IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (IEEE SmartGridComm 2021) — recording passcode: SGC-VC21
Technical Sessions
Session CNS2
Modelling and Simulation for Smart Grid Communications and Networking
Conference
9:00 AM
—
10:30 AM CEST
Local
Oct 27 Wed, 3:00 AM
—
4:30 AM EDT
Interoperability analysis of IEC61850 protocol using an emulated IED in a HIL microgrid testbed
Marziyeh Hemmati; Harshavardhan Palahalli; Giambattista Gruosso; Samuele Grillo
0
Expansion of distributed energy resources (DERs) leads to more complex and interconnected networks in smart grids. This increased the requirement of fast and standardized information exchanges for stable, resilient, and reliable operations in microgrids. To extend interoperability, modern power grids utilize a sophisticated network of Intelligent Electronic Devices (IEDs). These devices are able to communicate with one another using the IEC-61850 communication protocol. In this article, one particular architecture to inspect Generic Object Oriented Substation Event (GOOSE) services is proposed. phase one of the project resides in design details of an assumed micro- grid simulation as the testbed in Typhoon HIL, and modelling of the characteristics of a generic IED running on a Hardware-In-the-Loop device. While phase two of the project involves the HIL test setup as a novel methodology to approach the communication scenarios of mentioned commercial relays. one particular overload scenario is stated in more detail to investigate the performance of the protection mechanisms and GOOSE services emulated in the IED.
PLC-RF hybrid communication systems, model and simulation
Alfredo Sanz; Jose Carlos Ibar; Luis Lacasa
0
This paper is motivated by the evolution of PLC systems to new hybrid systems combining PLC and RF. In order to achieve a good performance and optimization of the system, the development of a simulator that allows debugging it is crucial. A fundamental part of this evolution is the modelling of the RF channel adapted to the MAC level simulators we are using. This paper presents an introduction to this evolution of PLC systems to new hybrid systems and the results of RF channel field measurements among with models of signal propagation and FER based on them.
Cyber-Power Co-Simulation for End-to-End Synchrophasor Network Analysis and Applications
Mohammed Mustafa Hussain; Dexin Wang; Sajan Sadanandan; Eshwar Nag Pilli; Renke Huang; Anurag. K Srivastava; Jianming Lian; Zhenyu Huang
0
The resiliency, reliability and security of the next generation cyber-power smart grid depend upon leveraging the advanced communication and computing technologies efficiently. Also, developing real-time data-driven applications is critical to enable wide-area monitoring and control of the cyber-power grid given high-resolution data from Phasor Measurement Units (PMUs). North American Synchrophasor Initiative Network (NASPlnet) provides guidance for PMU data exchanges. With the advancement in both networking and grid operation, it is necessary to evaluate the performance of different data flow architecture suggested by NASPInet and analyzing the impact on applications. Therefore, we need a cyber-power co-simulation framework that supports very large-scale co-simulation capable of running in parallel, high-performance computing platforms and capturing real-life network behavior. This work presents an end-to-end automated and user-driven cyber-power co-simulation using NS3 to model communication networks, GridPACK to model the power grid, and HELICS as a co-simulation engine. Comparative analysis of latency in synchrophasor networks and a performance evaluation of a power system stabilizer application utilizing PMU data in an IEEE 39 bus test system is presented using this co-simulation testbed.
Power and Communications Hardware-In-the-Loop CPS Architecture and Platform for DER Monitoring and Control Applications
Jianhua Zhang; Adarsh Hasandka; Kumaraguru Prabakar; S M Shafiul Alam; Bri-Mathias Hodge; Yazhou Jiang; Wenzhong Gao
0
The rapid growth of distributed energy resources (DERs) has prompted increasing interest in the monitoring and control of DERs through hybrid smart grid communications. The deployment of communications and computation has transformed the traditional physical power grid into a smart cyber-physical system (CPS). To fully understand the interdependency between physical grid and cyber netowrks, this study designed a power and communications hardware-in-the-loop (PCHIL) CPS architecture, which enables the flexible verification of DER monitoring and control with hybrid communications architectures and Internet protocols. Design, development and case study of a PCHIL testbed for the DER coordination are discussed in detail. In particular, the proposed platform integrates DER devices, Advanced Metering Infrastructures (AMIs), and a suite of hybrid communications networks for distribution automation applications. Case study on DER situational awareness and Volt-Var control validates the efficacy of this proposed PCHIL platform with hybrid communications designs. Results show that the HAN communication technologies play a critical role in hybrid designs and it is the bottleneck for DER applications. The high performance communication technologies is highly suggested applied in the HAN for enhanced monitoring and real-time control of DERs.
