The decentralisation of energy supply, largely driven by renewable energy sources, has led to increased volatility and imbalances in the power grid. To maintain grid stability, flexible balancing mechanisms are required. Ancillary services offer this flexibility by either injecting power (upward regulation) or absorbing it (downward regulation). Among these services, the Automatic Frequency Restoration Reserve (aFRR) plays a crucial role due to its large regulating capacity and direct impact on imbalance prices through its activation costs. As market volatility increases, short-term trading in intra-day and imbalance markets becomes more important, increasing the need for accurate forecasting. While day-ahead and intra-day market forecasting is well-established in Electricity Price Forecasting (EPF), there is limited research on imbalance markets, especially the Dutch aFRR market, highlighting the novelty of this study. The Dutch market uses a dual pricing system with separate bid ladders for upward and downward regulation.

This research focuses on forecasting aFRR bid ladders in the Dutch electricity market and explores their application in short-term trading strategies. Forecasts are generated three hours ahead of delivery, aligning with decision points for intra-day trading, aFRR participation, or opting for no action. The study includes a detailed market analysis, a review of forecasting methods, the development of a machine learning framework tailored to the aFRR market, evaluation of forecast performance, and the practical use of these forecasts in trading scenarios.

A structured machine learning pipeline is designed, encompassing data pre-processing, transformation, model selection, prediction, and evaluation. During the transformation phase, data is scaled and then reduced in dimensionality using Principal Component Analysis (PCA) to retain key variance. The resulting components are used as inputs for predictive models, including LASSO, XGBoost, and LSTM, which are benchmarked against preliminary bid ladders published three hours before delivery. Forecast accuracy is evaluated using point metrics (sMAPE), interval metrics (PICP and PINAW), and a novel self-developed metric called the Largest Knick Volume (LKV), which captures accuracy at key inflection points in the bid ladders that are most relevant for short-term trading.

The findings are twofold. First, model evaluations both at the PCA level and on the reconstructed bid ladders indicate that all models can track general market trends but do not outperform the benchmark. Diebold-Mariano tests confirm that benchmark performance is superior at the PCA level. After reconstructing the bid ladders, benchmark sMAPE scores are 7% for upward and 8% for downward regulation, outperforming LASSO (7%/10%), XGBoost (8%/11%), and LSTM (7%/11%). Second, the forecasts are integrated into Battery Energy Storage System (BESS) intra-day trading strategies, including one based solely on intra-day prices and three incorporating different aFRR bid ladder positions: gas turbine marginal cost, intra-day price with a premium, and LKV-based positioning. Integrating aFRR forecasts improves trading performance, with some high volume-price strategies boosting revenue by up to 18%.

Although the developed forecasting methods do not outperform the benchmark, the benchmark itself already captures much of the relevant market information available three hours ahead, limiting the scope for additional forecasting improvements. This limitation is attributed to the weak correlation between market fundamentals and bid outcomes, as well as the strong influence of individual actors in the relatively small Dutch aFRR market. Nonetheless, the study shows that using existing market data and participating across multiple markets can improve profitability. The results support the strategic value of data-driven, forecast-informed bidding for enhancing the economic performance of battery storage systems.

As the number of Electric Vehicles (EVs) and renewable energy sources (RES) increases rapidly, power grids struggle to adapt. In the coming years, power system flexibility is urgently required to use the limited capacity of the existing infrastructure efficiently. Over the long term, flexibility will remain essential to account for the variable and uncertain electricity production from RES. EVs have large batteries that are often only partly used for daily travel, particularly in densely populated areas. Smart charging and Vehicle-to-Grid (V2G) can harness the flexibility of EVs to support grid balancing and congestion management.

This thesis investigates the smart charging and V2G potential for EV aggregators, with a focus on workplace charging. State-of-the-art Reinforcement Learning (RL) techniques are applied to a case study involving a business parking lot. The objective is to maximize the profits of the EV aggregator while satisfying EV user and transformer power limit constraints. The modeled EV behavior is based on data from real EV measurements in the Netherlands. The real-time charging optimization problem is characterized by high uncertainty. RL is widely considered a promising algorithm for solving highly uncertain problems. However, the latest Deep RL algorithms often struggle to guarantee constraint-satisfying behavior. Safe RL, an emerging subfield, aims to reduce constraint violations in the learned behavior, thus making algorithms ‘safer’.

