IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (IEEE SmartGridComm 2021) — recording passcode: SGC-VC21
Technical Sessions
Session CNS1
Sensing and Communication in Smart Grids
Conference
9:00 AM
—
10:30 AM CEST
Local
Oct 26 Tue, 12:00 AM
—
1:30 AM PDT
Minimizing Age of Information for Distributed Control in Smart Grids
Leonard Fisser; Andreas Timm-Giel
1
Exchanging information between actively-managed components is key for establishing distributed control schemes in future Low Voltage Distribution Grids (LVDGs). Each nodes measurements and control data has to be disseminated to all other nodes in the electrical grid, to allow for safe operation and control. However, periodic all-to-all flooding procedures are challenging for wireless communication networks and especially so, if end-to-end reliability is required. The effectiveness of a flooding protocol is well captured in the Age of Information (AoI) performance indicator, combining the time it took to disseminate a data chunk and the time in between updates. Existing flooding protocols are not concerned with keeping AoI low in scenarios where status updates have to be continuously distributed. We propose a novel Parallel Sequential All-to-All Flooding (PSAA) protocol which is tailored to LVDGs and tries to minimize the average AoI. Special focus is given to the relation between AoI and retransmissions which are necessitated by unreliable communication channels. We show that PSAA is able to significantly outperform simple flooding schemes in characteristic LVDG topologies. Extensive simulation studies highlight the interaction between retransmission timer and AoI.
Scalable Integration of High Sampling Rate Measurements in Deterministic Process-level Networks
Fabian Hohn; Viktoria Fodor; Giovanni Zanuso; Lars Nordström
0
Travelling wave (TW) based protection functions, which process the time of arrival of TWs, require high sampling rates in the range of hundreds of kilohertz to several megahertz. In digital substations merging units (MU) publish the sampled values of current and voltage signals on process-level networks. However, publishing these highly sampled signals for TW-based protection functions limits severely the number of MUs in a process-level network due to the significant increase of communication load. To circumvent this problem, distributed signal processing units (DSPU) extract directly the necessary signal features and publish these at a lower publishing rate in order to decrease the communication load. This paper provides an mathematical analysis on the scalable integration of DSPUs in deterministic process-level networks based on the High-availability Seamless Redundancy (HSR) protocol, time-aware network nodes and traffic scheduling. It is shown that the distributed signal processing architecture provides a scalable integration of high sampling rate measurements for TW-based protection functions. Lastly the analytical model has been validated through a discrete-event simulation.
Prototyping Multi-Protocol Communication to enable semantic interoperability for Demand response Services
Nikolai Galkin; Valeriy Vyatkin; Chen-Wei Yang; Lars Nordström
0
In future demand response scenarios, a multitude of different types of resources are potentially to be used, e.g., electric vehicles, flexible residential loads, and battery storage systems. To solve the problem of realtime communication of data among the various resources, it is likely that several different communication protocols and most importantly differing semantic models, must be used. An aggregator utilizing several types of resources is therefore potentially faced with a problem of semantic interoperability. This paper addresses this challenge by presenting a testbed consisting of a microgrid model integrated with several controllers communicating with industrial grade protocols for demand response, including IEC 61850, OpenChargePoint protocol (OCPP), OpenADR and UDP. The testbed forms a basis for further development of a semantic canvas to enable forecasting, activation and clearing of heterogenous demand response resources.
A Generalized Nash Equilibrium analysis of the interaction between a peer-to-peer financial market and the distribution grid
Ilia Shilov; Helene Le Cadre; Ana Busic
0
We consider the interaction between the distribution grid (physical level) managed by the distributed system operator (DSO), and a financial market in which prosumers optimize their demand, generation, and bilateral trades in order to minimize their costs subject to local constraints and bilateral trading reciprocity coupling constraints. We model the interaction problem between the physical and financial levels as a noncooperative generalized Nash equilibrium problem. We compare two designs of the financial level prosumer market: a centralized design and a peer-to-peer fully distributed design. We prove the Pareto efficiency of the equilibria under homogeneity of the trading cost preferences. In addition, we prove that the pricing structure of our noncooperative game does not permit free-lunch behavior. Finally, in the numerical section we provide additional insights on the efficiency loss with respect to the different levels of agents' flexibility and amount of renewables in the network. We also quantify the impact of the prosumers' pricing on the noncooperative game social cost.