Impulse Noise Suppression for G.hn Broadband Power-Line Communication in Smart Grid
Szu-Lin Su; Nan Hsiung Huang
0
Power-line communication (PLC) system can exchange information over the existing electrical grid without much extra implementation cost. Such system will play an important role in the future Smart Grid. However, the system performance of PLC systems will be severely degraded by the multipath fading and random impulse noise (IN). This paper intends to evaluate the performance of different IN detection and reduction schemes combined with low-density parity-check (LDPC) decoding for the PLC systems based on G.hn (Gigabit Home Networking) specification. In particular, to improve the system performance, we modify the likelihood value calculation of the LDPC decoder with the information of signal-power and noise-power change due to the process of IN reduction and equalizer.
Session CYS2
Network attacks, detection, and control
Conference
9:00 AM
—
10:30 AM CEST
Local
Oct 27 Wed, 3:00 AM
—
4:30 AM EDT
Talking After Lights Out: An Ad Hoc Network for Electric Grid Recovery
Jan Janak; Hema Retty; Dana A Chee; Artiom Baloian; Henning Schulzrinne
1
When the electrical grid in a region suffers a major outage, e.g., after a catastrophic cyber attack, a "black start" may be required, where the grid is slowly restarted, carefully and incrementally adding generating capacity and demand. To ensure safe and effective black start, the grid control center has to be able to communicate with field personnel and with supervisory control and data acquisition (SCADA) systems. Voice and text communication are particularly critical. As part of the Defense Advanced Research Projects Agency (DARPA) Rapid Attack Detection, Isolation, and Characterization Systems (RADICS) program, we designed, tested and evaluated a self-configuring mesh network prototype called the Phoenix Secure Emergency Network (PhoenixSEN). PhoenixSEN provides a secure drop-in replacement for grid's primary communication networks during black start recovery. The network combines existing and new technologies, can work with a variety of link-layer protocols, emphasizes manageability and auto-configuration, and provides services and applications for coordination of people and devices including voice, text, and SCADA communication. We discuss the architecture of PhoenixSEN and evaluate a prototype on realistic grid infrastructure through a series of DARPA-led exercises.
Efficient Group-Key Management for Low-bandwidth Smart Grid Networks
Yacoub Hanna; Mumin Cebe; Suat Mercan; Kemal Akkaya
0
As Smart Grid comes with new smart devices and additional data collection for improved control decisions, this puts a lot of burden on the underlying legacy communication infrastructures that may be severely limited in bandwidth. Therefore, an alternative is to consider publish-subscribe architectures for not only enabling flexible communication options but also exploiting multicasting capabilities to reduce the number of data messages transmitted. However, this capability needs to be complemented by a communication-efficient group key management scheme that will ensure security of multicast messages in terms of confidentiality, integrity and authentication. In this paper, we propose a group-key generation and renewal mechanism that minimizes the number of messages while still following the Diffie-Hellman (DH) Key exchange. Specifically, the Control Center (CC) utilizes Shamir's secret key sharing scheme to compute points for each device using random pairs sent by group members. Such points are then utilized to derive the group key based on Lagrange interpolation. The hash-chain concept is employed to renew the group key without requiring further message exchanges, essentially achieving key renewal in a single message. We evaluated our protocol by creating an MQTT-based testbed supporting multicasting. The results show that number of messages are decreased significantly compared to alternative approaches.