This thesis applies recent Safe RL algorithms and compares their performance to Deep RL baselines and conventional As-Fast-As-Possible (AFAP) charging. The proposed method, Constrained Variational Policy Optimization (CVPO), achieved performance comparable to that of the optimal offline Gurobi solver in simulation scenarios where sufficient transformer capacity was available and overloads could not occur. The learned behavior generalized well to unseen levels of charger occupation. However, in scenarios with more inflexible loads and a smaller transformer power limit, transformer overloading risk made the problem more constrained, resulting in a decline in CVPO’s performance.

This thesis explores how accurate state estimation can support Distribution System Operators (DSO) in managing MV grids. As the energy transition drives increasing complexity in electricity networks, more precise and reliable data is necessary for planning and operational decisions. Although state estimation offers a way to enhance data, DSOs struggle to implement it. The main objective of this thesis is to develop a general strategy for placing measurements to achieve accurate state estimation using as few measurements as possible.

Multiple cases were designed to assess the impact of specific measurement placement. A reference ”perfect case” network load scenario was created, and the values were corrupted using normally distributed noise, with standard deviations reflecting the expected accuracy of each measurement type. These noisy measurements were fed into the PandaPower Weighted Least Squares (WLS) state estimation algorithm, implemented in Python. The resulting state estimates were compared to the perfect case values to evaluate the state estimator’s accuracy and the placement strategy’s effect.

The results show that Power Injection Measurements (PIMs) primarily improve accuracy at the node where they are placed. In contrast, Medium Voltage Measurement Units (MVMUs) offer broader improvements across the entire feeder where they are installed. One strategic measurement location was identified based on financial grounds. Results also indicate that improving the accuracy of an inaccurate node is possible without improving its measurement, but requires widespread deployment elsewhere, which is rarely justifiable economically. Retrofitting stations solely for measurement purposes is generally not considered worthwhile.

The main focus should be on identifying the substations with the poorest measurement accuracy, typically pseudo-measurements. As a consequence, overall pseudo-measurement accuracy will also improve. This makes it more likely that the state estimates will fall within predetermined limits. The number of measurements that need to be placed depends on this predetermined limit.

Decarbonisation and efforts to reduce living expenses are driving interest in multicarrier energy systems (MCES) that integrate electricity, heat and e-mobility. These integrated networks require sophisticated energy management strategies to address uncertainties in energy supply, demand, and weather conditions.

This thesis seeks to develop a reinforcement learning (RL)-based energy management system (EMS) for a multicarrier residential building at TU Delft's Green Village. The case study household integrates photovoltaic and solar thermal systems, an electric vehicle (EV), lithium-ion battery, heat pump, and thermal storage. Current management uses white-box model predictive control (MPC) which, while effective, demands significant computational resources. In addition, its implementation requires expertise in optimal control and physics-based modelling.

The developed RL-based EMS provides a computationally efficient alternative, leveraging data-driven methods to learn system dynamics. The RL approach was benchmarked against the existing Expert, a day-ahead MPC planner. Performance was evaluated in terms of operational safety, grid energy exchange costs, and satisfaction of EV state of charge (SoC) demands.

The RL agent achieved performance comparable to that of the Expert in managing the MCES, while improving accessibility for developers lacking control theory expertise. The RL agent displayed consistent and near-optimal performance, resulting in only a 4% increase in grid exchange costs while improving both EV charging compliance and safety constraint adherence.

A literature review of RL applications for residential EMS is presented alongside an investigation of advanced policy update algorithms, deep neural network architectures, temporal feature engineering, and reward shaping strategies to analyse their impact on EMS performance.

The implemented RL-based control solution has been shown to manage electrical and thermal subsystems while maintaining safety and minimising costs. The computational efficiency and reduced modelling requirements of an RL agent highlight its suitability for small-scale MCES applications, such as residential or office buildings, where MPC may not be practical.

