substantial improvements in thermal performance or reconcentration capacity are observed unless specific design modifications are introduced.

Finally, a novel operational strategy is proposed to enable the seamless, autonomous, and optimal integration of a solar-driven MED system. The method explicitly determines when to start and stop each subsystem while considering a two-day prediction horizon. This allows the optimization to account not only for immediate performance but also for the effect of present decisions on future production. The approach is based on an experimentally validated system model that includes the electrical consumption of each component, combined with the most comprehensive data-driven MED model currently available in the literature. The control architecture follows a hierarchical three-layer structure, in which the upper operational layer solves a mixed-integer nonlinear programming problem aimed at maximizing water production while minimizing operating costs. Results from a week-long system simulation are compared with two alternative strategies, a baseline rule-based operation and a fixed-schedule optimized operation. The proposed method significantly increases system performance by 32% relative to the heuristic baseline and by 21% relative to the fixed-schedule strategy. These gains arise from the method’s ability to fully exploit the temporal flexibility provided by thermal storage while accounting for operational costs.

The best way to interact with the document is via a devcontainer and using the provided workspace. Following the following steps:

1. Clone or download the repository
2. Open it with VSCode
3. Reopen in devcontainer (requires Docker)
4. Open workpsace
5. That's it! Try building the PDF (the build gets automatically triggered after every save)

Most of the writing environment customizations were done following a guide from Paul Wintz

There are a lot of snippets for including attachments in the document, make sure to check them out in latex.code-snippets.

• Compile only the parts you are working on, otherwise the build time will be significant.
• The glossary compilation is very buggy (You can't use a prefix with begin-group character {'.<to be read again> {). So it's better to leave that part commented out in the main.tex: printunsrtglossary. Generate the manuscript PDF, then comment everything else and uncomment the glossary part, and compile again. Join both PDFs where it should be appearing and hopefully hyperlinks will still work (spoiler it does not).
• Important! Validate the CITATION.cff using uvx cffconvert --validate. Otherwise, Zenodo will not be able to parse it and break the integration.
• Check the .zenodo.json file to make sure all the metadata is correct before uploading a new version to Zenodo. This file is used by Zenodo to populate the repository metadata.

The document is compiled using latexmk which uses the configuration file latexmkrc. This file in turn uses lualatex as the interpreter. Output build files including the manuscript PDF are saved in the out_dir folder.

This build process is defined as a "recipe" in LaTeX workshop, and as the default in the devcontainer.

When ordering printed copies of the thesis, it is often more economical to print black & white (B&W) pages at a lower cost than color pages. However, many pages that appear to be B&W may still contain small amounts of color information, leading to higher printing costs. Two utility scripts are provided to analyze and optimize PDF files for printing:

The bash script count_and_enumerate_bw_pages.sh quickly analyzes a PDF to identify which pages are black & white vs. color:

bash scripts/count_and_enumerate_bw_pages.sh out_dir/main.pdf [threshold]

Parameters:

• out_dir/main.pdf: Path to the PDF file to analyze
• threshold (optional): Color threshold for C+M+Y ink coverage (default: 0.005)

Example output:

B&W pages: 95
  Pages: 2,5,6,8,9,10,11,12,...

Color pages: 153
  Pages: 1,3,4,7,25,27,28,...

Threshold (C+M+Y): 0.005

This script uses Ghostscript's inkcov device to measure ink coverage and classify pages.

The Python script count_and_generate_modified_bw_pdf.py analyzes a PDF and creates a new version with B&W pages converted to grayscale while keeping color pages in color:

uv run scripts/count_and_generate_modified_bw_pdf.py out_dir/main.pdf [threshold]

Parameters:

• out_dir/main.pdf: Path to the PDF file to process
• threshold (optional): Color threshold for C+M+Y ink coverage (default: 0.005)

Output:

• Creates out_dir/main_bw.pdf in the same directory as the input file
• B&W pages are converted to grayscale (reduces file size and ensures true B&W printing)
• Color pages remain in color

Example output:

Threshold (C+M+Y): 0.005
B&W pages (95): [2, 5, 6, 8, 9, ...]
Color pages (153): [1, 3, 4, 7, 25, ...]

Extracting pages...
Merging pages...

Output written to: out_dir/main_bw.pdf

This is particularly useful for reducing printing costs when ordering printed copies, as B&W pages will be truly grayscale and charged at B&W rates.