An IEC 61850 MMS Traffic Parser for Customizable and Efficient Intrusion Detection
Heng Chuan Tan; Vyshnavi Mohanraj; Binbin Chen; Daisuke Mashima; Shing Kham Shing Nan; Aobo Yang
0
Manufacturing Message Specification (MMS) is a widely used protocol for communication in IEC 61850-based substations or in general, industrial control systems (ICSs). It does not have any built-in security mechanism, and thus it is vulnerable to cyber attacks. Zeek is an open-source network security monitoring tool that can be used as a network intrusion detection system (IDS) to analyze network packets and detect malicious traffic patterns in ICS. To our knowledge, no MMS protocol parser has been developed for Zeek. This is largely due to the vast number of MMS services (i.e., different packet types) and the complex data structure, making the design of an MMS parser particularly challenging. In addition, sending every parsed packet to Zeek for analysis may reduce the throughput and increase the packet processing latency. In this work, we explain the challenges involved in parsing MMS packets and detail our design choices when designing the MMS parser. To reduce the processing load, we implement filtering rules in our parser to customize which packets to send to Zeek script. Finally,
we formulated test cases to validate the correctness of our parser and conducted experiments to evaluate its throughput and latency. Results show that our MMS parser can achieve higher throughput and lower delay through custom filtering of packets compared to no filtering at all.
we formulated test cases to validate the correctness of our parser and conducted experiments to evaluate its throughput and latency. Results show that our MMS parser can achieve higher throughput and lower delay through custom filtering of packets compared to no filtering at all.
RICSel21 Data Collection: Attacks in a Virtual Power Network
Chih-Yuan Lin, August Fundin, Erik Westring, Tommy Gustafsson, and Simin Nadjm-Tehrani
0
Attacks against Supervisory Control and Data Acquisition (SCADA) systems operating critical infrastructures have increased since the appearance of Stuxnet. To defend critical infrastructures, security researchers need realistic datasets to evaluate and benchmark their defense mechanisms such as Intrusion Detection Systems (IDS). However, real-world data collected from critical infrastructures are too sensitive to share openly. Therefore, testbed datasets have become a viable option to balance the requirement of openness and realism. This study provides a data generation framework based on a virtual testbed with a commercial SCADA system and presents an openly available dataset called RICSel21, containing twelve attack scenarios against the SCADA system for power management.
Session LabTour
Laboratory Tour in E.ON ERC organised by OPAL-RT
Conference
10:30 AM
—
11:00 AM CEST
Local
Oct 27 Wed, 4:30 AM
—
5:00 AM EDT
Session COS3
Cyber-physical power systems
Conference
11:00 AM
—
12:30 PM CEST
Local
Oct 27 Wed, 5:00 AM
—
6:30 AM EDT
Analysis of Moving Target Defense in Unbalanced and Multiphase Distribution Systems Considering Voltage Stability
Mengxiang Liu; Chengcheng Zhao; Zhenyong Zhang; Ruilong Deng; Peng Cheng
1
Moving Target Defense (MTD) is a new technology to defend against the false data injection attacks (FDIAs) on distribution system state estimation (DSSE). It works by proactively perturbing the branch reactances. However, due to the challenges induced by the nonlinear dynamics and non-decoupling phases in the three-phase AC DSSE model, the analysis on the effectiveness and hiddenness of MTD, which are two basic metrics to evaluate the performance of MTD, has not yet been conducted. In this paper, we quantify the effectiveness and hiddenness using measurement residuals: First, the residuals are approximated by applying sensitivity analysis to the nonlinear AC SE problem, where the sensitivity of inexplicit branch admittance matrix to branch reactance is represented as the function of explicit impedance matrix; Based on the approximated residuals, we formulate an optimization problem to maximize the effectiveness of MTD while guaranteeing its hiddenness, and to ensure the voltage stability through minimizing the voltage variation introduced by MTD. The original problem is transformed to a polynomial optimization problem, where the near-optimum result can be obtained, based on the observation that the projection matrix involved in the approximated residual is almost invariant under MTD. Finally, extensive simulations are conducted on IEEE 13-bus test feeder to evaluate the performance of the proposed MTD.