The integration of distributed energy resources (DERs) and the increasing penetration of renewable energy generation have significantly increased the complexity and uncertainty of modern distribution networks. These developments necessitate advanced dis
patch algorithms capable of handling the variability and operational constraints inherent in such systems. This thesis focuses on developing model-free deep reinforcement learning (DRL) algorithms to ensure reliable, safe, cost-effective operation in distribution networks (DNs). The research questions addressed in this thesis explore various challenges associated with the enforcement of operational constraints, learning efficiency, and computational cost reduction in DRL-based optimal operation of DNs.....

With the energy transition already underway in the Netherlands, one of the largest obstacles currently facing this process is the widespread grid congestion.
To deal with this grid congestion, many different flexibility solutions have been proposed. One of these solutions is the use of non-firm grid connections, which dynamically allocate grid capacity to connected parties based on the free space available on their specific part of the grid.
A key design parameter of this flexibility options is the manner in which the available capacity is allocated across the non-firm connections.
In this work, a comparison is made between five methodologies for allocating capacity among non-firm grid connections: a contract age based methodology (and a variation upon it which also takes into account the location of a connection in the grid), a rotating priority list, a pure mathematical optimisation of total allocated capacity as well as a proportional allocation methodology which tries to allocate equally between loads.
To compare these different methodologies, a set of metrics is devised, which attempt to capture the three key dimensions of relevance in this space, each representing a stakeholder: the customer/connection side, the grid operator side and the societal/wider grid user side.
It is found from this comparison across these metrics that each of the methodologies excels at different metrics, making each of them viable in their own respect.

In the current grid infrastructure, the power flow reaching the limits of the grid capacity induces several problems. A way to circumvent this issue is the introduction of flexibility in order to alleviate the impact on the energy network. However, the estimation of the required flexibility remains a problem in the current state of the art. In this thesis, an approach is proposed to estimate the flexibility of distribution systems by identifying assets based on aggregated transformer measurements. Flexibility can be seen as the ability of the system to adapt to changes coming from demand and generation. Different assets can be available such as solar panels, and heat pumps and have their specificities. The problem thus translates into the identification process in aggregated data. Several steps are important in the process, such as the consumption baseline, the study of the different characteristics of the profiles, and the number of iterations. To allow the accuracy to reach high levels and to match the standards of distribution system operators (DSO), the process requires the minimization of squared error between the data and the chosen profiles. Different cases are investigated and validated prior to the study of the results. For the identification of heat pumps and solar panels, individual approaches have been investigated in the literature. However, the combined identification of assets has not yet been studied. A mixed identification process is proposed in this work, and the method has been tested, reaching a high level of accuracy. Within this new framework aiming at separating the assets, options are proposed to enhance the accuracy of the process.

Energy arbitrage is a key activity in privately-owned energy storage system portfolios. However, the profitability of battery energy storage systems in this domain is limited by two primary challenges: the high cost of battery manufacturing and the insufficient temporal volatility of electricity prices within individual bidding area.

In this context, this study explores the economic viability of relocating the utility-scale battery energy storage systems. To enhance profitability, the proposed portable energy storage system integrates a semi-empirical battery degradation model to accurately account for degradation costs, while its mobility
allows for travel between different bidding zones to exploit inter-zonal electricity price differences.

A mixed-integer linear programming model is built to optimize the system’s daily charging/discharging and transportation scheduling. The system’s daily operations within the Norwegian day-ahead market are simulated using a Python model.

The investigation includes comparative analyses to assess how battery location, chemistry, and degradation models influence system profitability. The study compares stationary and portable battery energy storage systems: in the stationary case, it evaluates profitability with different battery chemistries (lithium ferrophosphate and lithium manganese) in different locations, while the portable case explores the impact of different battery chemistries and degradation models (semi-empirical vs. conventional energy throughput).

The findings reveal that none of the stationary energy storage systems can achieve break-even through conventional temporal arbitrage in the Norwegian day-ahead market. In contrast, relocating the battery proves profitable for all four portable energy storage systems.