The Python script combine_covers.py combines front cover, back cover, and spine PDFs into complete book covers ready for printing. The script creates a single-page PDF with all components laid out horizontally: back-cover | spine | front-cover.

uv run scripts/combine_covers.py assets/full_covers/front-cover.pdf assets/full_covers/back-cover.pdf assets/full_covers/spines/

Parameters:

• First argument: Path to the front cover PDF
• Second argument: Path to the back cover PDF
• Third argument: Path to a spine PDF file or folder containing multiple spine PDFs
• -o, --output (optional): Output directory or file path (defaults to back cover's directory)

Behavior:

• Creates a single-page PDF with components arranged horizontally (back | spine | front)
• Automatically scales spine height if it differs from the cover height (with warning)
• Raises an error if front and back covers have different dimensions
• When a folder of spines is provided, generates a separate combined PDF for each spine variation
• Output files are named combined_cover_<spine_name>.pdf

Example output:

Warning: Spine height (793.00 pts) differs from covers (792.00 pts).
         Scaling spine by factor 0.9987
Created: assets/full_covers/combined_cover_spine_g520.pdf
  Dimensions: 1317.88 x 792.00 pts
  Layout: Back (612.00) | Spine (93.88) | Front (612.00)
...
Successfully created 8 combined cover(s)

Additional examples:

# Generate cover with a specific spine
uv run scripts/combine_covers.py assets/full_covers/front-cover.pdf assets/full_covers/back-cover.pdf assets/full_covers/spines/spine_g520.pdf

# Specify custom output directory
uv run scripts/combine_covers.py assets/full_covers/front-cover.pdf assets/full_covers/back-cover.pdf assets/full_covers/spines/ -o output_covers/

# Show help
uv run scripts/combine_covers.py --help

This thesis is done in LaTeX using the kaobook class with some modifications inspired by... In particular:

(Fill here all the changes and which files have been changed)

• TODO Changed colors for paragraphs in the margins (margin@par) to a lighter gray (kao.sty)
• TODO Changed font to Fira Sans, as used in TheoWinterhalter-PhD (main.tex)
• Changed side citations to a dark gray, as used in TheoWinterhalter-PhD (main.tex)

The document supports two output variants to easily switch between digital and printed (B&W friendly) versions by changing a single boolean variable in main.tex:

% Set to true for printed black & white version (darker colors for better contrast)
% Set to false for digital color version (lighter colors, green links)
\newif\ifprintedversion
\printedversionfalse  % Change to \printedversiontrue for printed version

Digital version (\printedversionfalse): Uses lighter grays and OliveGreen color for links and accents.
Printed version (\printedversiontrue): Uses darker grays (black!50!Gray) for better contrast in B&W printing.

The following elements are automatically adjusted based on this setting:

• Hyperlinks (ToC, cross-references, citations, URLs)
• Margin citations (bibliography references in margins)
• Nomenclature group titles
• Sidenotes and margin captions

Implementation details can be found in macros.tex where colors are defined conditionally and applied throughout the document.

Some are listed here, check the kaobox class repository for more.

\begin{definition}
\labdef{openset}
Let $(X, d)$ be a metric space. A subset $U \subset X$ is an open set 
if, for any $x \in U$ there exists $r > 0$ such that $B(x, r) \subset 
U$. We call the topology associated to d the set $\tau\textsubscript{d}$ 
of all the open subsets of $(X, d).$
\end{definition}

\begin{remark}
Let $(X, d)$ be a metric space. A subset $U \subset X$ is an open set 
if, for any $x \in U$ there exists $r > 0$ such that $B(x, r) \subset 
U$. We call the topology associated to d the set $\tau\textsubscript{d}$ 
of all the open subsets of $(X, d).$
\end{remark}

\begin{marginfigure}[-5.5cm]
    \includegraphics[]{figures/shareholders-driven-apocalysis.png}
    \caption{``Yes, the planet got destroyed. But for a beautiful moment in time, we created a lot of value for shareholders''}
    % \labfig{}
\end{marginfigure}

Useful to indicate to yourself or your supervisors that the section is not completed.

\wipbox{Additional text to add}

\begin{kaobox}[title=Kaobox with custom color,colback=MediumPurple2!25!white,colbacktitle=MediumPurple2!25!white]
    What is thiiiis
    \begin{enumerate}
      \item Enumerations
      \item should
      \item be
      \item okay
    \end{enumerate}

    Even fancy figures\texttrademark{}. \\

    \begin{minipage}{0.48\linewidth}
    \centering
    \includegraphics[width=\linewidth]{figures/cc-intro-hybrid-cooler-diagram-no-bg.png}\\
    \small Hybrid cooler
    \end{minipage}
    \hfill
    \begin{minipage}{0.48\linewidth}
    \centering
    \includegraphics[width=\linewidth]{figures/cc-intro-combined-cooler-diagram-no-bg.png}\\
    \small Combined cooler
    \end{minipage}
\end{kaobox}

Just a command for kaobox with some custom color

\annotation{Process \textit{vs} waste heat take on efficiency}{
    In a process heat driven system, between two plants that produce the same
    amount of useful product, the most efficient one is the one that uses the
    least external heat to do so, whereas in a waste heat driven system, the
    two plants would be considered as efficient since the unused heat would be
    \textit{wasted} to the environment. A more intuitive definition would be:
    \\
    Given two plants that consume the same waste heat, the most efficient one
    is the one that produces more product with the available heat. 
}

