Search

The minimum output variance filtered reference least mean square (MOV-FxLMS algorithm is a effective algorithm that utilizes the penalty mechanism to help the active noise control (ANC system achieve noise cancellation with constrained output variance or power. As it can constrain output power, the MOV-FxLMS algorithm can freely determine the ANC system's control effort, avoid output saturation, and improve system stability. However, its performance is determined by a penalty factor, which is normally chosen by trial and error. Hence, this work proposes an optimal penalty factor and its feasible estimation that does not require any assumptions of Gaussian reference signal or input independence. This factor assists MOV-FxLMS in achieving the optimal solution under the target output-variance constraint. Numerical simulations on measure paths demonstrate its effectiveness for various types of noise.

Active noise control (ANC) headphone is widely used to attenuate the noise around human' s ear. The microphone mounted on the conventional ANC headphones collected the mixed reference signals when more than one noise sources are often present in the surrounding. In this case, the uncorrelated noise sources involved in the mixed reference usually deteriorate the noise reduction performance of the ANC headphones. To solve this problem, wireless microphones are proposed to install close to each potential noise source in the environment. The microphones pick up the clean reference signals and transmit them to the ANC controller embedded in the headphones with time-advance wirelessly. Every reference signal selected by a coherence-based-selection algorithm is provided individual control filter in each ear. Each control filter updated by using a single clean reference offers better noise reduction performance for ANC headphones. Furthermore, numerical simulations and real-time experiment results in this paper demonstrate the improvement of the proposed method compared with conventional ANC headphones.

Hierarchical state estimation (HSE) is often deployed to evaluate the states of an interconnected power system from telemetered measurements. By HSE, each low-level control center (LCC) takes charge of the estimation of its internal states, whereas a trusted high-level control center (HCC) assumes the coordination of boundary states. However, a trusted HCC may not always exist in practice; a cloud server can take the role of an HCC in case no such facility is available. Since it is prohibited to release sensitive power grid data to untrustworthy cloud environments, considerations need to be given to avoid breaches of LCCs’ privacy when outsourcing the coordination tasks to the cloud server. To this end, this article proposes a privacy-preserving HSE framework, which rearranges the regular HSE procedure to integrate a degree-2 variant of the Thresholded Paillier Cryptosystem (D2TPC). Attributed to D2TPC, computations by the cloud-based HCC can be conducted entirely in the ciphertext space. Even if the HCC and some LCCs conspire together to share the information they have, the privacy of non-conspiring LCCs is still assured. Experiments on various scales of test systems demonstrate a high level of accuracy, efficiency, and scalability of the proposed framework.

This paper presents a hybrid metaheuristic optimizer called CHDJ to identify accurate and reliable values for the model parameters of solid oxide fuel cells (SOFC). CHDJ is a hybridizer of differential evolution (DE) and Jaya algorithm. The core of CHDJ lies in a developed competitive hybrid operation. In this operation, three vectors including the DE mutant vector, the Jaya mutant vector, and the crossover vector that constituted by the foregoing two mutant vectors form a tri-competitive mechanism together. This mechanism can contribute to exploring the global space extensively and exploiting the local region meticulously to generate promising candidate solutions. To verify the performance of CHDJ, two SOFC models including a cylindrical cell and a tubular stack operating at different pressures and temperatures are employed. Experimental results demonstrate that these three vectors interact and cooperate organically. They make CHDJ take full advantages of the exploration of DE and the exploitation of Jaya effectively. The optimal values of the mean square error (MSE) between the measured voltage and the calculated voltage obtained by CHDJ on the cylindrical cell at 1073K, 1173K, 1213K, and 1273K are 2.9215E-06, 1.4610E-06, 1.9093E-06, and 2.4682E-06, respectively. On the tubular stack, for the operating conditions of 3atm constant pressure at 873K, 923K, 973K, 1023K, and 1073K, the optimal MSE values are 2.7489E-05, 1.6169E-04, 6.9207E-04, 1.4833E-03, and 1.5876E-03, respectively. For the operating conditions of 1073K constant temperature under 1atm, 2atm, 3atm, 4atm, and 5atm, the results are 1.5293E-03, 1.6289E-03, 1.5876E-03, 1.5868E-03, and 1.5870E-03, respectively. Comparisons with other algorithms show that CHDJ achieves better overall performance. It can skip local optima and find optimal solutions, and thus can serve as a hopeful candidate method for parameter identification of SOFC models.

