Search

This paper proposes a novel high-performance deep learning framework for load recognition. The framework consists of a deep-shallow model and a fast backpropagation (FBP) algorithm. In the deep-shallow model, power-related wave patterns are perceived by a convolutional neural network (CNN), and feature statistics of the power consumption data are analyzed by a sparse feed-forward neural network. The new architecture improves model interpretability and prevents information loss problems in CNNs. The architecture also greatly boosts the convergence speed and significantly enhances the test set accuracy of a neural net-work. Compared with conventional CNN models utilized by many load recognition applications, the FBP algorithm that constitutes four sub-algorithms converges faster at the start of the training process and reduces at least 87.5% of the filter gradient compu-tations on average. The deep-shallow-fast model that combines the deep-shallow model and the FBP algorithm attains 97.62% accuracy on the test set in the load recognition task. To fully utilize the training data, a data augmentation technique is in-vented that transforms the voltage and current time series into an image-like 4-D tensor. Experiments illustrate that the proposed framework is much more accurate and converges considerably faster than the conventional CNN model that many deep-learning-based load recognition applications are based up-on.

Destroying image integrity in scientific papers may result in serious consequences. Inappropriate duplication and fabrication of images are two common misconducts in this aspect. The rapid development of artificial-intelligence technology has brought to us promising image-generation models that can produce realistic fake images. Here, we show that such advanced generative models threaten the publishing system in academia as they may be used to generate fake scientific images that cannot be effectively identified. We demonstrate the disturbing potential of these generative models in synthesizing fake images, plagiarizing existing images, and deliberately modifying images. It is very difficult to identify images generated by these models by visual inspection, image-forensic tools, and detection tools due to the unique paradigm of the generative models for processing images. This perspective reveals vast risks and arouses the vigilance of the scientific community on fake scientific images generated by artificial intelligence (AI) models.

Risk preference is an important factor in electricity market strategy analysis and decision-making. The existing methods of risk preference analysis need to design and execute questionnaires or experiments on the subjects, and hence are costly and time-consuming for bidding in electricity markets. This article proposes a new method of data-driven risk preference analysis for power generation plants based on historical data and inverse reinforcement learning. Historical data are transformed to the transition function model according to the specific market mechanism. An adjusted inverse reinforcement learning model is thereafter proposed along with the optimization objective and technical constraints. The proposed method is tested in a simulated electricity market environment using the Australian Energy Market Operator (AEMO) day-ahead bidding data. Simulation results show that 1) thermal power plants prefer to adjust risk preferences within the day; 2) apart from the thermal power plants, the rest types of power plants are risk-neutral; 3) the daily risk preference trend of the thermal power plants varies in different seasons and is closely related to the load level.

Residential demand response (DR) is recognized as a promising approach to improve grid energy efficiency and relieve the network stress. Many studies have been conducted to design home energy management systems that directly schedule and control the household appliances. Distinguished from existing works, this article proposes a personalized recommendation system (PRS) to learn energy-efficient household appliance usage experiences from a large scale of residential users, and recommends suitable appliance usage plans to users while taking their lifestyles into account. The proposed system is based on a collaborative filtering recommendation technique. The PRS first classifies a collection of users as “highly responsive users” and “less responsive users” based on their DR degree analysis. Then, for each less responsive user, the PRS infers the user's lifestyle from usage profiles of nonshiftable appliances and finds out users who have similar habits with the target user from the set of highly responsive users. Based on this, the PRS evaluates the lifestyle similarity between the target user and each smart user, aggregates the appliance usage experiences of highly responsive users, and makes appliance-use recommendations to the target user. Experiments based on a residential data simulator “SimHouse” are designed to validate the proposed system.

Cooperative wind farm control is a complex problem due to wake effect, and it is hard to find the proper model. Reinforcement learning can find the optimal policy in a dynamic environment using “trial and error,” but may damage the machine and cause high cost during the learning process. In order to address this challenge, this article proposes the knowledge-assisted reinforcement learning framework by combining the low-fidelity analytical model with a reinforcement learning framework. Moreover, the knowledge-assisted deep deterministic policy gradient (KA-DDPG) algorithm and three kinds of knowledge-assisted learning methods are proposed based on the framework. The proposed methods are tested in nine different scenarios of WFSim. The simulation results show that the KA-DDPG algorithm can reach the maximum power output and ensure safety during learning. In addition, the learning cost is reduced by accelerating the learning process.