Data-driven Electric Vehicle Charging Station Placement for Incentivizing Potential Demand
Chenxi Sun; Tongxin Li; Xiaoying Tang
0
It is believed that Electric Vehicles (EVs) will play an increasing important role in making the city greener and smarter. However, a critical challenge raised by the transportation electrification process is the proper planning of city-wide EV charging infrastructures, i.e. the siting and sizing of charging stations, especially for the cities that just start promoting the adoption of EVs. In this paper, we investigate the following problem: For a city with a limited budget for public EV charging infrastructure construction, where should the charging stations be deployed in order to promote the transition of EVs from traditional cars? We propose a ..-nearest model that captures people's satisfaction towards a certain design and formulate the EV charging station placement problem as a monotone submodular maximization problem, equipped with gridded population data and trip data. We then propose a greedy based algorithm to sub-optimally solve the problem efficiently with a provable approximation ratio. A case study using fine-grained Haikou population data, POI data and trip data is also provided to demonstrate the effectiveness of our approach.
Session COS1
Power system operation
Conference
9:00 AM
—
10:30 AM CEST
Local
Oct 26 Tue, 12:00 AM
—
1:30 AM PDT
Measurement-based Condition Monitoring of Railway Signaling Cables
Rathinamala Vijay; Gautham Prasad; Yinjia Huo; Sachin Sm; Prabhakar TV
0
We propose a composite diagnostics solution for railway infrastructure monitoring. In particular, we address the issue of soft-fault detection in underground railway cables. We first demonstrate the feasibility of an orthogonal multi-tone time domain reflectometry based fault detection and location method for railway cabling infrastructure by implementing it using software defined radios. Our practical implementation, comprehensive measurement campaign, and our measurement results guide the design of our overall composite solution. With several diagnostics solutions available in the literature, our conglomerated method presents a technique to consolidate results from multiple diagnostics methods to provide an accurate assessment of underground cable health. We present a Bayesian framework based cable health index computation technique that indicates the extent of degradation that a cable is subject to at any stage during its lifespan. We present the performance results of our proposed solution using real-world measurements to demonstrate its effectiveness.
Analyzing Extrapolation-based Co-simulation Methods with Control Systems: A Linearized Two-Area Power System with Automatic Generation Control
Andrés F. Acosta; Ernesto Perez; Jairo Espinosa; Antonello Monti
0
This article deals with the incorporation of control systems in co-simulation scenarios. We develop a framework where control systems are included as additional subsystems in an existing co-simulation, called open-loop co-simulation, and with disturbances and external control loops included as external inputs. Moreover, we derive a specific co-simulation method using Zero-Order-Hold extrapolation and interpolation of the coupling and external inputs, respectively, assuming that the external inputs are known. A similar process can be used to derive co-simulation scenarios based on higher degree extrapolation methods. We demonstrate the proposed framework in a co-simulation of a linearized two-area power system and show how the incorporation of an Automatic Generation Control impacts the global co-simulation error for a given macro step size.
On the specification of requirements for the activation of Frequency Containment Reserves
Philipp Maucher; Hendrik Lens
0
The dynamic requirements for the provision of Frequency Containment Reserves (FCR) in Continental Europe are defined in the respective network codes (e.g. System Operation Guideline). However, this definition is precise only for a sudden frequency deviation of ±200 mHz. The requirements for smaller and/or slower frequency deviations are only described indirectly by referring to the case of ±200 mHz. As a result, different interpretations are possible, among which requiring activation dynamics that a) correspond to a linear time-invariant (LTI) system or b) exhibit a constant rate of change of power (RoCoP). This paper assesses the effects of these two different requirement interpretations on FCR providers and system stability by comparing their effect for different frequency deviations. It turns out that the RoCoP interpretation is disadvantageous, as it provides a slower response for large and fast frequency deviations and a fast response for small frequency deviations. Apart from Battery Energy Storage Systems (BESS), most FCR providers cannot
perform FCR activation with a fixed RoCoP.
