33 results found
Zeng L, Qiu D, Sun M, 2022, Resilience enhancement of multi-agent reinforcement learning-based demand response against adversarial attacks, Applied Energy, Vol: 324, ISSN: 0306-2619
Demand response improves grid security by adjusting the flexibility of consumers meanwhile maintaining their demand–supply balance in real-time. With the large-scale deployment of distributed digital communication technologies and advanced metering infrastructures, data-driven approaches such as multi-agent reinforcement learning (MARL) are being widely employed to solve demand response problems. Nevertheless, the massive interaction of data inside and outside the demand response management system may lead to severe threats from the perspective of cyber-attacks. The cyber security requirements of MARL-based demand response problems are less discussed in the existing studies. To this end, this paper proposes a robust adversarial multi-agent reinforcement learning framework for demand response (RAMARL-DR) with an enhanced resilience against adversarial attacks. In particular, the proposed RAMARL-DR first constructs an adversary agent that aims to cause the worst-case performance via formulating an adversarial attack; and then adopts periodic alternating robust adversarial training scenarios with the optimal adversary aiming to diminish the severe impacts induced by adversarial attacks. Case studies are conducted based on an OpenAI Gym environment CityLearn, which provides a standard evaluation platform of MARL algorithms for demand response problems. Empirical results indicate that the MARL-based demand response management system is vulnerable when the adversary agent occurs, and its performance can be significantly improved after periodic alternating robust adversarial training. It can be found that the adversary agent can result in a 41.43% higher metric value of Ramping than the no adversary case, whereas the proposed RAMARL-DR can significantly enhance the system resilience with an approximately 38.85% reduction in the ramping of net demand.
Bi J, Luo F, He S, et al., 2022, False Data Injection- and Propagation-Aware Game Theoretical Approach for Microgrids, IEEE Transactions on Smart Grid, Vol: 13, Pages: 3342-3353, ISSN: 1949-3053
Defending microgrids against cyberattacks has been recognized as a significant task in modern energy systems. This paper studies a False Data Injection Attack (FDIA) scenario targeting microgrids, in which the attacker maliciously propagates false data malware in the communication network of a microgrid system, aiming at misleading the microgrid's operation. Firstly, this paper establishes the propagation and attack models for the false data malware; then, the impact of the FDIA is quantified through the formulation of a network-constrained microgrid dispatch model. Based on this, a false data malware propagation-aware game between the attacker and the defender is established and an effective computational approach is developed for solving the proposed game and obtaining the potential Nash equilibrium strategy pair. Extensive numerical simulations are conducted to validate the proposed framework.
Li B, Wan C, Luo F, et al., 2022, A bi-level transactive control model for integrating decision-making and DLMP-pricing in distribution networks, IET GENERATION TRANSMISSION & DISTRIBUTION, Vol: 16, Pages: 3814-3824, ISSN: 1751-8687
Lu R, Bai R, Luo Z, et al., 2022, Deep Reinforcement Learning-Based Demand Response for Smart Facilities Energy Management, IEEE Transactions on Industrial Electronics, Vol: 69, Pages: 8554-8565, ISSN: 0278-0046
This work proposes a novel deep reinforcement learning (DRL)-based demand response algorithm for smart facilities energy management to minimize electricity costs while maintaining a satisfaction index. Specifically, to accommodate the characteristics of the decision-making problem, long short-term memory (LSTM) units are adopted to extract discriminative features from past electricity price sequences and fed into fully connected multi-layer perceptrons (MLPs) with the measured energy and time information, then a deep Q-network is developed to approximate the optimal policy. After that, an experimental setup is constructed to investigate the effectiveness of the proposed DRL-based demand response algorithm to bridge the gap between theoretical studies and practical implementations. Numerical results demonstrate that the proposed algorithm can handle energy management well for multiple smart facilities. Moreover, the proposed algorithm outperforms the model predictive control (MPC) strategy and uncontrolled solution and is close to the theoretical optimal control method.