In this paper, we present the problem formulation and methodology framework of Super Resolution Perception (SRP) on smart meter data. With the widespread use of smart meters, a massive amount of electricity consumption data can be obtained. Smart meter data is the basis of automated billing and pricing, appliance identification, demand response, etc. However, the provision of high-quality data may be expensive in many cases. In this paper, we propose a novel problem - the SRP problem as reconstructing high-quality data from unsatisfactory data in smart grids. Advanced generative models are then proposed to solve the problem. This technology makes it possible for empowering existing facilities without upgrading existing meters or deploying additional meters. We first mathematically formulate the SRP problem under the Maximum a Posteriori (MAP) estimation framework. The dataset namely Super Resolution Perception Dataset (SRPD) is designed for this problem and released. A case study is then presented, which performs SRP on smart meter data. A network namely Super Resolution Perception Convolutional Neural Network (SRPCNN) is proposed to generate high-frequency load data from low-frequency data. Experiments demonstrate that our SRP models can reconstruct high-frequency data effectively. Moreover, the reconstructed high-frequency data can lead to better appliance identification results.

The prevalence of distributed energy resources encourages the concept of an electricity “Prosumer (Producer and Consumer)”. This paper proposes a distributed electricity trading system to facilitate the peer-to-peer electricity sharing among prosumers. The proposed system includes two layers. In the first layer, a multi-agent system is designed to support the prosumer network, and an agent coalition mechanism is proposed to enable the prosumers to form coalitions and negotiate electricity trading. In the second layer, a Blockchain based transaction settlement mechanism is proposed to enable the trusted and secure settlement of electricity trading transactions formed in the first layer. Simulations are conducted based on the java agent development environment to validate the proposed electricity trading process.

Cyber security of modern power systems has drawn increasing attention in recent years. This paper proposes a class of cyber-topology attacks derived from false data injection attacks, with the aim of disturbing the operation of power systems. Three kinds of cyber-topology attack are proposed: 1) line-addition attack; 2) line-removal attack; and 3) line-switching attack. By directly misleading the decision-making process of the independent system operator, the proposed cyber-topology attack consequently affects the economic operation and security of the system. We establish optimal attack models for different cyber-topology attack scenarios, and use a recently proposed metaheuristic optimization algorithm the natural aggregation algorithm to solve the resulting attack models. Experiments based on the IEEE 39-node benchmark system show that the proposed class of attacks poses a significant threat to modern power systems.

This paper proposes a generalized false data injection attack-based cyber topology attack capable of disturbing conventional transactions in the electricity market. The proposed attack aims to mislead customers pay higher electricity bills in the wholesale market by making small, undetected price deviations during each attack interval over an extended period. In this paper, the Australian electricity market trading mechanism is considered, in which an extended attack over a single day is proposed, spanning totally 288 5-minute dispatch (attack) intervals. The proposed attack poses no security threat to the power system, affecting only the economic costs borne by customers. A defining feature of the proposed model is the joint consideration of both topology error processing and bad data detection constraints in order to ensure validity of the proposed attack. The rolling horizon optimization technique is used to make appropriate adjustment to the attacker’s strategy considering the practical operating condition. A new metaheuristic optimization algorithm previously proposed by the authors, Natural aggregation algorithm, is used to find the corresponding attack strategies in this paper. Case studies on the IEEE-118 benchmark system are conducted to validate the proposed attack methodology.

Modern power systems are rapidly evolving into complex cyber-physical systems. The increasingly complex interaction among different energy entities calls for a secure, efficient, and robust cyber infrastructure. As an emerging distributed computing technology, Blockchain provides a secure environment to support such interactions. This paper gives a prospective on using Blockchain as a secure, distributed cyber infrastructure for the future grid. Firstly, the basic principles of Blockchain and its state-of-the-art are introduced. Then, a Blockchain based smart grid cyber-physical infrastructure model is proposed. Afterwards, some promising application domains of Blockchain in future grids are presented. Following this, some potential challenges are discussed.