Additionally, the application of the depth-of-discharge-aware semi-empirical model enhances profitability of portable systems with both battery chemistries, with the lithium manganese system showing a more than 15% increase in investment return rate. This makes lithium manganese a more attractive option than the lithium ferrophosphate battery due to its lower investment cost, shorter payback period, and higher return on investment.

The transportation sector continues decarbonizing with the increasing number of Electric Vehicles (EVs) replacing gasoline and diesel cars every year. However, the integration of vast amounts of EVs introduces complexities in energy distribution and grid stability. Charge Point Operators (CPOs), positioned at the intersection of EVs and the grid, play a critical role in managing these complexities. They ensure that the charging infrastructure meets the needs of both EV users and the grid, highlighting the importance of smart charging strategies.

In this thesis, a smart charging approach is proposed from the point of view of a CPO. The proposed approach aims to optimize the charging schedules for EVs parked at a commercial building's parking lot. The objective of the optimization problem is to minimize the Power Setpoint Tracking (PST) error, which indicates the error between the contracted energy in the day-ahead market by the CPO and the aggregated consumption of charging stations the next day. This optimization involves complex sequential decision-making, where the uncertain nature of EV arrivals and departures demands a fast and adaptive solution. Thus, this thesis proposes a Markov Decision Process (MDP) formulation and solves it using the Deep Deterministic Policy Gradient (DDPG) algorithm to minimize the PST error by scheduling the charging of EVs. DDPG is chosen for its ability to efficiently handle complex problems with continuous state and action spaces, making it ideal, considering the uncertainties inherent to the arrival of EVs and the charging process. Additionally, DDPG's application in a commercial building's parking lot, where EV arrival and departure patterns are usually consistent, further solidifies DDPG as a strong alternative.

Evaluating the proposed DDPG approach with alternative benchmarks, such as the uncontrolled "charge as fast as possible" (CAFAP) and the optimal solution obtained through a Mixed Integer Non-Linear Programming (MINLP) formulation, signifies DDPG's superior performance in several metrics. Specifically, it outperforms the CAFAP algorithm by achieving a reduction in PST error by an average of 34% for a parking lot with 10 chargers over 12 hours of charging for a day. This highlights DDPG's efficacy in optimizing EV charging schedules over the CAFAP algorithm. Moreover, DDPG's model benefits from the ability to be trained offline with historical data and deployed online once trained. This approach allows for rapid, dynamic rescheduling of charging in real-world operations, offering speed advantages over the theoretically optimal solution, which requires prior knowledge of arrival and departure times and State of Charge (SoC) of EVs. All experiments validating these findings were conducted within the EV2Gym, a Gym environment specifically designed to simulate the EV charging scenarios.

Lastly, this thesis contributes to the field by demonstrating how RL, through the use of DDPG, can optimize PST for EV charging in a commercial building's parking lot. By offering a detailed comparison with other algorithms and showcasing the scalability and adaptability of DDPG, the research provides valuable insights for CPOs and stakeholders in the energy sector.

In an era defined by the rapid evolution of energy technologies and the necessity of sustainable power generation, effective voltage control is a critical component in optimizing the performance of the power distribution grid and the seamless integration of renewable energy sources. This thesis presents the development and evaluation of an adaptive voltage control policy for high voltage to medium voltage transformers aimed at mitigating voltage limit violations in modern power distribution networks.

The motivation for this research arises from the increasing integration of distributed energy resources and the growing complexity of power grids, which require dynamic and robust voltage management strategies. The proposed policy utilizes a voltage control curve, optimized using a genetic algorithm, which systematically identifies the best parameters for the curve to ensure voltage stability and minimize violations across diverse grid configurations.

The methodology involves quasi-dynamic simulations performed on 30 real-world distribution grids, which test the effectiveness of the voltage control curves in maintaining grid stability under varying load and generation conditions. This was done by setting a stricter ±3% voltage limit for the medium voltage grid. Results show an average reduction of 88.54% in voltage limit violations and a 17.19% decrease in the maximum difference between the maximum and minimum voltage levels in the medium voltage grid throughout a complete year (2023). This study highlights the importance of a grid-specific voltage control curve, as every distribution grid exhibited unique voltage regulation needs.