Zhang J, Wang Y, Sun M, et al., 2022, Two-Stage Bootstrap Sampling for Probabilistic Load Forecasting, IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, Vol: 69, Pages: 720-728, ISSN: 0018-9391
Wang Y, Jia M, Gao N, et al., 2022, Federated Clustering for Electricity Consumption Pattern Extraction, IEEE Transactions on Smart Grid, Vol: 13, Pages: 2425-2439, ISSN: 1949-3053
The wide popularity of smart meters enables the collection of massive amounts of fine-grained electricity consumption data. Extracting typical electricity consumption patterns from these data supports the retailers in their understanding of consumer behaviors. In this way, diversified services such as personalized price design and demand response targeting can be provided. Various clustering algorithms have been studied for electricity consumption pattern extraction. These methods have to be implemented in a centralized way, assuming that all smart meter data can be accessed. However, smart meter data may belong to different retailers or even consumers themselves who are not willing to share their data. In order to better protect the privacy of the smart meter data owners, this paper proposes two federated learning approaches for electricity consumption pattern extraction, where the k-means clustering algorithm can be trained in a distributed way based on two frequently used strategies, namely model-averaging and gradient-sharing. Numerical experiments on two real-world smart meter datasets are conducted to verify the effectiveness of the proposed method.
Huang DW, Luo F, Bi J, et al., 2022, An Efficient Hybrid IDS Deployment Architecture for Multi-Hop Clustered Wireless Sensor Networks, IEEE Transactions on Information Forensics and Security, Vol: 17, Pages: 2688-2702, ISSN: 1556-6013
Deploying Intrusion Detection Systems (IDSs) is an essential way to enhance the security of Multi-hop Clustered Wireless Sensor Networks (MCWSNs). The conventional IDS deployment architectural designs show limitations in ensuring the security of MCWSNs due to the limited monitoring range of the nodes. This paper proposes an efficient IDS deployment architecture for MCWSNs. The architecture is a hybrid design, in which the cluster heads and the sink collaboratively act as IDS agents to monitor the entire network and perform intrusion detection. Based on the new architecture, this paper proposes a resource allocation model to optimally allocate resources among the IDS agents so that the network's security metric can be maximized. A comprehensive analytical framework is proposed to analyze the optimality of the model's solution, and an efficient computational approach is developed to obtain the optimal resource allocation strategy. Extensive numerical simulation and comparison studies are conducted to validate the proposed method.
Bugaje AAB, Cremer JL, Sun M, et al., 2021, Selecting decision trees for power system security assessment, Energy and AI, Vol: 6
Power systems transport an increasing amount of electricity, and in the future, involve more distributed renewables and dynamic interactions of the equipment. The system response to disturbances must be secure and predictable to avoid power blackouts. The system response can be simulated in the time domain. However, this dynamic security assessment (DSA) is not computationally tractable in real-time. Particularly promising is to train decision trees (DTs) from machine learning as interpretable classifiers to predict whether the system-wide responses to disturbances are secure. In most research, selecting the best DT model focuses on predictive accuracy. However, it is insufficient to focus solely on predictive accuracy. Missed alarms and false alarms have drastically different costs, and as security assessment is a critical task, interpretability is crucial for operators. In this work, the multiple objectives of interpretability, varying costs, and accuracies are considered for DT model selection. We propose a rigorous workflow to select the best classifier. In addition, we present two graphical approaches for visual inspection to illustrate the selection sensitivity to probability and impacts of disturbances. We propose cost curves to inspect selection combining all three objectives for the first time. Case studies on the IEEE 68 bus system and the French system show that the proposed approach allows for better DT-selections, with an 80% increase in interpretability, 5% reduction in expected operating cost, while making almost zero accuracy compromises. The proposed approach scales well with larger systems and can be used for models beyond DTs. Hence, this work provides insights into criteria for model selection in a promising application for methods from artificial intelligence (AI).