Following the new round of electricity market reform, the spot market in Guangdong province has successfully entered the trial operation stage recently and is considered as the first pilot in China. In this paper, we review the performance of the trial operation of the electricity spot market in Guangdong. We discuss two main challenges encountered in Guangdong's market reform including the imbalanced funds caused by dual-track pricing and the compensation to gas power generators. The dual-track pricing refers to the two different settlement systems in the Guangdong trial market, one for the non-market electricity traded via the "priority power generation plan," and another traded via the spot market. While the dual-track pricing is inevitable since the Chinese government is accustomed to a gradual approach to economic reforms, it does lead to the serious problem of imbalanced funds which caused a lot of controversies. Furthermore, we discuss a case when a large number of renewable energy generators join the market. For the gas generator compensation, the market operator has three alternatives based on the present long-term transaction market. We employ a Hybrid Experimental Learning (HEL) approach to study which alternative is most in the interest of the gas generators.

With rapid advances in sensor, computer, and communication networks, modern power systems have become complicated cyber-physical systems. Assessing and enhancing cyber-physical system security is, therefore, of utmost importance for the future electricity grid. In a successful false data injection attack (FDIA), an attacker compromises measurements from grid sensors in such a way that undetected errors are introduced into estimates of state variables such as bus voltage angles and magnitudes. In evading detection by commonly employed residue-based bad data detection tests, FDIAs are capable of severely threatening power system security. Since the first published research on FDIAs in 2009, research into FDIA-based cyber-attacks has been extensive. This paper gives a comprehensive review of state-of-the-art in FDIAs against modern power systems. This paper first summarizes the theoretical basis of FDIAs, and then discusses both the physical and the economic impacts of a successful FDIA. This paper presents the basic defense strategies against FDIAs and discusses some potential future research directions in this field.

This paper proposes a cooperative control framework for the coordination of multiple microgrids. The framework is based on the multiagent system. The control framework aims to encourage the resource sharing among different autonomous microgrids and solve the energy imbalance problems by forming the microgrid coalition self-adaptively. First, the conceptual model of the integrated microgrids and the layered cooperative control framework is presented. Then, an advanced dynamic coalition formation scheme and corresponding negotiation algorithm are introduced to model the coordination behaviors of the microgrids. The proposed control framework is implemented by the Java Agent Development Framework. A loop distribution system with multiple interconnected microgrids is simulated, and the case studies are conducted to prove the efficiency of the proposed framework.

Smart home technology powered by cutting-edge intelligent algorithms creates a secure and comfortable living environment for electricity end users. In this paper, we propose an object surveillance algorithm and apply it to smart home environment. A combination usage of batch-normalized convolutional neural network (CNN) and data augmentation technique has been introduced for accomplishing object surveillance task by overseeing an object for long duration. Three aspects are considered including algorithm convergence, model creation, and dataset preparation. Firstly, mathematical demonstrations are offered to study the vanishing gradient phenomenon, and batch normalization is proposed for resolving this issue. Secondly, a revised version of AlexNet with batch normalization layer added after each convolution layer significantly boosts training speed and retains test set accuracy. Thirdly, only one picture taken on-premise is required for dataset preparation, and satisfactory predictive results are obtained.

The impact analysis of cyber-attacks on the modern power system is of great importance for researchers to grasp more precise information of adversaries, especially for contemplated cyber-attacks that may happen in the future. Different from studies based on existing proposed cyber-attacks, this paper proposes a new kind of cyber-attack which is launched on the demand side, namely, household level smart meter manipulation attack. It achieves attacking goals by manipulating household's smart meter measurements. And the attacking goals are diverse, either achieving huge economic losses to customers or causing security problems to the power grid. Instead of studying one specific model, this paper mainly focuses on the economic impact analysis for households to seek the relationship between the maximum electricity fee deviation and the tampering ratio by simulations. Six scenarios are proposed to simulate three kinds of situations: single household target, multiple household target, and whole society target. Conclusions and future works are also listed.

The false data injection attack (FDIA) is a form of cyber-attack capable of affecting the secure and economic operation of the smart grid. With DC model-based state estimation, this paper analyzes ways of constructing a successful attacking vector to fulfill specific targets, i.e., pre-specified state variable target and pre-specified meter target according to the adversary's willingness. The grid operator's historical reading experiences on meters are considered as a constraint for the adversary to avoid being detected. Also from the viewpoint of the adversary, we propose to take full advantage of the dual concept of the coefficients in the topology matrix to handle with the problem that the adversary has no access to some meters. Effectiveness of the proposed method is validated by numerical experiments on the IEEE-14 benchmark system.