Cyber-Physical Disaster Response of Power Supply Using a Centralised-to-Distributed Framework
Pudong Ge; Charalambos Konstantinou; Fei Teng
0
This paper proposes a cyber-physical cooperative recovery framework to maintain critical power supply, enhancing power systems resilience under extreme events such as earthquakes and hurricanes. Extreme events can possibly damage critical infrastructure in terms of power supply, on both cyber and physical layers. Microgrid (MG) has been widely recognised as the physical-side response to such blackouts, however, the recovery of cyber side is yet fully investigated, especially the cooperatively recovery of cyber-physical power supply. Therefore, a centralised-to-distributed resilient control framework is designed to maintain the power supply of critical loads. In such resilient control, controller-to-controller (C2C) wireless network is utilised to form the emergency distributed communication without a centralised base station. Owing to the limited reliable bandwidth that can be employed in C2C networks, the inevitable delay is considered in designing a discrete control framework, and the corresponding stability criteria are given quantitatively. Finally, the cyber-physical recovery framework is demonstrated effectively through simulations in MATLAB/Simulink.
Hybrid Modeling of Cyber-Physical Distribution Grids
Sina Hassani; Jan Bendtsen; Rasmus Olsen
0
Penetration of distributed generation into distribution grids brings new demands for both centralized and distributed control at the low-voltage level. In particular, when trying to coordinate the production from distributed generation, communication becomes an important aspect of control design. However, whereas local control typically occurs at sub-second resolution, communication between geographically separate locations based on e.g., smart meter data, commonly takes place at much lower frequencies, such as on an hourly basis or even slower. Therefore, novel distribution grids should be analyzed and controlled within the context of cyber-physical systems. Hybrid systems, which cover systems that have both continuous and discrete dynamics, provide the natural setting for such analysis. In this paper, a hybrid model of the distribution grid considering both the continuous states of the power network and the discrete nature of the communication is presented, capturing the different update rates of centralized and local controllers in the modeling process. Simulation results show good agreement with data from a real-life system.
Deep Reinforcement Learning For Online Distribution Power System Cybersecurity Protection
Tyson Bailey; Jay Johnson; Drew Levin
0
The sophistication and regularity of power system cybersecurity attacks has been growing in the last decade, leading researchers to investigate new innovative, cyber-resilient tools to help grid operators defend their networks and power systems. One promising approach is to apply recent advances in deep reinforcement learning (DRL) to aid grid operators in making real-time changes to the power system equipment to counteract malicious actions. While multiple transmission studies have been conducted in the past, in this work we investigate the possibility of defending distribution power systems using a DRL agent who has control of a collection of utility-owned distributed energy resources (DER). A game board using a modified version of the IEEE 13-bus model was simulated using OpenDSS to train the DRL agent and compare its performance to a random agent, a greedy agent, and human players. Both the DRL agent and the greedy approach performed well, suggesting a greedy approach can be appropriate for computationally tractable system configurations and a DRL agent is a viable path forward for systems of increased complexity. This work paves the way to create multi-player distribution system control games which could be designed to defend the power grid under a sophisticated cyber-attack.
Deep Learning Based Multi-Label Attack Detection for Distributed Control of AC Microgrids
Sheik Mohiuddin; Junjian Qi; Sasha Fung; Yu Huang; Yufei Tang
0
This paper presents a deep learning based multilabel attack detection approach for the distributed control in AC microgrids. The secondary control of AC microgrids is formulated as a constrained optimization problem with voltage and frequency as control variables which is then solved using a distributed primal-dual gradient algorithm. The normally distributed false data injection (FDI) attacks against the proposed distributed control are then designed for the distributed generator's output voltage and active/reactive power measurements. In order to detect the presence of false measurements, a deep learning based attack detection strategy is further developed. The proposed attack detection is formulated as a multi-label classification problem to capture the inconsistency and co-occurrence dependencies in the power flow measurements due to the presence of FDI attacks. With this multi-label classification scheme, a single model is able to identify the presence of different attacks and load change simultaneously. Two different deep learning techniques are compared to design the attack detector, and the performance of the proposed distributed control and the attack detector is demonstrated through simulations on the modified IEEE 34-bus distribution test system.