In a further step, we consider the effects of the different requirement interpretations on system stability. For a constant RoCoP, it is assumed that the FCR is provided by BESS, while a conventional power plant model is used to implement LTI behavior. The comparison is performed both with model parameters corresponding to the current grid and with model parameters corresponding to a future grid. For each grid model, two scenarios are considered: The first scenario considers active power imbalances caused by load noise only (normal operation), while the second takes an additional significant generation outage into account (contingency). The results show that, in the load noise scenario, FCR activation with constant RoCoP reduces the frequency deviations slightly at the cost of higher total FCR provision and higher maximum FCR activation. However, in case of an additional generation outage, constant RoCoP activation results in a larger maximum frequency deviation, which means that the stability margin of the system is reduced.
perform FCR activation with a fixed RoCoP.
In a further step, we consider the effects of the different requirement interpretations on system stability. For a constant RoCoP, it is assumed that the FCR is provided by BESS, while a conventional power plant model is used to implement LTI behavior. The comparison is performed both with model parameters corresponding to the current grid and with model parameters corresponding to a future grid. For each grid model, two scenarios are considered: The first scenario considers active power imbalances caused by load noise only (normal operation), while the second takes an additional significant generation outage into account (contingency). The results show that, in the load noise scenario, FCR activation with constant RoCoP reduces the frequency deviations slightly at the cost of higher total FCR provision and higher maximum FCR activation. However, in case of an additional generation outage, constant RoCoP activation results in a larger maximum frequency deviation, which means that the stability margin of the system is reduced.
Online Distributed Optimization in Radial Power Distribution Systems: Closed-Form Expressions
Rabayet Sadnan; Tom Asaki; Anamika Dubey
0
The limitations of centralized optimization methods in managing power distribution systems operations motivate distributed control and optimization algorithms. However, the existing distributed optimization algorithms are inefficient in managing fast varying phenomena, resulting from highly variable distributed energy resources (DERs). Related online distributed control methods are equally limited in their applications. They require thousands of time-steps to track the network-level optimal solutions, resulting in slow performance. We have previously developed an online distributed controller that leverages the system's radial topology to achieve network-level optimal solutions within a few time steps. However, it requires solving a node-level nonlinear programming problem at each time step. This paper analyzes the solution space for the node-level optimization problem and derives the analytical closed-form solutions for the decision variables. The theoretical analysis of the node-level optimization problem and obtained closed-form optimal solutions eliminate the need for embedded optimization solvers at each distributed agent and significantly reduce the computational time and optimization costs.
Reliability-Security Trade-Off for Distributed Reactive Power Control in Transactive Grid
Muhammad Ramadan Saifuddin; David Yau; Xin Lou
3
Under the trend of deregulated Volt/VAR ancillary service market, power distribution grid (PDG) is seeing a growing demand for personally owned distributed energy resources (DERs) installed behind-the-meter as value adding participants. A trustworthy cyber-physical network thus becomes essential for coordinating these decentralized participants (e.g., by aggregators) in supporting Volt/VAR optimisation, a critical conservation voltage reduction (CVR) operation. Meanwhile, oversized inverters, which reserve a larger reactive power (VAR) capacity than needed for real power generation, provide incentive payouts during market participation; they are thus likely to be adopted by future customers. This adoption, as our findings show however, inaugurates a fundamental reliability-security trade-off, when the surplus VAR capacity, in the wrong hands of cyber attackers, can become a stronger weapon for damaging voltage control as a malicious intent. This paper presents novel analysis of key mechanisms and impacts of a class of data integrity attacks against voltage control during CVR. Evaluation results using a realistic 118-bus test system show that tampering with Volt/VAR control in prosumer-side DER and metering devices, which service D-STATCOM, can cause harmful power quality degradation (e.g., excessive voltage dips) or even power interruption. The results also quantify (i) trade-offs between better Volt/VAR control (i.e., increased reliability) and heightened potency of data integrity attacks (i.e. weakened security) under DER inverter oversizing; and (ii) impacts of these attacks under salient global trends such as increasing DER adoption.