Output saturation, mainly caused by the power amplifier, is a critical issue influencing the performance and stability of an adaptive system, such as in active noise control. In this paper, a quadratically constrained quadratic program (QCQP) is defined to achieve optimal control under the averaging-output-power constraint, which ensures the output of the system operates linearly and hence, avoids the output saturation. To solve this QCQP problem recursively in practice, this paper utilizes one of the leaky-based filtered-x least mean square algorithm with an optimal leak factor. However, this method only can be applied when the statistical feature of the control signal with maximum output-power is known, which is difficult to obtain in practice. Hence, by incorporating the adaptive inverse modeling technique, we can derive a practical estimation of the optimal leaky factor, which is applicable to different noise types. Furthermore, as the optimal output-constraint control forces the output to operate linearly, the nonlinear amplifier model is not required for the leak factor estimate. The simulation of the proposed algorithm is carried out on measured nonlinear paths to validate its efficacy.

Leveraging multiple heterogeneous measurements to predict wind power has long been a challenging task in the electrical community. In this paper, a deep architecture incorporated with multitask learning and multimodal learning for wind power prediction, termed predictive stacked autoencoder (PSAE), is presented. PSAE is a unified framework integrating multiple stacked autoencoders (SAEs), one feature fusion layer, and one prediction terminal layer, which expands the architecture from two spatial dimensions, including the depth and width, compared to conventional prediction models. Initially, the SAEs at the bottom of PSAE extracted features from multiple kinds of measurements respectively. Following, the feature fusion layer encodes the high-order features extracted by different SAEs into a unified feature that is more informative and representative for wind power prediction. Finally, the prediction terminal layer functions as a regression machine which generates the predicted targets based on the fusion features. Trained in an end-to-end (E2E) manner, PSAE is capable of learning heterogeneous features jointly and achieving the prediction task of sequence-to-sequence (S2S). Experiments for multi-step short-term predictions are conducted on real-world data, and the results demonstrate the superiority of PSAE to prior methods.

Shutting the window is usually the last resort in mitigating environmental noise, at the expense of natural ventilation. We describe an active sound control system fitted onto the opening of the domestic window that attenuates the incident sound, achieving a global reduction in the room interior while maintaining natural ventilation. The incident sound is actively attenuated by an array of control modules (a small loudspeaker) distributed optimally across the aperture. A single reference microphone provides advance information for the controller to compute the anti-noise signal input to the loudspeakers in real-time. A numerical analysis revealed that the maximum active attenuation potential outperforms the perfect acoustic insulation provided by a fully shut single-glazed window in ideal conditions. To determine the real-world performance of such an active control system, an experimental system is realized in the aperture of a full-sized window installed on a mockup room. Up to 10-dB reduction in energy-averaged sound pressure level was achieved by the active control system in the presence of a recorded real-world broadband noise. However, attenuation in the low-frequency range and its maximum power output is limited by the size of the loudspeakers. Ministry of National Development (MND) National Research Foundation (NRF) Published version This research is supported by the Singapore Ministry of National Development and the National Research Foundation, Prime Minister’s Office under the Cities of Tomorrow (CoT) Research Programme (CoT Award No. COT-V4-2019-1). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of the Singapore Ministry of National Development and National Research Foundation, Prime Minister’s Office, Singapore.

Multichannel active noise control (MCANC) is widely recognized as an effective and efficient solution for acoustic noise and vibration cancellation, such as in high-dimensional ventilation ducts, open windows, and mechanical structures. The feedforward multichannel filtered-x least mean square (FFMCFxLMS) algorithm is commonly used to dynamically adjust the transfer function of the multichannel controllers for different noise environments. The computational load incurred by the FFMCFxLMS algorithm, however, increases exponentially with increasing channel count, thus requiring high-end field-programmable gate array (FPGA) processors. Nevertheless, such processors still need specific configurations to cope with soaring computing loads as the channel count increases. To achieve a high-efficiency implementation of the FFMCFxLMS algorithm with floating-point arithmetic, a novel architecture based on multiple-parallel-branch with folding (MPBF) technique is proposed. This architecture parallelizes the branches and reuses the multiplier and adder in each folded branch so that the tradeoff between throughput and the usage of the hardware resources is balanced. The proposed architecture is validated in an experimental setup that implements the FFMCFxLMS algorithm for the MCANC system with 24 reference sensors, 24 secondary sources, and 24 error sensors, at a sampling and throughput rates of 25 kHz and 260 Mb/s, respectively. NRF (Natl Research Foundation, S’pore) Accepted version

The parameter extraction problem of solar photovoltaic (PV) models is a highly nonlinear multimodal optimization problem. In this paper, quadratic interpolation learning differential evolution (QILDE) is proposed to solve it. Differential evolution (DE) is a preeminent metaheuristic algorithm with good exploration. However, its exploitation is poor, resulting in low searching precision when applied to the problem. To overcome this deficiency, in QILDE, quadratic interpolation (QI) is embedded in the crossover operation of DE to construct a QI learning-backup crossover operation to enhance the performance of DE. The mutation scheme of DE is primarily responsible for exploring the new search space while QI is mainly in charge of exploiting the local solution space around the best individual, which, therefore, can achieve a good trade-off between exploitation and exploration. QILDE is applied to six different PV cases. The experimental results demonstrate that QI coupled with the mutation scheme DE/best/2 can obtain superior results in solving the parameter extraction problem of PV models. Besides, compared with other advanced algorithms, QILDE shows highly competitive performance in terms of solution quality, extraction accuracy, robust stability, convergence property, computational time, and statistical significance. In addition, the current–voltage characteristics provided by QILDE agree well with the measured data for different PV models under different operating conditions.