\marginannotation[*-5]{Title of the annotation}{
	Is everything in this life just a wrapped kaobox? What if I told you that a
	kaobox is just a wrapper of tcolorbox?
}

They are defined in kao.sty and not only are customized boxes, but also they have their own counter like figures or tables which can be used to reference them and create a list of at the beginning of the manuscript. Support for referencing them must be added in kaorefs.sty

\begin{modelcounter}{Test}
	\begin{align*}
    T_{cc,out},\,C_{e}&,\,C_{w},\,T_{c,out} = \mathrm{combined\:cooler\:model}(q_{c}, R_{p}, R_{s}, \omega_{dc}, \omega_{wct},T_{amb},HR_i,T_{v},\dot{m}_{v}) \\
    & T_{cc,in}=T_{c,out} \\
    & T_{dc,in}=T_{cc,in} \\
    % three-way valves
    & q_{dc} = q_{c} \cdot (1-R_{p}) \\
    & q_{wct,p} = q_{c} \cdot R_{p} \\
    & q_{wct,s} = q_{dc} \cdot R_{s} \\
    % dc model
    & T_{dc,out},\,C_{e,dc} = \mathrm{dc\:model}(q_{dc},\, \omega_{\text{dc}},\, T_{\text{amb}},\, T_{dc,in}) \\
    % first mixer
    & q_{wct},\,T_{wct,in} = \mathrm{mixer\:model}(q_{wct,p},\,T_{T{cc,in}},\, q_{wct,s},\, T_{dc,out}) \\
    \end{align*}
	\labmod{test}
\end{modelcounter}

As can be seen in \refmod{test}, the counter is working.
As can be seen in \refprob{test}, the counter is working.

\begin{problemcounter}{Test}
	\blindtext
  \begin{equation*}
	\min_{\mathbf{x},\, \mathbf{e};\, \boldsymbol{\theta}} \quad J = f(\mathbf{x}, \mathbf{e}; \boldsymbol{\theta}) = f(x)
  \end{equation*}

  \textbf{with}:
  \begin{itemize}
	\item Model name model
	\[
	  out_1,\, out_2 = f(in_1,\, in_2,\, \ldots,\, in_N)
	\]
	\item Decision variables
	\[
	  \mathbf{x} = [x_1,\, x_2]
	\]
	\item Environment variables
	\[
	  \mathbf{e} = [e_1,\, e_2,\, \ldots,\, e_3]
	\]
	\item Fixed parameters
	\[
	  \boldsymbol{\theta} = [\theta_1 = X,\, \theta_2 = Y]
	\]
  \end{itemize}

  \textbf{subject to}:
  \begin{itemize}
	\item Box-bounds
	\begin{itemize}
	  \item \( x_{1} \in [\underline{x}_{1}, \overline{x}_{1}] \)
	  \item \( x_{2} \in [\underline{x}_{2}, \overline{x}_{2}] \)
	\end{itemize}
	\item Constraints
	\begin{itemize}
	  \item \( \left| out_X - out_Y \right| \leq \epsilon_1 \)
	  \item \( out_X \leq out_Z - \Delta Z \)
	\end{itemize}
  \end{itemize}
  \labprob{test}
\end{problemcounter}

\reminder{Optimization problem definition}{
    The general optimization function is defined as:\footnote{See \nrefsec{intro:optimization}}
    \begin{equation*}
    \min_{\mathbf{x},\, \mathbf{e};\, \boldsymbol{\theta}} \quad J = f(\mathbf{x}, \mathbf{e}; \boldsymbol{\theta}) 
        \quad \text{s.t.} \quad g_i(\mathbf{x}) \leq 0, \quad i = 1, \ldots, m
    \end{equation*}

    where \(x\) is the decision vector, \(e\) represents the environment, and \(\theta\) contains the fixed parameters.
}

The problem is designed as an optimization problem with a shrinking
horizon\marginreminder[*-2]{Shrinking horizon optimization}{
An optimization where the horizon end is fixed, and as time progresses, the
start of the horizon moves forward.\footnote{See
\nrefch{intro:optimization}}
}. The horizon size should be large enough so that
decisions on how to operate the system are made with perspective...

• Add \listofmodels
• Add \listofproblems
• Glosary is not being printed

Si finalmente se hace un capíutlo de instalaciones experimentales, se puede hacer una introducción donde se ponga esta imagen con marcas:

Indicando el campo solar, el circuito de intercambio, generación de vapor, refrigeración combinada, etc. y la MED

This work is licensed under a
Creative Commons Attribution 3.0 IGO License (CC BY 3.0 IGO).

You are free to share and adapt the material for any purpose, including commercially,
provided appropriate credit is given.

See the full license text in LICENSE.md or at
https://creativecommons.org/licenses/by/3.0/igo/