Session COS2
Flexibility and demand response
Conference
11:00 AM
—
12:30 PM CEST
Local
Oct 26 Tue, 2:00 AM
—
3:30 AM PDT
Capturing Battery Flexibility in a General and Scalable Way Using the FlexOffer Model
Fabio Lilliu; Torben B Pedersen; Laurynas Siksnys
0
To solve the problems caused by the intermittent generation of Renewable Energy Sources, the concept of energy flexibility is of utmost importance, and batteries are devices with high potential in this regard. However, current exact mathematical models specifying battery flexibility cannot scale (exponentially growing runtime) with long time horizons and many batteries. In this paper, we propose to use the FlexOffer (FO) model for this purpose, because: 1) FO is a general model, capturing all types of flexible assets in a unified format and 2) being approximate, it scales very well in terms of number of devices and time horizons. First, we describe the different types of FOs: \emph{standard}, \emph{total-energy constraint} and \emph{dependency-based}(DFOs). Then, we present and discuss FO generation techniques, and provide an analytic method for generating DFOs. Finally, we perform simulations for measuring flexibility in economic terms and time needed for optimization and aggregation. When the time for charging/discharging the battery is known in advance, \textit{total energy constraints} and DFOs can retain up to 100% of flexibility; if not, DFOs can capture up to 61.2% of the optimal profits. The FO model optimizes battery usage in under 0.137 seconds, which is 5 orders of magnitude faster than the 4.7 hours for the exact model, making the latter infeasible in practice. FO aggregation retains over 72.4% of the total flexibility and can be performed for up to 6,000 batteries or 96 time units, while exact models fail at 21 time units and 400 batteries. Lastly, DFO analytic generation is 3 orders of magnitude faster than the original method for a 24 time units horizon
Optimization Strategy for Energy Allocation through Cooperative Storage Management
Johann Leithon; Stefan Werner; Visa Koivunen
0
We propose a strategy to optimize energy utilization through battery management in a cooperative environment where households share access to a community-owned energy farm. The households are equipped with lossy rechargeable batteries, which exhibit a non-linear discharging behavior. To devise our strategy, we first design the battery discharging operation in each household, and then we optimize the energy allocation policy among participating users. Our proposed strategy seeks to minimize the collective energy expenditure, and accounts for time- and location-dependent electricity prices. Both the battery discharging operation and the energy allocation policy are designed by solving constrained optimization problems. Specifically, calculus of variations and optimal control theory are used to provide explicit solutions and determine closed-form performance estimates. Extensive simulations are presented to validate our analysis and evaluate the impact of different system parameters.
Short-Term Residential Load Forecasting Based on Federated Learning and Load Clustering
Yu He; Fengji Luo; Gianluca Ranzi; Weicong Kong
0
Power load forecasting plays a fundamental role in modern energy systems' operations. While traditional load forecasting applies to bus-level aggregated load data, widespread deployment of advanced metering infrastructure creates an opportunity to fine-grained monitor the power consumption of single households and to predict their load requirements. This paper proposes a distributed residential load forecasting framework that combines federated learning and load clustering techniques. The system firstly applies a K-means clustering algorithm to divide a group of residential users into multiple clusters based on their historical power consumption patterns. For each cluster, the system then applies a federated learning process to enable the users in that cluster to collaboratively train their local load prediction models without physically sharing their load data. Experiments and comparison studies are conducted based on a real Australian residential load dataset to validate the proposed approach and to highlight its ease of use.
Optimal Cycling of a Heterogenous Battery Bank via Reinforcement Learning
Vivek Deulkar; Jayakrishnan Nair
0
We consider the problem of optimal charging/discharging of a bank of heterogenous battery units, driven by stochastic electricity generation and demand processes. The batteries in the battery bank may differ with respect to their capacities, ramp constraints, losses, as well as cycling costs. The goal is to minimize the degradation costs associated with battery cycling in the long run; this is posed formally as a Markov decision process. We propose a linear function approximation based Q-learning algorithm for learning the optimal solution, using a specially designed class of kernel functions that approximate the structure of the value functions associated with the MDP. The proposed algorithm is validated via an extensive case study.