Zhang T, Sun M, Cremer JL, et al., 2021, A Confidence-Aware Machine Learning Framework for Dynamic Security Assessment, IEEE TRANSACTIONS ON POWER SYSTEMS, Vol: 36, Pages: 3907-3920, ISSN: 0885-8950
Wang Y, Bennani IL, Liu X, et al., 2021, Electricity Consumer Characteristics Identification: A Federated Learning Approach, IEEE TRANSACTIONS ON SMART GRID, Vol: 12, Pages: 3637-3647, ISSN: 1949-3053
Cheng C, Ma G, Zhang Y, et al., 2020, A Deep Learning-Based Remaining Useful Life Prediction Approach for Bearings, IEEE-ASME TRANSACTIONS ON MECHATRONICS, Vol: 25, Pages: 1243-1254, ISSN: 1083-4435
Huyghues-Beaufond N, Tindemans S, Falugi P, et al., 2020, Robust and automatic data cleansing method for short-term load forecasting of distribution feeders, Applied Energy, Vol: 261, Pages: 1-17, ISSN: 0306-2619
Distribution networks are undergoing fundamental changes at medium voltage level. To support growing planning and control decision-making, the need for large numbers of short-term load forecasts has emerged. Data-driven modelling of medium voltage feeders can be affected by (1) data quality issues, namely, large gross errors and missing observations (2) the presence of structural breaks in the data due to occasional network reconfiguration and load transfers. The present work investigates and reports on the effects of advanced data cleansing techniques on forecast accuracy. A hybrid framework to detect and remove outliers in large datasets is proposed; this automatic procedure combines the Tukey labelling rule and the binary segmentation algorithm to cleanse data more efficiently, it is fast and easy to implement. Various approaches for missing value imputation are investigated, including unconditional mean, Hot Deck via k-nearest neighbour and Kalman smoothing. A combination of the automatic detection/removal of outliers and the imputation methods mentioned above are implemented to cleanse time series of 342 medium-voltage feeders. A nested rolling-origin-validation technique is used to evaluate the feed-forward deep neural network models. The proposed data cleansing framework efficiently removes outliers from the data, and the accuracy of forecasts is improved. It is found that Hot Deck (k-NN) imputation performs best in balancing the bias-variance trade-off for short-term forecasting.
Ye Y, Qiu D, Sun M, et al., 2020, Deep reinforcement learning for strategic bidding in electricity markets, IEEE Transactions on Smart Grid, Vol: 11, Pages: 1343-1355, ISSN: 1949-3053
Bi-level optimization and reinforcement learning (RL) constitute the state-of-the-art frameworks for modeling strategic bidding decisions in deregulated electricity markets. However, the former neglects the market participants' physical non-convex operating characteristics, while conventional RL methods require discretization of state and / or action spaces and thus suffer from the curse of dimensionality. This paper proposes a novel deep reinforcement learning (DRL) based methodology, combining a deep deterministic policy gradient (DDPG) method with a prioritized experience replay (PER) strategy. This approach sets up the problem in multi-dimensional continuous state and action spaces, enabling market participants to receive accurate feedback regarding the impact of their bidding decisions on the market clearing outcome, and devise more profitable bidding decisions by exploiting the entire action domain, also accounting for the effect of non-convex operating characteristics. Case studies demonstrate that the proposed methodology achieves a significantly higher profit than the alternative state-of-the-art methods, and exhibits a more favourable computational performance than benchmark RL methods due to the employment of the PER strategy.
Sun M, Zhang T, Wang Y, et al., 2020, Using Bayesian deep learning to capture uncertainty for residential net load forecasting, IEEE Transactions on Power Systems, Vol: 35, Pages: 188-201, ISSN: 0885-8950
Decarbonization of electricity systems drives significant and continued investments in distributed energy sources to support the cost-effective transition to low-carbon energy systems. However, the rapid integration of distributed photovoltaic (PV) generation presents great challenges in obtaining reliable and secure grid operations because of its limited visibility and intermittent nature. Under this reality, net load forecasting is facing unprecedented difficulty in answering the following question: how can we accurately predict the net load while capturing the massive uncertainties arising from distributed PV generation and load, especially in the context of high PV penetration? This paper proposes a novel probabilistic day-ahead net load forecasting method to capture both epistemic uncertainty and aleatoric uncertainty using Bayesian deep learning, which is a new field that combines Bayesian probability theory and deep learning. The proposed methodological framework employs clustering in subprofiles and considers residential rooftop PV outputs as input features to enhance the performance of aggregated net load forecasting. Numerical experiments have been carried out based on fine-grained smart meter data from the Australian grid with separately recorded measurements of rooftop PV generation and loads. The results demonstrate the superior performance of the proposed scheme compared with a series of state-of-theart methods and indicate the importance and effectiveness of subprofile clustering and high PV visibility.