A Multigraph Modeling Approach to Enable Ecological Network Analysis of Cyber Physical Power Networks
Abheek Chatterjee; Hao Huang; Katherine Davis; Astrid Layton
1
The design of resilient power grids is a critical engineering challenge for the smooth functioning of society. Bio-inspired design, using a framework called the Ecological Network Analysis (ENA), is a promising solution for improving the resilience of power grids. However, the existing ENA framework can only account or for one type of flow in a network. Thus, the previous applications of ENA in power grid design were limited to the design and evaluation of the power flows only and could not account for the monitoring and control systems and their interactions that are critical to the operation of energy infrastructure. The present work addresses this limitation by proposing a multigraph-based modeling approach and modified ENA metrics that enable evaluation of the network organization and comparison to biological ecosystems for design inspiration. This work also compares the modeling features of the proposed model and the conventional graphical model of Cyber Physical Power Networks found in the literature to understand the implications of the different modeling approaches.
Session CYS3
Mitigation
Conference
11:00 AM
—
12:30 PM CEST
Local
Oct 27 Wed, 5:00 AM
—
6:30 AM EDT
Achieving Runtime State Verification Assurance in Critical Cyber-Physical Infrastructures
Abel O Gomez Rivera; Deepak K Tosh; Sachin Shetty
0
Industrial Cyber-Physical Systems (ICPS) are an essential backbone of national critical infrastructures. They help monitor and control crucial cyber-enabled services such as energy generation. Commonly ICPS monitors the physical process through Supervisory Control and Data Acquisition (SCADA) systems. The SCADA ecosystem takes critical real-time and future system operational decisions based on the runtime state behavior of field sensors. Traditional SCADA systems use legacy and insecure communication protocols such as the Modbus protocol that lack adequate security mechanisms to provide robust runtime state behavior assurance of constrained field sensors. Therefore, constrained field sensors are commonly vulnerable to standard semantic attacks that gradually change the behavior state of infected devices. This paper discusses process integrity assurance techniques necessary to enhance the security of behavior-based protocols such as the Modbus protocol. The Runtime State Verification (RSV) protocol proposed in this paper aims to address semantic attacks in the SCADA ecosystem by integrating behavior-based Mandatory Results Automata (MRA) and a Hyperledger Fabric (HLF) network. The RSV protocol provides high process integrity assurance through enhanced behavior-based MRA suitable for the constrained field devices. A proof of concept of the RSV protocol has been evaluated in an emulated water-tube boiler. Preliminary evaluations of the RSV protocol aimed to measure the efficiency of the proposed protocol by monitoring an Combustion Efficiency (CE) process necessary to preserve optimal combustion, thus minimizing costs and future maintenance of water-tube boilers. We analyze the overall network overhead and latency of the proposed RSV protocol by evaluating the HLF network performance and comparing the proposed RSV protocol with the state-of-art BloSPAI protocol. Through the preliminary evaluations of the proposed RSV protocol, this paper demonstrates that the proposed RSV protocol overcomes the shortcomings and network overhead of the BloSPAI protocol by integrating behavior-based authentication through novel MRAs and HLF networks.
An Investigation into Ecological Network Analysis for Cyber-Physical Power Systems
Hao Huang; Abheek Chatterjee; Astrid Layton; Katherine Davis
0
Power systems consist of interdependent cyber and physical networks: the physical network supplies energy to the cyber network for data exchange, while the data exchange provides for observation and operation of the power system. This mix of a physical and a cyber/information network means that network disturbances can be synthesized in both physical and cyber forms. Cyber incidents in particular have been increasing, highlighting the importance of both designing for and measuring the reliability and robustness of cyber networks. Industry guidelines exist to inform network designs for security and availability, but they are limited when it comes to being able to rigorously account for cyber-physical interdependencies in these networks. This presents cyber-physical network designers with a lack of design tools to guide network creation. This paper introduces a bio-inspired approach that has been successfully applied to the physical component of power networks, extending it for evaluation and guidance in a cyber-physical power system.