Asset Participation and Aggregation in Incentive-Based Demand Response Programs
Utkarsha Agwan; Costas Spanos; Kameshwar Poolla
2
In order to manage peak-grid events, utilities run incentive-based demand response (DR) programs in which they offer an incentive to assets who promise to curtail power consumption, and impose penalties if they fail to do so. We develop a stochastic model for the curtailment capability of these assets, and use it to derive analytical expressions for the optimal participation (i.e., promised curtailment) and profitability from the DR asset perspective. We also investigate the effects of risk-aversion and curtailment uncertainty on both promised curtailment and profit. We use the stochastic model to evaluate the benefits of forming asset aggregations for participation in DR programs, and develop a numerical test to estimate asset complementarity. We illustrate our results using load data from commercial office buildings.
Session CYS1
False data injection attacks
Conference
11:00 AM
—
12:30 PM CEST
Local
Oct 26 Tue, 2:00 AM
—
3:30 AM PDT
CHIMERA: A Hybrid Estimation Approach to Limit the Effects of False Data Injection Attacks
Xiaorui Liu; Yaodan Hu; Charalambos Konstantinou; Yier Jin
0
The reliable operation of power grid is supported by energy management systems (EMS) that provide monitoring and control functionalities. Contingency analysis is a critical application of EMS to evaluate the impacts of outages based on the grid state variables, and allow operators to prepare for system failures. However, false data injection attacks (FDIAs) against state estimation have demonstrated the possibility of compromising sensor measurements and falsifying the estimated power system states. As a result, FDIAs may mislead the system operations and other EMS applications including contingency analysis and optimal power flow. In this paper, we assess the effect of FDIAs on contingency analysis and demonstrate that such attacks can affect the resulted number of contingencies. In order to mitigate the FDIA impact on contingency analysis, we propose CHIMERA, a hybrid attack-resilient state estimation approach that integrates model-based and data-driven methods. CHIMERA combines the physical grid information with a Long Short Term Memory (LSTM)-based deep learning model by considering a static loss of weighted least square errors and a dynamic loss of the difference between the temporal variations of the actual and the estimated active power. Our simulation experiments based on the load data from New York state demonstrate that CHIMERA can effectively mitigate 91.74% of the cases in which FDIAs can maliciously modify the contingencies.
Detecting Attacks on Synchrophasor Protocol Using Machine Learning Algorithms
Kolten C Knesek; Patrick Wlazlo; Hao Huang; Abhijeet Sahu; Ana E Goulart; Katherine Davis
0
Phasor measurement units (PMUs) are commonly used in power grids across North America to measure the amplitude, phase, and frequency of an alternating voltage or current. PMU's use the C37.118 protocol to send telemetry to phasor data collectors (PDC) and human machine interface (HMI) workstations in a control center. In today's modern grid, the C37.118 protocol utilizes internet protocols without any authentication mechanism. This means that the protocol is vulnerable to false data injection (FDI) and false command injection (FCI). In order to study different scenarios in which C37.118 protocol's integrity and confidentiality can be compromised, we created a testbed that emulates a C37.118 communication network. In this testbed we are able to conduct FCI and FDI attacks on real-time C37.118 data packets using a packet manipulation tool called Scapy. Using this platform, we generated C37.118 FCI and FDI datasets which are processed by multi-label machine learning classifier algorithms, such as Decision Tree (DT), k-Nearest Neighbor (kNN), and Naive Bayes (NB), to find out how effective machine learning can be at detecting such attacks. Our results show that the DT classifier had the best precision and recall rate.