Sun M, Wang Y, Teng F, et al., 2019, Clustering-based residential baseline estimation: a probabilistic perspective, IEEE Transactions on Smart Grid, Vol: 10, Pages: 6014-6028, ISSN: 1949-3061
Demand Response (DR) is one of the most cost-effective solutions for providing flexibility to power systems. The extensive deployment of DR trials and the roll-out of smart meters enable the quantification of consumer responsiveness to price signals via baseline estimation. The traditional deterministic baseline estimation approach can provide only a single value without consideration of uncertainty. This paper proposes a novel probabilistic baseline estimation framework that consists of a daily load profile pool construction stage, a deep learning-based clustering stage, an optimal cluster selection stage, and a quantile regression forests model construction stage. In particular, the concept of a daily load profile pool is introduced, and a deep-learning-based clustering approach is employed to handle a large number of daily patterns to further improve the baseline estimation performance. Case studies have been conducted on fine-grained smart meter data collected from a real dynamic time-of-use (dTOU) tariffs trial of the Low Carbon London (LCL) project. The superior performance of the proposed method is demonstrated based on a series of evaluation metrics regarding both deterministic and probabilistic estimation results.
Sun M, Djapic P, Aunedi M, et al., 2019, Benefits of smart control of hybrid heat pumps: an analysis of field trial data, Applied Energy, Vol: 247, Pages: 525-536, ISSN: 0306-2619
Smart hybrid heat pumps have the capability to perform smart switching between electricity and gas by employing a fully-optimized control technology with predictive demand-side management to automatically use the most cost-effective heating mode across time. This enables a mechanism for delivering flexible demand-side response in a domestic setting. This paper conducts a comprehensive analysis of the fine-grained data collected during the world’s first sizable field trial of smart hybrid heat pumps to present the benefits of the smart control technology. More specifically, a novel flexibility quantification framework is proposed to estimate the capability of heat pump demand shifting based on preheating. Within the proposed framework, accurate estimation of baseline heat demand during the days with interventions is fundamentally critical for understanding the effectiveness of smart control. Furthermore, diversity of heat pump demand is quantified across different numbers of households as an important input into electricity distribution network planning. Finally, the observed values of the Coefficient of Performance (COP) have been analyzed to demonstrate that the smart control can optimize the heat pump operation while taking into account a variety of parameters including the heat pump output water temperature, therefore delivering higher average COP values by maximizing the operating efficiency of the heat pump. Finally, the results of the whole-system assessment of smart hybrid heat pumps demonstrate that the system value of smart control is between 2.1 and 5.3 £ bn/year.
Sun M, Strbac G, Djapic P, et al., 2019, Preheating quantification for smart hybrid heat pumps considering uncertainty, IEEE Transactions on Industrial Informatics, Vol: 15, Pages: 4753-4763, ISSN: 1551-3203
The deployment of smart hybrid heat pumps can introduce considerable benefits to electricity systems via smart switching between electricity and gas while minimizing the total heating cost for each individual customer. In particular, the fully-optimized control technology can provide flexible heat that redistributes the heat demand across time for improving the utilization of low-carbon generation and enhancing the overall energy efficiency of the heating system. To this end, accurate quantification of preheating is of great importance to characterize the flexible heat. This paper proposes a novel data-driven preheating quantification method to estimate the capability of heat pump demand shifting and isolate the effect of interventions. Varieties of fine-grained data from a real-world trial are exploited to estimate the baseline heat demand using Bayesian deep learning while jointly considering epistemic and aleatoric uncertainties. A comprehensive range of case studies are carried out to demonstrate the superior performance of the proposed quantification method and then, the estimated demand shift is used as an input into the whole-system model to investigate the system implications and quantify the range of benefits of rolling-out the smart hybrid heat pumps developed by PassivSystems to the future GB electricity systems.
Wang Y, Chen Q, Zhang N, et al., 2019, Fusion of the 5G Communication and the Ubiquitous Electric Internet of Things: Application Analysis and Research Prospects, Dianwang Jishu/Power System Technology, Vol: 43, Pages: 1575-1585, ISSN: 1000-3673
The ubiquitous electric Internet of Things (IoT) is a concrete manifestation of the IoT in the power industry. It is a deep integration of interconnected power network and communication network, and an important measure to implement the Energy Internet. The fifth generation mobile communication (5G communication) is favored by all walks of life because of its advantages of high bandwidth, low delay and low power consumption. It will also play an important role in the ubiquitous electric IoT. This paper discusses the deep fusion of the 5G communication technology and the ubiquitous electric IoT. Specifically, the potential application scenarios of 5G communication in ubiquitous electric IoT are analyzed. In addition, the key technologies of the 5G communication that support ubiquitous electric IoT are also summarized. Since the base stations of 5G communication network are dense and the energy consumption is large in the future, how ubiquitous electric IoT supports 5G communication network and the coordinated interaction between the two networks are also studied. Finally, future research on the fusion of the 5G communication technology and the ubiquitous electric IoT are prospected.