Furthermore, the policy was evaluated under projected grid scenarios, demonstrating sustained effectiveness across two future timeframes: one set five years and the other ten years into the future. In the first scenario (2028), an 80.06% reduction in voltage limit violations was achieved, while the second scenario (2033) still showed a reduction of 62.25%.

Through the genetic algorithm optimization, the voltage control policy adapts to fluctuating grid dynamics, contributing to improved power quality and the successful integration of distributed energy resources. While effective in medium voltage grids, further research is needed to explore the policy’s applicability to low voltage networks and in environments with limited data availability.

The simplicity and universality of the proposed adaptive voltage control strategy make it a practical solution for real-world deployment, as it can be implemented on existing hardware. In contrast, the proposed strategy alone will not be enough to completely eliminate all the voltage stability problems in distribution grids and additional measures are necessary. However, by minimizing voltage violations and improving grid resilience, this work offers a robust framework for voltage management in modern, evolving power distribution networks, particularly in the context of the global energy transition towards sustainable energy sources.

Climate change results in governments making policies for more renewable electricity generation. At the same time, electricity prices are rising. These factors result in an increasing number of photovoltaic (PV) systems being installed and connected to the low voltage (LV) grid. More than half of the areas in The Netherlands have capacity issues related to distributed power generation, which can be solved by reinforcing the electricity grid or reducing the combined power being fed into the grid. This thesis aims at investigating the possibilities and effect of reducing PV power put into the grid.

Grid reinforcement can not happen instantly and requires a considerable investment while currently not enough skilled labourers are available. Together with this labour shortage and time delay, a large cost is associated with this action. Another factor is the eventual over generation from distributed energy resources (DER’s) reducing power grid stability. All these considerations provide the basis for research into the possibilities of LV PV curtailment. There is a significant amount of research on the effects of curtailment, however this research mainly focuses on the LV effects or looks at not technically feasible curtailment options.

The current overloaded grid can be solved by different ways of curtailing LV PV alongside medium voltage (MV) and high voltage (HV) curtailment. Different options result in more curtailed energy correlated with lower grid loading or less curtailed energy but more residual network loading. The current German-implemented method sets a power limit of 70% of the installed power for small PV systems and requires a gradual curtailment possibility for larger systems. This approach has proven to provide a good trade-off between loading reduction and generation losses for now. Other curtailment may be more effective in the future. Therefore the research question of this thesis is How will LV connected PV curtailment affect the loading of the MV and HV network?.

In order to determine the impact of different curtailment methods a simulation based approach was chosen. Further on in this thesis the generation of power consumption and generation data will be discussed, as well as the implemented and explored curtailment methods. The selected networks are based on the amount of LV PV present together with the network capacity relative to the total amount of installed PV generation.

At the moment MV to LV transformers are the main bottleneck, which can only be solved by installing larger transformers or curtailing LV connections. When analysing all other assets an optimisation approach reduces grid loading the most, however this is far from necessary. In order to keep MV cable loading and HV to MV transformer loading within limits any form of curtailment is functional, even curtailment methods which do not impact LV connections.

The main takeaways from this thesis are that curtailment will become necessary in the coming years. Combined with the fact that LV curtailment has to be implemented within 10 years in order to avoid overloading the MV to LV transformers.

All in all a system wide approach is preferred, as this reduces loading with minimal energy curtailment. This can be achieved with an optimisation approach or propagating rules associated to the maximally loaded asset. Considerations, which should be explored further, are the financial impact, social impact of curtailment and complexity of the system. As these impact customers both directly and indirectly.