A Stealthier False Data Injection Attack against the Power Grid
Weili Yan; Xin Lou; David Yau; Ying Yang; Muhammad Ramadan Saifuddin; Jiyan Wu; Marianne Winslett
2
We use discrete-time adaptive control theory to design a novel false data injection (FDI) attack against automatic generation control (AGC), a critical system that maintains a power grid at its requisite frequency. FDI attacks can cause equipment damage or blackouts by falsifying measurements in the streaming sensor data used to monitor the grid's operation. Compared to prior work, the proposed attack (i) requires less knowledge on the part of the attacker, such as correctly forecasting the future demand for power; (ii) is stealthier in its ability to bypass standard methods for detecting bad sensor data and to keep the false sensor readings near historical norms until the attack is well underway; and (iii) can sustain the frequency excursion as long as needed to cause real-world damage, in spite of AGC countermeasures. We validate the performance of the proposed attack on realistic 37-bus and 118-bus setups in PowerWorld, an industry-strength power system simulator trusted by real-world operators. The results demonstrate the attack's improved stealthiness and effectiveness compared to prior work.
Defense against Power System Time Delay Attacks via Attention-based Multivariate Deep Learning
Shahram Ghahremani; Rajvir Sidhu; David Yau; Ngai-Man Cheung; Justin Albrethsen
1
Time delay attacks pose a threat to power systems that conventional cybersecurity methods do not adequately address. Conventional methods analyze the contents of network packets to identify threats; this is not effective against time delay attacks, which do not alter packet contents. To detect and identify time delay attacks, a new method is needed. In this paper, a novel and data-driven deep learning (DL) approach is developed to detect time delay attacks on power systems and simultaneously identify both the time of attack and attack magnitude. While conventional DL networks struggle with multivariate long time series data generated by power systems, this can be improved using attention mechanisms. In this paper, dual attention mechanisms (DA) are used to focus and improve a gated recurrent unit (GRU) network for detecting and identifying time delay attacks. A comparative analysis shows the proposed GRU-DA approach outperforms conventional DL, machine learning (ML), and statistical methods while maintaining low model complexity.
Attack Detection and Localization in Smart Grid with Image-based Deep Learning
Mostafa Mohammadpourfard; V. M. Istemihan Genc; Subhash Lakshminarayana; Charalambos Konstantinou
0
Smart grid's objective is to enable electricity and information to flow two-way while providing effective, robust, computerized, and decentralized energy delivery infrastructure. This necessitates the use of state estimation-based techniques and real-time analysis to ensure that effective controls are deployed properly. However, the reliance on communication technologies makes such systems susceptible to sophisticated data integrity attacks imposing serious threats to the overall reliability of smart grid. To detect such attacks, advanced and efficient anomaly detection solutions are needed. In this paper, a two-stage deep learning-based framework is carefully designed by embedding power system's characteristics enabling precise attack detection and localization. First, we encode temporal correlations of the multivariate power system time-series measurements as $2$D images using image-based representation approaches such as Gramian Angular Field (GAF) and Recurrence Plot (RP) to obtain the latent data characteristics. These images are then utilized to build a highly reliable and resilient deep Convolutional Neural Network (CNN)-based multilabel classifier capable of learning both low and high level characteristics in the images to detect and discover the exact attack locations without leveraging any prior statistical assumptions. The proposed method is evaluated on the IEEE 57-bus system using real-world load data. Also, a comparative study is carried out. Numerical results indicate that the proposed multi-class cyber-intrusion detection framework outperforms the current conventional and deep learning-based attack detection methods.
Session GC1
5G for Energy
Conference
11:00 AM
—
12:30 PM CEST
Local
Oct 26 Tue, 2:00 AM
—
3:30 AM PDT
5G as a Game Changer for the Energy Transition
Antonello Monti (RWTH), Kristian Winter (Vattenfall), Kalina Barboutov (Hitachi Energy), Sergio Ramos Pinto (EUTC), Fiona Williams (Ericsson), Chair
0
This talk does not have an abstract.