Wang Y, Gan D, Sun M, et al., 2019, Probabilistic individual load forecasting using pinball loss guided LSTM, Applied Energy, Vol: 235, Pages: 10-20, ISSN: 0306-2619
The installation of smart meters enables the collection of massive fine-grained electricity consumption data and makes individual consumer level load forecasting possible. Compared to aggregated loads, load forecasting for individual consumers is prone to non-stationary and stochastic features. In this paper, a probabilistic load forecasting method for individual consumers is proposed to handle the variability and uncertainty of future load profiles. Specifically, a deep neural network, long short-term memory (LSTM), is used to model both the long-term and short-term dependencies within the load profiles. Pinball loss, instead of the mean square error (MSE), is used to guide the training of the parameters. In this way, traditional LSTM-based point forecasting is extended to probabilistic forecasting in the form of quantiles. Numerical experiments are conducted on an open dataset from Ireland. Forecasting for both residential and commercial consumers is tested. Results show that the proposed method has superior performance over traditional methods.
Sun M, Wang Y, Strbac G, et al., 2019, Probabilistic peak load estimation in smart cities using smart meter data, IEEE Transactions on Industrial Electronics, Vol: 66, Pages: 1608-1618, ISSN: 0278-0046
Adequate capacity planning of substationsand feeders primarily depends on an accurate estimationof the future peak electricity demand. Traditional coinci-dent peak demand estimation is carried out based on theempirical metric, after diversity maximum demand (ADMD),indicating individual peak consumption levels and of de-mand diversification across multiple residents. With theprivilege of smart meters in smart cities, this paper pro-poses a data-driven probabilistic peak demand estima-tion framework using fine-grained smart meter data andsocio-demographic data of the consumers, which drivefundamental electricity consumptions across different cat-egories. In particular, four main stages are integrated inthe proposed approach: load modeling and sampling viathe proposed variable truncated R-vine copulas (VTRC)method; correlation-based customer grouping; probabilis-tic normalized maximum diversified demand (NMDD) esti-mation; and probabilistic peak demand estimation for newcustomers. Numerical experiments have been conductedon real demand measurements across 2,639 households inLondon, collected from Low Carbon London (LCL) projectssmart-metering trial. The mean absolute percentage error(MAPE) and pinball loss function are used to quantitativelydemonstrate the superiority of the proposed approach interms of the point estimate value and the probabilisticresult, respectively.
Sun M, Teng F, Zhang X, et al., 2019, Data-driven representative day selection for investment decisions: a cost-oriented approach, IEEE Transactions on Power Systems, Vol: 34, Pages: 2925-2936, ISSN: 0885-8950
Power system investment planning problems become intractable due to the vast variability that characterizes system operation and the increasing complexity of the optimization model to capture the characteristics of renewable energy sources (RES). In this context, making optimal investment decisions by considering every operating period is unrealistic and inefficient. The conventional solution to address this computational issue is to select a limited number of representative operating periods by clustering the input demand-generation patterns while preserving the key statistical features of the original population. However, for an investment model that contains highly complex nonlinear relationship between input data and optimal investment decisions, selecting representative periods by relying on only input data becomes inefficient. This paper proposes a novel investment costoriented representative day selection framework for large scale multi-spacial investment problems, which performs clustering directly based on the investment decisions for each generation technology at each location associated with each individual day. Additionally, dimensionality reduction is performed to ensure that the proposed method is feasible for large-scale power systems and high-resolution input data. The superior performance of the proposed method is demonstrated through a series of case studies with different levels of modeling complexity.
Konstantelos I, Sun M, Tindemans S, et al., 2019, Using vine copulas to generate representative system states for machine learning, IEEE Transactions on Power Systems, Vol: 34, Pages: 225-235, ISSN: 0885-8950
The increasing uncertainty that surrounds electricity system operation renders security assessment a highly challenging task; the range of possible operating states expands, rendering traditional approaches based on heuristic practices and ad hoc analysis obsolete. In turn, machine learning can be used to construct surrogate models approximating the system's security boundary in the region of operation. To this end, past system history can be useful for generating anticipated system states suitable for training. However, inferring the underlying data model, to allow high-density sampling, is problematic due to the large number of variables, their complex marginal probability distributions and the non-linear dependence structure they exhibit. In this paper we adopt the C-Vine pair-copula decomposition scheme; clustering and principal component transformation stages are introduced, followed by a truncation to the pairwise dependency modelling, enabling efficient fitting and sampling of large datasets. Using measurements from the French grid, we show that a machine learning training database sampled from the proposed method can produce binary security classifiers with superior predictive capability compared to other approaches.