With the phasing out of the Net Energy Metering (NEM) scheme, the energy market is shifting towards alternative solutions like independent energy storage, already successful in countries like Belgium and Germany.
However, a single solution dominating the market is unlikely due to continuous innovation and the limitations of individual battery systems for prosumers and Distribution System Operators (DSOs). Community energy storage (CES) emerges as a promising alternative but lacks a defined business model, particularly for Dutch residential communities.
This study delves into the implementation of centralized community energy storage systems to boost prosumer profitability and mitigate grid congestion in the Dutch solar residential market, in the wake of the NEM scheme phase-out. Community energy storage applications are identified, along with their respective potential business models. The optimal application, in terms of prosumer profitability and grid relief, is selected, and its associated business model is developed using the Morphological business model designed for energy communities. Furthermore, a practical approach for integration is proposed, based on regulatory and market constraints, to enhance the potential for large-scale emergence. This approach includes defining key roles and responsibilities of stakeholders within the community and the corresponding allocation of value. Subsequently, a technical system design topology is outlined for each defined community. This system design delves into engineering details to analyze the energy interaction possibilities between consumers and the grid, along with the corresponding financial implications. Accordingly, the CES application’s performance is simulated and evaluated both technically and financially. The potential is presented by simulating the interactions between the community, the grid, and the optimal battery system. This optimal interaction arises from an optimization problem formulated to provide the optimal battery size and its corresponding energy profiles that minimize the total community cost. Finally, an energy distribution mechanism is carried out through conditional decision making to evaluate the cost and profitability allocation among consumers within the community.
The findings highlights the optimal application of CES, combining energy sharing with energy arbitrage, which significantly enhances the value of prosumers’ surplus PV energy, outperforming standard tariffs and avoiding grid feedback charges. This approach also provides consumers with access to more affordable shared community energy, while aiding DSOs in alleviating grid congestion and improving infrastructure capacity. The study suggests that the most effective strategy for widespread CES adoption involves collaboration between housing cooperatives and Energy Service Companies (ESCOs). Financially, this model entails
community managers overseeing initial investments, complemented by household contributions via usagebased or fixed service fees. The business model’s success is influenced by the type of grid connection, with Behind-The-Meter (BTM) offering flexibility but lacking standardization, and Front-of-The-Meter (FTM) encountering challenges related to community energy taxation. Modelling the optimal operation for both BTM and FTM connections demonstrates a significant decrease of energy costs and contribution to grid relief, highlighting load smoothing and peak shaving as key benefits. The research concludes that centralized CES systems can substantially elevate prosumer profitability and reduce grid congestion, leading to considerable energy savings and enhanced grid performance in the Dutch solar residential market.
To support the expansion of Community Energy Storage (CES) systems and energy communities, policymakers are advised to revise energy taxation policies and create frameworks aiding community grid formation, including simplifying regulations and offering incentives for residential initiatives. Researchers should adopt a multidisciplinary approach to explore regulatory, technical, economic, social, and environmental impacts on CES, focusing on regulatory effects, grid dynamics, cost-benefit models, community engagement, and environmental benefits. Industry stakeholders, such as Distribution System Operators, energy providers, Energy Service Companies, and housing cooperatives, should apply these research insights to develop and implement CES systems, fostering partnerships to address challenges and innovate in energy solutions, particularly in the evolving landscape post-Net Energy Metering, to enhance the role of community storage in sustainable energy systems.

Battery energy storage systems (BESS) that are integrated with the electricity grid have proven to be a suitable solution for reducing costs by providing flexible demand, but are often uneconomical when used solely for this purpose. Despite the research efforts showing the benefits of an energy-integrated harbor-area smart grid (HASG), the integration of onboard BESS in synergy with shore power connections has not been considered in the literature. Moreover, not from the financial perspective of the vessel owner.

This thesis addresses the integration of onboard BESS with a shore power system. More specifically, by presenting a cost-effective energy management system (EMS) that uses a stochastic approximation to define the charging and discharging decisions based on electricity prices. Additionally, the system takes the uncertainty of wind power production into account and reduces the power strain on the grid. Besides this, the research provides an analysis of the battery parameters that influence the cost-reducing ability of the EMS. As a result, the additional cost reduction presented while at berth may allow for previously uneconomical investments in onboard BESS for SSCV operators.

The wait-and-see (WS) approach was applied to provide an optimal energy scheduling solution with regard to uncertainty in wind power generation. Within the WS, three strategies were applied. By optimally scheduling a 5MWh BESS, the arbitrage, arbitrage + peak shaving, and peak shaving strategies respectively achieved a 1.4%, 15.7%, and 10.2% yearly reduction in electricity costs for the vessel operator during a 100-day stay in port.