The power system's cyber-physical network is modeled here as an ecological food web. The potential benefits of selected ecological metrics related to food web resilience are evaluated, including robustness (R ECO ), average mutual information (AMI), cyclicity (λ max) and cycling index (CI). The paper investigates the use of these metrics and our understanding of food web characteristics to enhance the resilience and robustness of cyber- physical network design and data routing. Two cases are explored to highlight this potential, a 3-substation and an 8-substation cyber- physical network design and data routing. Two cases are explored to highlight this potential, a 3-substation and an 8-substation cyber-physical system. The analysis suggests that increasing redundancy in the network design and more equally distributing data flow can improve the security and availability of data being transferred to operators.
The power system's cyber-physical network is modeled here as an ecological food web. The potential benefits of selected ecological metrics related to food web resilience are evaluated, including robustness (R ECO ), average mutual information (AMI), cyclicity (λ max) and cycling index (CI). The paper investigates the use of these metrics and our understanding of food web characteristics to enhance the resilience and robustness of cyber- physical network design and data routing. Two cases are explored to highlight this potential, a 3-substation and an 8-substation cyber- physical network design and data routing. Two cases are explored to highlight this potential, a 3-substation and an 8-substation cyber-physical system. The analysis suggests that increasing redundancy in the network design and more equally distributing data flow can improve the security and availability of data being transferred to operators.
Preventing Outages under Coordinated Cyber-Physical Attack with Secured PMUs
Yudi Huang; Ting He; Nilanjan Chaudhuri; Thomas La Porta
0
Due to the severe consequences of the coordinated cyber-physical attack (CCPA), the design of defenses has gained a lot of attention. A popular defense approach is to eliminate the existence of attacks by either securing existing sensors or deploying secured PMUs. In this work, we improve this approach by lowering the defense target from eliminating attacks to preventing outages in order to reduce the required number of secured PMUs. To this end, we formulate the problem of PMU Placement for Outage Prevention (PPOP) as a tri-level non-linear optimization and transform it into a bi-level mixed-integer linear programming (MILP) problem. Then, we propose an alternating optimization algorithm to solve it optimally. Finally, we evaluate our algorithm on IEEE 30-bus, 57-bus, and 118-bus systems, which demonstrates the advantage of the proposed approach in significantly reducing the required number of secured PMUs.
A Localized Cyber threat Mitigation Approach for Wide Area Control of FACTS
Abhiroop Chattopadhyay; Alfonso Valdes; Peter Sauer; Reynaldo Nuqui
0
We propose a localized oscillation amplitude monitoring (OAM) method for the mitigation of cyber threats directed at the wide area control (WAC) system used to coordinate control of Flexible AC Transmission Systems (FACTS) for power oscillation damping (POD) of active power flow on inter-area tie lines. The method involves monitoring the inter-area tie line active power oscillation amplitude over a sliding window. We use system instability - inferred from oscillation amplitudes growing instead of damping - as evidence of an indication of a malfunction in the WAC of FACTS, possibly indicative of a cyber attack. Monitoring the presence of such a growth allows us to determine whether any destabilizing behaviors appear after the WAC system engages to control the POD. If the WAC signal increases the oscillation amplitude over time, thereby diminishing the POD performance, the FACTS falls back to POD using local measurements. The proposed method does not require an expansive system-wide view of the network. We simulate replay, control integrity, and timing attacks for a test system and present results that demonstrate the performance of the OAM method for mitigation.
Session GAS2
Forecasting methods and energy storage
Conference
3:00 PM
—
4:30 PM CEST
Local
Oct 27 Wed, 9:00 AM
—
10:30 AM EDT
A Comparative Analysis of Machine Learning Methods for Short-Term Load Forecasting Systems
Alejandro Parrado-Duque; Sousso Kelouwani; Kodjo Agbossou; Sayed Saeed Hosseini; Nilson Henao; Fatima Amara
0
End-users' electricity consumption is highly affected by weather conditions. The uncertain nature of these circumstances can highly challenge energy supply and demand balancing. The identification of explanatory variables that influence energy usage plays a key role in addressing this issue. This paper conducts a benchmark study of several machine learning methods to compare their ability to determine the most significant weather-related variables and estimate energy demand. Accordingly, it investigates fifteen climate features as predictors. These components are entered into eight algorithms that select different sets of meaningful features. The selected characteristics are exploited by five other techniques to predict energy usage. Subsequently, the outcomes are evaluated to define the most efficient forecasting process. The results of the selection procedure demonstrate that mains water and dry outdoor temperatures are the most descriptive variables. With regard to both algorithmic steps, the random forest method provides the best results with 60.78% forecasting ability. Indeed, the remarks, elaborated by this study, can assist with designing the effective load forecasting structures.