This paper studies the coordinated scheduling of residential energy resources in a smart home environment. The particularity of this paper is to consider the uncertainties of the must-run appliance load demand forecast errors and to addresses the residential energy resource scheduling through a multi-objective optimization approach. Multiple 1-day must-run appliance power demand scenarios are firstly generated from the house’s historical energy consumption data. Based on this, a stochastic day-ahead appliance scheduling model is formulated, aiming to minimize the 1-day energy costs while maximizing the preference of the homeowner simultaneously. A new multi-objective optimization tool, i.e. Multi-Objective Natural Aggregation Algorithm (MONAA), is proposed to solve the stochastic day-ahead appliance scheduling model. Simulations are designed for the validation of the proposed method. Australian Research Council

Cyber-attacks pose significant hazards to the secure and economic operation of modern power systems. This paper proposes a composite false data injection attack (FDIA) and cyber topology attack capable of disturbing conventional transactions in the electricity market. The Australian electricity market mechanism is considered, wherein each trading period is comprised of six 5-minute security-constraint economic dispatch (SCED) intervals. In the proposed attack, the manipulation of cyber-topology information over the 30-minute horizon is considered, with the attack objective being the malicious increase in the costs borne by customers in the market. The proposed attack poses no security threat to the power system, affecting only the economic costs to customers. A defining feature of the proposed algorithm is the joint consideration of both topology error processing and bad data detection constraints in order to ensure validity of the proposed attack. The differential evolution (DE) algorithm is used to find the corresponding attack strategies. Case studies are conducted to validate the proposed attack methodology.

High speed growth in residential air-conditioning load requires increasing penetration of renewable energy to cater for peak load in current electricity networks. However, voltage over oscillation happens during high renewable penetration periods and peak load demand time. This paper attempts to provide a distributed control strategy to coordinate aggregated air-conditioning loads (ACLs) on network voltage regulation. Through limited communication between aggregators, required active power from ACLs are shared among participating groups, in the meanwhile, residents' thermal comfort is guaranteed. A novel two-parameter thermal model is introduced to calculate indoor air temperature more precisely. Case study and simulation results are given to demonstrate the effectiveness of proposed control scheme.

This paper considers two important challenges in the distribution network with high renewable penetrations. One is that during the peak load, generation is normally low or zero and it will cause the voltage drop. Meanwhile, during peak generation period, the generated power will exceed the load and be injected to the grid, which will subsequently cause the voltage rise. This paper proposes a MILP (Mixed Integer Linear Programming) approach to accommodate more wind generation in distribution networks. By coordinating the status of controllable load, battery energy storage system (BESS) and wind power, this paper proposes an optimal scheduling model for the distribution system to minimize the total costs from grid. Rolling optimization and weighting factor were implemented to obtain a good operation strategy under real-time operation. Detailed case studies are conducted to demonstrate the feasibility of the proposed method.

In a false data injection attack (FDIA), an adversary stealthily compromises measurements from electricity grid sensors in a coordinated fashion, with a view to evading detection by the power system bad data detection module. A successful FDIA can cause the system operator to perform control actions that compromise either the physical or economic operation of the power system. In this letter, we consider some implications for FDIAs arising from the late 2015 Ukraine Blackout event.

Static security assessment (SSA) is an important procedure to ensure the static security of the power system. Researches recently show that cyber-attacks might be a critical hazard to the secure and economic operations of the power system. In this paper, the influences of false data injection attack (FDIA) on the power system SSA are studied. FDIA is a major kind of cyber-attacks that can inject malicious data into meters, cause false state estimation results, and evade being detected by bad data detection. It is firstly shown that the SSA results could be manipulated by launching a successful FDIA, which can lead to incorrect or unnecessary corrective actions. Then, two kinds of targeted scenarios are proposed, i.e., fake secure signal attack and fake insecure signal attack. The former attack will deceive the system operator to believe that the system operates in a secure condition when it is actually not. The latter attack will deceive the system operator to make corrective actions, such as generator rescheduling, load shedding, etc. when it is unnecessary and costly. The implementation of the proposed analysis is validated with the IEEE-39 benchmark system.