The increased diffusion of Renewable Energy Sources (RES) into energy distribution systems gives rise to a number of issues that the Distribution Network Operators (DNOs) need to face. The dependency of RES on weather, along with their intermittent and non-dispatchable nature, urges us to develop frameworks that can guarantee a stable network, despite the power fluctuations.
In this thesis, we develop a Data-Driven Optimal Power Flow (OPF) formulation in order to achieve voltage regulation against operational problems that may occur because of the high PV penetration into energy distribution systems. The unpredictable PV generation deteriorates the system’s reliability and introduces instability. Thus, we develop appropriate formulations to define a limit on exports from residential PV owners to the grid. For this purpose, we employ Distribunally Robust Chance Constraint Programming (DRCCP), as a method that can handle constraints that depend on the uncertain PV generation and
residential demand. We capture the distributional uncertainties with an ambiguity set and we utilize the Wasserstein metric to parameterize the range of this set.
We divide the thesis into two case studies. In the first case study, we perform the DRCCP optimization with only few recorded data. We evaluate the results of our algorithm in a dataset of 500 days of recorded data and we achieve to reduce significantly the overvoltage instances. We tune the DRCCP in a way that we achieve at least 95% satisfaction of constraints that emerge from the DRCCP, avoiding overconservative solutions and high generation cuts that
could harm the PV owners.
In the second case study, we intend to exploit plenty of historical data, but the computational burden hinders us from using them in the DRCCP. Therefore, we employ Wasserstein Barycenters, and we utilize the Wasserstein distance in order to cluster the data. With Wasserstein Barycenters we reduce significantly the running time and we provide an efficient and robust output, for at least 95% satisfaction of the model’s constraints.

The electrification of all sectors is essential to achieve carbon neutrality by 2050. In terms of mobility, many nations around the globe envisage a zero-emission fleet resulting in a massive deployment of Electric Vehicles (EV). Nevertheless, this goes hand-in-hand with some challenges related to congestion and voltage problems in power grids caused by inadequate grid capacity. When grid reinforcement cannot keep up with the increasing demand for electricity, both in terms of planning and excessive financial means, smarter, more efficient solutions should solve the problem.

This thesis aims at providing an optimal strategy for consumers’ flexibility distribution by smartly charging EVs in Low Voltage (LV) grids. In that way, grid reinforcement should be reduced to a minimum. In order to do so, the main contribution of this work is covered by the development of a smart, optimal charging model that tries to comply with grid constraints including voltage and congestion boundaries. Mathematical (near) real-time optimisation is used together with the receding horizon optimisation principle. The work of this thesis relates to a typical urban LV grid in the Netherlands.

First, the required level of grid reinforcement was investigated by means of quantifying the voltage and congestion problems in the LV grid. This was done by applying uncontrolled charging scenarios up to 2050 for both winter and summer.

After investigating the uncontrolled scenarios, the need for an optimal smart charging strategy became apparent. This was developed afterwards. The results showed that in 2050, 94.5% of line congestion and 100% of transformer congestion and voltage problems could be bypassed with smart charging only. This was achieved by implementing grid constraints at the most vulnerable locations in the power grid.
Consequently, the need for grid reinforcement could be almost completely avoided till at least 2050.

Lastly, this work distinguishes itself from others by extensively reflecting on the integration of three concepts: a novel recently developed tariff structure, the effect of bidirectional charging, as well as the necessity for grid constraints. This allows summarising the benefits for all stakeholders including the distribution system operator, the charge point operator and EV owners. The final conclusion can be made that including these three concepts results in one of the most favourable investigated strategies for all three parties together.