LSTM-based Multi-Step SOC Forecasting of Battery Energy Storage in Grid Ancillary Services
Ardiansyah Ardiansyah; Yeonghyeon Kim; Deokjai Choi
0
Battery energy storage (BES) participation in the grid ancillary services markets is increasing rapidly in recent years. To facilitate optimal participation, the need for accurate BES state-of-charge (SOC) forecasting is indispensable. In grid ancillary services, the development of SOC forecasting models should deal with uncertainties and corresponding stochastic processes that determine the BES SOC periodically. Several traditional and state-of-the-art machine learning (ML) techniques, ranging from decision-tree to deep learning methods, were used to solve this problem. However, developing a multi-step SOC forecasting model remains a challenge in this subject that is essential for optimal BES economic dispatch and unit commitment. Taking advantage of the Long short-term memory (LSTM) deep learning and its variants techniques which are proven to be a robust method for predicting sequentially dependent data in the time-series domain, this paper proposes LSTM-based multi-step SOC forecasting for BES operating in frequency regulation. Various developed models, i.e., Vanilla-LSTM, Vanilla-Gated Recurrent Units (GRU), Bidirectional-LSTM (Bi-LSTM), and Bi-GRU, are evaluated using real-world datasets. The evaluation results show that the developed models outperform the existing methods in terms of root mean square error (RMSE) and mean absolute error (MAE) evaluation metrics.
Neural network interpretability for forecasting of aggregated renewable generation
Yucun Lu; Ilgiz Murzakhanov; Spyros Chatzivasileiadis
0
With the rapid growth of renewable energy, lots of small photovoltaic (PV) prosumers emerge. Due to the uncertainty of solar power generation, there is a need for aggregated prosumers to predict solar power generation and whether solar power generation will be larger than load. This paper presents two interpretable neural networks to solve the problem: one binary classification neural network and one regression neural network. The neural networks are built using TensorFlow. The global feature importance and local feature contributions are examined by three gradient-based methods: Integrated Gradients, Expected Gradients, and DeepLIFT. Moreover, we detect abnormal cases when predictions might fail by estimating the prediction uncertainty using Bayesian neural networks. Neural networks, which are interpreted by the gradient-based methods and complemented with uncertainty estimation, provide robust and explainable forecasting for decision-makers.
Reinforcement Learning for Battery Energy Storage Dispatch augmented with Model-based Optimizer
Gayathri Krishnamoorthy
0
Reinforcement learning has been found useful in solving optimal power flow (OPF) problems in electric power distribution systems. However, the use of largely model-free reinforcement learning algorithms that completely ignore the physics-based modeling of the power grid compromises the optimizer performance and poses scalability challenges. This paper proposes a novel approach to synergistically combine the physics-based models with learning-based algorithms using imitation learning to solve distribution-level OPF problems. Specifically, we propose imitation learning-based improvements in deep reinforcement learning (DRL) methods to solve the OPF problem for a specific case of battery storage dispatch in the power distribution systems. The proposed imitation learning algorithm uses the approximate optimal solutions obtained from a linearized model-based OPF solver to provide a good initial policy for the DRL algorithms while improving the training efficiency. The effectiveness of the proposed approach is demonstrated using IEEE 34-bus and 123-bus distribution feeders with numerous distribution-level battery storage systems.
Session GC3
Regulation
Conference
3:00 PM
—
4:30 PM CEST
Local
Oct 27 Wed, 9:00 AM
—
10:30 AM EDT
Made with in Toronto · Privacy Policy · SmartGridComm 2020 · © 2021 Duetone Corp.