Sun M, Cremer J, Strbac G, 2018, A novel data-driven scenario generation framework for transmission expansion planning with high renewable energy penetration, Applied Energy, Vol: 228, Pages: 546-555, ISSN: 0306-2619
Transmission expansion planning (TEP) is facing unprecedented challenges with the rise of integrated renewable energy resources (RES), flexible load elements, and the potential electrification of transport and heat sectors. Under this reality, the inadequate information of the stochastic parameters’ behavior may lead to inefficient expansion decisions, especially in the context of very high renewable penetration. This paper proposes a novel data-driven scenario generation framework for the TEP problem to generate unseen but important load and wind power scenarios while capturing inter-spatial dependencies between loads and wind generation units’ output in various locations, using a vine-copula based high-dimensional stochastic variable modeling approach. The superior performance of the proposed model is demonstrated through a case study on a modified IEEE 118-bus system. The expected result of using the expected value problem solution (EEV) and the net benefits of transmission expansion (NBTE) are used as the evaluation metrics to quantitatively illustrate the advantages of the proposed approach. In addition, the case of very high wind penetration is carried out to further highlight the importance of the multivariate stochastic dependence of load and wind power generation. The results demonstrate that the proposed scenario generation method can result in near-optimal investment decisions for the TEP problem that make more net benefits than using limited number of historical data.
Sun M, Konstantelos I, Strbac G, 2018, A Deep Learning-Based Feature Extraction Framework for System Security Assessment, IEEE Transactions on Smart Grid, Vol: 10, Pages: 5007-5020, ISSN: 1949-3053
The ongoing decarbonisation of modern electricity systems has led to a substantial increase of operational uncertainty, particularly due to the large-scale integration of renewable energy generation. However, the expanding space of possible operating points renders necessary the development of novel security assessment approaches. In this paper we focus on the use of security rules where classifiers are trained offline to characterize previously unseen points as safe or unsafe. This paper proposes a novel deep learning-based feature extraction framework for building security rules. We show how deep autoencoders can be used to transform the space of conventional state variables (e.g., power flows) to a small number of dimensions where we can optimally distinguish between safe and unsafe operation. The proposed framework is data-driven and can be useful in multiple applications within the context of security assessment. To achieve high accuracy, a novel objective-based loss function is proposed to address the issue of imbalanced safe/unsafe classes that characterize electricity system operation. Furthermore, an R-vine copula-based model is proposed to sample historical data and generate large populations of anticipated system states for training. The superior performance of the proposed framework is demonstrated through a series of case studies and comparisons using the load and wind generation data from the French transmission system, which have been mapped to the IEEE 118-bus system.
Wang Y, Qixin C, Sun M, et al., 2018, An ensemble forecasting method for the aggregated load with sub profiles, IEEE Transactions on Smart Grid, Vol: 9, ISSN: 1949-3061
With the prevalence of smart meters, fine-grained subprofiles reveal more information about the aggregated load and further help improve the forecasting accuracy. Ensemble is an effective approach for load forecasting. It either generates multiple training datasets or applies multiple forecasting models to produce multiple forecasts. In this letter, a novel ensemble method is proposed to forecast the aggregated load with subprofiles where the multiple forecasts are produced by different groupings of subprofiles. Specifically, the subprofiles are first clustered into different groups and forecasting is conducted on the grouped load profiles individually. Thus, these forecasts can be summed to form the aggregated load forecast. In this way, different aggregated load forecasts can be obtained by varying the number of clusters. Finally, an optimal weighted ensemble approach is employed to combine these forecasts and provide the final forecasting result. Case studies are conducted on two open datasets and verify the effectiveness and superiority of the proposed method.