Session GAS1
Data driven grid services and frequency control
Conference
3:00 PM
—
4:30 PM CEST
Local
Oct 26 Tue, 6:00 AM
—
7:30 AM PDT
Data-Driven Frequency Regulation Reserve Prediction Based on Deep Learning Approach
Shiyao Zhang; Ka-Cheong Leung
0
Day-ahead frequency regulation reserves can be procured to compensate power imbalance capacity for the purpose of stabilizing the power system. Due to the intermittency and uncertainty characteristics of renewable generations in a general power system, the dynamic nature of multi-scale system features cannot be fully captured through the existing approaches. This further causes inaccurate prediction and ineffective system operation. To tackle this issue, the deep learning techniques can be utilized to accurately predict the amount of reserves needed for the next-day system operations in an efficient manner. In this paper, we propose a deep learning approach on predicting frequency regulation reserves in a general power system. First, we use the power flow model to generate the net active power imbalance, frequency regulation reserves, and power matrix of a general power system. Second, we combine multiple dynamic system features into a complete input dataset and perform data pre-processing before model training and testing. Third, the proposed deep long short-term memory (DLSTM) model is developed to accurately predict the net active power imbalance in the system, as well as predicting the frequency regulation reserves. Our simulation results show that, when considering the entire power network information, our proposed deep learning approach outperforms the four baseline techniques on predicting the frequency regulation reserves in a general power system. These promising results contributes to large economical benefits in power system operations.
Exploring deterministic frequency deviations with explainable AI
Johannes Kruse; Benjamin Schäfer; Dirk Witthaut
0
Deterministic frequency deviations (DFDs) critically affect power grid frequency quality and power system stability. A better understanding of these events is urgently needed as frequency deviations have been growing in the European grid in recent years. DFDs are partially explained by the rapid adjustment of power generation following the intervals of electricity trading, but this intuitive picture fails especially before and around noonday. In this article, we provide a detailed analysis of DFDs and their relation to external features using methods from eXplainable Artificial Intelligence. We establish a machine learning model that well describes the daily cycle of DFDs and elucidate key interdependencies using SHapley Additive exPlanations. Thereby, we identify solar ramps as critical to explain patterns in the Rate of Change of Frequency (RoCoF).
Solving Unit Commitment Problems with Multi-step Deep Reinforcement Learning
Jingtao Qin; Nanpeng Yu; Yuanqi Gao
1
Solving the unit commitment (UC) problem in a computationally efficient manner is a critical issue of electricity market operations. Optimization-based methods such as heuristics, dynamic programming, and mixed-integer quadratic programming (MIQP) often yield good solutions to the UC problem. However, the computation time of optimization-based methods grows exponentially with the number of generating units, which is a major bottleneck in practice. To address this issue, we formulate the UC problem as a Markov decision process and propose a novel multi-step deep reinforcement learning (RL)-based algorithm to solve the problem. We approximate the action-value function with neural networks and design an algorithm to determine the feasible action space. Numerical studies on a 5-generator test case %bench-test UC problem show that our proposed algorithm significantly outperforms the deep Q-learning and yields similar level of performance as that of MIQP-based optimization in terms of optimality. The computation time of our proposed algorithm is much shorter than that of MIQP-based optimization methods.
Mining energy consumption data of industrial systems to identify and characterize energy flexibility capabilities
Alejandro Tristan; Can Kaymakci
0
Industrial energy flexibility can play a pivotal supporting role in the transition towards renewable energy sources. Nonetheless, to harness the vast potential of industrial energy flexibility operation-friendly energy flexibility measures need to be identified and characterized. This work presents a step by step approach to mine historical energy consumption data of an industrial system using the k-means algorithm with support of the average silhouette score method to establish the system's typical operation profiles. These profiles can then be used not only to identify specific energy flexibility measures but their energy flexibility potential among other characterization parameters. The paper presents two representative use case examples and concludes by enumerating the benefits and providing an outlook of improvement opportunities for the developed approach.
Session GC2
Open Source for Energy
Conference
3:00 PM
—
4:30 PM CEST
Local
Oct 26 Tue, 6:00 AM
—
7:30 AM PDT
Made with in Toronto · Privacy Policy · SmartGridComm 2020 · © 2021 Duetone Corp.