Automated negotiation is a key form of interaction in systems composed of multiple autonomous agents with different preferences. Such interactions aim to reach agreements through an iterative process of making offers. With the growth of Peer-to-Peer (P2P) energy markets due to the development and deployment of a variety of small-scale electricity generation and storage devices (DERs), automated negotiation is seen as one of the advanced techniques that can improve the efficiency of energy distribution with the consideration of preferences of different entities. Opponent modeling is one of the essential abilities of automated negotiation agents that can further benefit automated negotiation. This project introduces a new opponent modeling technique considering the specific characteristics of P2P energy markets. These particular characteristics are a) Two automated negotiation agents can negotiate with each other many times, and b) The preferences of the users of agents are decided mainly by their energy consumption patterns, which usually do not have massive fluctuation across the year. The proposed opponent modeling method is developed from the idea of modeling the policy of a Reinforcement Learning agent. It uses a neural network to approximate the bidding strategy of the opposite automated negotiation agent. The network is learned based on the observations of offers exchanged in negotiations. With the learned network, the negotiation agent can predict the future actions of its opponent and make better decisions. We evaluated our opponent modeling with an existing automated negotiation system designed for off-grid energy trading. In experiments, the introduced opponent modeling always performs better than a random-guess model while modeling basic bidding strategies. Its performance is stable in dynamic environments where its opponent’s preference and bidding strategy may change randomly. It is also proved that the introduced method has potential for further improvement with the help of advanced opponent modeling techniques, which model the preference profile of the opponent. With our new opponent modeling method, the automated negotiators who take part in the P2P energy markets should be able to find better joint agreements that are preferred by both itself and its opponents. And in this case, a better joint agreement means a more efficient way of distributing energy.

In this thesis the build plans for a new residential neighbourhood are used as a basis for a distribution network of the neighbourhood of the future. Data is used to simulate power flows and an analysis is given on the current method used by DSO's. Furthermore the discussion is opened to suggest different methods for constructing sustainable distribution networks.

Increasing distributed generation of and demand for electrical energy results ever more in problems like congestion. Forecasting the demand, photovoltaic power generation and the number of electric vehicles connected to charging station for a residential neighbourhood can be an important part of a smarter distribution network for such a neighbourhood and can thereby increase the usage of renewable energy. In this thesis an overview of the design steps for making such a forecasting system is given and the steps are applied to a fictional dutch residential neighbourhood of the future. Some key findings are that for accurately forecasting the load, the temperature and time are the most important features. For accurately forecasting the PV power generation, especially the irradiation is most important, but time, horizontal view and humidity are also important features. Furthermore, it is shown that random forest regression models can accurately forecast both the demand and PV power generation with an accuracy above 90%. Artificial neural networks are also adequate models for the forecasting problems, but because they are harder to understand and not necessarily better, it is recommended to start with random forest regressors before making neural networks. Support vector machines seem less suitable for these particular forecasting problems.

Smart buildings, including photovoltaic (PV) generation, controllable electricity consumption, and a battery energy storage system, are expected to fulfill a crucial role in balancing out supply and demand in future power systems.
Energy management systems (EMS) present in these smart buildings control the operation of these various components.
Achieving an optimal dynamic control strategy is still challenging due to the stochastic nature of PV generation, electricity consumption patterns, and market prices.
Hence, this research developed an EMS that minimizes day-ahead electricity costs based on reinforcement learning (RL) with linear function approximation.
The proposed Q-learning with tile coding (QLTC) EMS is compared to the solutions found by the deterministic mixed-integer linear programming (MILP) model, which is needed to validate if the proposed approach reaches good-quality solutions. Furthermore, the QLTC's generalization capabilities are evaluated, a missing feature in literate.
A case study on an industrial manufacturing company in the Netherlands with historical electricity consumption, PV generation, and wholesale electricity prices is carried out to examine the QLTC EMS's performance.
The results show that the QLTC's returns convergence consistently to the MILP's negative electricity costs, indicating that the QLTC reaches a good-quality control policy.
The EMS effectively adjusts its power consumption to favorable price moments during one week of operation, where the electricity costs made by the QLTC comes 99\% close to MILP's electricity costs.
Furthermore, the results demonstrate that the QLTC approach can deploy a decent control policy without encountering the exact day of data by generalizing on previously learned control policies.
On average, it can deploy a control policy of 80\% near the MILP's optimum on a test week after being trained for 85 days of data.