Sun M, Teng F, Konstantelos I, et al., 2018, An objective-based scenario selection method for transmission network expansion planning with multivariate stochasticity in load and renewable energy sources, Energy, Vol: 145, Pages: 871-885, ISSN: 0360-5442
Transmission Network Expansion Planning (TNEP) in modern electricity systems is carried out on a cost-benefit analysis basis; the planner identifies investments that maximize the social welfare. As the integration of Renewable Energy Sources (RES) increases, there is a real challenge to accurately capture the vast variability that characterizes system operation within a planning problem. Conventional approaches that rely on a large number of scenarios for representing the variability of operating points can quickly lead to computational issues. An alternative approach that is becoming increasingly necessary is to select representative scenarios from the original population via clustering techniques. However, direct clustering of operating points in the input domain may not capture characteristics which are important for investment decision-making. This paper presents a novel objective-based scenario selection framework for TNEP to obtain optimal investment decisions with a significantly reduced number of operating states. Different clustering frameworks, clustering variable s and clustering techniques are compared to determine the most appropriate approach. The superior performance of the proposed framework is demonstrated through a case study on a modified IEEE 118-bus system.
Zhang J, Wang Y, Sun M, et al., 2018, Constructing probabilistic load forecast from multiple point forecasts: a bootstrap based approach, IEEE PES ISGT Asia 2018, Publisher: IEEE
Probabilistic load forecast presents more informa-tion on the possible deviation of forecast than the point forecast.There are sufficient regression models that can make pointforecasts. An intuitive question can be raised:Is there a wayto combine the point forecasts to construct a probability or intervalforecast?In this paper, a bootstrap based ensemble approach isput forward to construct forecast intervals from multiple pointforecasts. Specifically, multiple point forecasting models are firsttrained based on the bootstrap sampled training datasets anddifferent forecasting models. Then, bootstrap is applied again tothe multiple point forecasts. Finally, the quantiles are estimatedaccording to the distribution of the sampled point forecasts.Two common machine learning methods, random forest (RF)and gradient boosting regression tree (GBRT), are combinedto test the feasibility of the proposed forecasting framework.Compared with quantile RF (Q-RF) and quantile GBRT (Q-GBRT), numerical experiments demonstrate its advantage overQ-RF and Q-GBRT.
Yang Y, Hao J, Sun M, et al., 2018, Deep Multiagent Q-Learning for Autonomous Agents in Future Smart Grid, Goran Strbac
Sun M, Konstantelos I, Strbac G, 2016, Transmission network expansion planning with stochastic multivariate load and wind modeling, PMAPS 2016, Publisher: IEEE
The increasing penetration of intermittent energy sources along with the introduction of shiftable load elements renders transmission network expansion planning (TNEP) a challenging task. In particular, the ever-expanding spectrum of possible operating points necessitates the consideration of a very large number of scenarios within a cost-benefit framework, leading to computational issues. On the other hand, failure to adequately capture the behavior of stochastic parameters can lead to inefficient expansion plans. This paper proposes a novel TNEP framework that accommodates multiple sources of operational stochasticity. Inter-spatial dependencies between loads in various locations and intermittent generation units' output are captured by using a multivariate Gaussian copula. This statistical model forms the basis of a Monte Carlo analysis framework for exploring the uncertainty state-space. Benders decomposition is applied to efficiently split the investment and operation problems. The advantages of the proposed model are demonstrated through a case study on the IEEE 118-bus system. By evaluating the confidence interval of the optimality gap, the advantages of the proposed approach over conventional techniques are clearly demonstrated.
Sun M, Konstantelos I, Strbac G, 2016, C-Vine copula mixture model for clustering of residential electrical load pattern data, IEEE Transactions on Power Systems, Vol: 32, Pages: 2382-2393, ISSN: 0885-8950
The ongoing deployment of residential smart meters in numerous jurisdictions has led to an influx of electricity consumption data. This information presents a valuable opportunity to suppliers for better understanding their customer base and designing more effective tariff structures. In the past, various clustering methods have been proposed for meaningful customer partitioning. This paper presents a novel finite mixture modeling framework based on C-vine copulas (CVMM) for carrying out consumer categorization. The superiority of the proposed framework lies in the great flexibility of pair copulas towards identifying multi-dimensional dependency structures present in load profiling data. CVMM is compared to other classical methods by using real demand measurements recorded across 2,613 households in a London smart-metering trial. The superior performance of the proposed approach is demonstrated by analyzing four validity indicators. In addition, a decision tree classification module for partitioning new consumers is developed and the improved predictive performance of CVMM compared to existing methods is highlighted. Further case studies are carried out based on different loading conditions and different sets of large numbers of households to demonstrate the advantages and to test the scalability of the proposed method.
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.