more corrections

docs/3-osier/32-genetic-algorithms.tex

@@ -51,8 +51,8 @@ \subsection{Specific \Aclp{ga}} The variety of \acp{ga} comes from different
 types of operators being applied to the selection, crossover, and mutation
 steps. Section \ref{section:moo-in-energy} showed that \ac{nsga2} is a popular
 genetic algorithm choice. However, this algorithm performs poorly with greater
-than three objectives \cite{deb_fast_2002, seada_unified_2016}. In this thesis,
-I use a more modern algorithm, \ac{unsga3}. \ac{unsga3} builds on its
+than three objectives \cite{deb_fast_2002, seada_unified_2016}. This thesis
+uses a more modern algorithm, \ac{unsga3}. \ac{unsga3} builds on its
 predecessors \ac{nsga2} and \ac{nsga3} by unifying efficient solutions of mono-,
 multi-, and many-objective problems in a single algorithm.
 
@@ -76,7 +76,8 @@ \subsection{Specific \Aclp{ga}} The variety of \acp{ga} comes from different
 blank_generating_2021}. In this thesis, I use the Riesz s- Energy method
 described by Blank et al. to calculate these points for a problem with an
 arbitrary number of objectives \cite{blank_generating_2021}. Figure
-\ref{fig:ref-dirs} illustrates a set of initialized reference directions.
+\ref{fig:ref-dirs} illustrates a set of initialized reference directions used
+in such methods.
 
 \begin{figure}[h]
   \centering
@@ -90,6 +91,7 @@ \subsection{Specific \Aclp{ga}} The variety of \acp{ga} comes from different
 objectives by introducing the binary tournament from \ac{nsga2} and reducing to
 the most efficient algorithm for the problem at hand \cite{seada_unified_2016}.
 Chapter \ref{chapter:benchmark-results} demonstrates these three algorithms in 
+Section \ref{section:4-exercise-1}.
 
 \subsection{Hyperparameter Tuning}
 Similar to other machine learning models, \acp{ga} have several hyperparameters
@@ -99,9 +101,9 @@ \subsection{Hyperparameter Tuning}
 hyperparameters is often performed using either a grid search or random sampling
 \cite{bergstra_random_2012}. This thesis adopts the approach from Blank and Deb
 \cite{blank_pymoo_2020} using a genetic algorithm to identify the ideal
-hyperparameters. A problem is converted into a single objective problem using
-the desired algorithm, then a second genetic algorithm drives the problem where
-the decision variables are hyperparameters of the desired algorithm.
+hyperparameters. A multi-objective problem is converted into a single objective problem 
+by choosing one objective and optimized with the target \ac{ga}, then a second \ac{ga} 
+drives a new problem where the decision variables are hyperparameters of the desired algorithm.
 
 \subsection{Convergence}
 There are several ways to stop a simulation in \ac{pymoo}. A simulation may end
@@ -135,7 +137,10 @@ \subsection{\acs{pymoo} and \acs{deap}}
 tools and hyperparameter tuning. \ac{deap} is another open-source library
 offering a toolkit for constructing \acp{ga} and therefore has fewer prepackaged
 algorithms than \ac{pymoo}. There are robust reasons to use both libraries, so
-\ac{osier} facilitates both.
+\ac{osier} facilitates both. \ac{pymoo} provides a broader range of tools 
+while \ac{deap} enables greater flexibility with \ac{ga} implementation and more
+straightforward procedures for stopping a simulation and restarting at a later
+time.
 
 
 

docs/3-osier/33-dispatch-model.tex

@@ -2,13 +2,13 @@ \section{Dispatch Model}
 \ac{osier} offers two models for dispatching energy. The first is a merit order
 dispatch model that optimally dispatches electricity according to the marginal
 cost of generation with perfect foresight. This model is formulated as a
-\acf{lp} and written in \ac{pyomo}. Similarly, the second dispatch model follows
+linear program and written in \ac{pyomo}. Similarly, the second dispatch model follows
 merit-order but electricity is dispatched according to a hierarchy of rules and
-is myopic. This approach facilitates parallelization and reduces the overhead
-cost of problem setup. However, dispatch solutions may be sub-optimal compared 
-to the \ac{lp} formulation.
+is myopic, unaware of future timesteps. This approach facilitates parallelization 
+and reduces the overhead cost of problem setup. However, dispatch solutions may be 
+sub-optimal compared to the \ac{lp} formulation.
 
-\subsection{Economic/Merit-order dispatch}
+\subsection{Optimal dispatch}
 \label{section:merit_order}
 
  The economic dispatch model minimizes the generation cost subject to physical
@@ -59,7 +59,9 @@ \subsection{Economic/Merit-order dispatch}
 charging and discharging from storage technologies. I used this approach because
 constraining this behavior requires a binary variable that makes the problem
 non-convex and therefore requires a more sophisticated solver. A small but
-sufficiently large $\pi$ will always nullify the penalty term. This dispatch
+sufficiently large $\pi$ will always nullify the penalty term. That is,
+the model will never prefer simultaneous charging and discharging over avoiding 
+even a slight penalty. This dispatch
 model reflects the minimum physical constraints for an energy system without
 considering fine-scale operational details such as frequency control. Equation
 \ref{eq:dispatch-objective} assumes that the retail cost for generating
@@ -75,7 +77,8 @@ \subsection{Hierarchical dispatch}
 electricity demand time series. A second loop runs through the sorted list of 
 technology objects and calculates the power output for each technology. There
 are three broad types of technologies in \ac{osier}: Standard technologies,
-ramping technologies, and storage technologies. The power output for a standard
+ramping technologies (e.g.,  conventional coal and nuclear plants), and storage 
+technologies (e.g., fly-wheels, \acp{lib}). The power output for a standard 
 technology is given by
 
 \begin{align}

docs/acros.tex

@@ -57,4 +57,6 @@
 \acro{mca}[MCA]{municipal choice aggregation}
 \acro{imea}[IMEA]{Illinois Municipal Electric Agency}
 \acro{ppa}[PPA]{power purchase agreement}
-\acro{pypsa}[\texttt{PyPSA}]{Python for Power Systems Analysis}
+\acro{pypsa}[\texttt{PyPSA}]{Python for Power Systems Analysis}
+\acro{lib}[LiB]{lithium-ion battery}
+\acrodefplural{lib}[LiBs]{lithium-ion batteries}