Leveraging both temporal and spatial correlations to predict wind speed remains one of the most challenging and less studied areas of wind speed prediction. In this paper, the problem of predicting wind speeds for multiple sites is investigated by using the spatio-temporal correlation. We proposed a deep architecture termed predictive spatio-temporal network (PSTN), which is a unified framework integrating a convolutional neural network (CNN) and a long short-term memory (LSTM). Initially, the spatial features are extracted from the spatial wind speed matrices by the CNN at the bottom of the model. Then, the LSTM captures the temporal dependencies among the spatial features extracted from contiguous time points. Finally, the predicted wind speeds are given by the last state of the top layer of the LSTM, which are generated by using the spatial features and temporal dependencies. Though composed of two kinds of architectures, PSTN is trained with one loss function in an end-to-end manner, which can learn temporal and spatial correlations jointly. Experiments for short-term predictions are conducted on real-world data, whose results demonstrate that PSTN outperforms prior methods.

Phasor measurement unit (PMU) data manipulation attacks (PDMAs) may blind the control centers to the real-time operating conditions of power systems. Detecting these attacks accurately is essential to ensure the normal operation of power system monitoring and control. Using long-term accumulated historical PMU measurements to train a machine learning model to detect PDMAs has shown promising results. In this paper, deep-autoencoder-based anomaly measurers are deployed throughout the power system to build a distributed PDMA detection framework. The architecture of a deep autoencoder and its training process are introduced. How to convert the historical PMU measurements into data samples for learning is also elaborated upon. Once trained, an anomaly measurer can assess the PDMA existence possibility of the new PMU measurements. By integrating the results of different anomaly measurers, the proposed distributed PDMA detection framework can detect PDMAs in the whole power system. The effectiveness and detection performance of the framework are discussed through experiments.

This paper presents an evaluation method of strength measure for the static voltage stability analysis of the hybrid multi-infeed dc (HMIdc) system, which comprises both the line-commutated converter-based high voltage dc (LCC-HVdc) and voltage source converter-based HVdc (VSC-HVdc) links. For this purpose, the coupled single-port model proposed originally to assess the static voltage stability of ac power systems is first extended in the HMIdc system. In particular, how the LCC-HVdc and VSC-HVdc links should be modeled in the extended coupled single-port model is further investigated. Then, a generalized effective short-circuit ratio (GESCR) index is proposed as a strength measure for the static voltage stability analysis of the HMIdc system using the extended coupled single-port model. The GESCR can resolve the ambiguity encountered by earlier indices arising from empirical reasoning; thus, it is able to indicate the static voltage stability of the HMIdc system more accurately. Finally, the simulated results based on a classical simplified HMIdc system and a modified IEEE 118-bus system, both including two LCC-HVdc links and a VSC-HVdc link, validate the proposed GESCR.

The leaky filtered-input least mean square (LFxLMS) algorithm is widely used in active noise control applications to minimize the degradation of attenuation performance due to output saturation distortion. However, the leak factor, which is critical in determining the steady-state error and robustness of the algorithm, is usually selected through trial and error. This letter proposes a leak factor selection approach, which ensures the LFxLMS algorithm converges to its optimal solution under the average-output-power constraint and can be readily derived in practice. Both broadband and narrowband cases are considered in the derivation without the independence assumption, and the simulations are conducted based on real primary and secondary paths to verify its effectiveness. Accepted version

This study presents a novel faulted line identification method that is applicable in power systems with only sparse synchronised measurements, and at the same time, with significantly reduced computational burden so that it can be a competitive choice for wide-area backup protection schemes. By revealing the behaviour pattern of bus voltages from the perspective of high-dimensional unitary space, the method is enabled to avoid unnecessary computation and thus to remarkably improve its efficiency. Analysis in high-dimensional unitary space also gives a direct and feasible guideline for phasor measurement unit (PMU) placement, from which a PMU optimal placement strategy is developed. Performed on the IEEE benchmark systems of 39 and 118 buses, the proposed method gives encouraging results.

Feedforward active noise control (ANC) is widely utilized to attenuate the broadband noise picked up by the reference microphone. However, in some situations, it is impractical to obtain a clean reference signal when the noise source is far away from the controller. Hence, we adopt a wireless reference microphone to pick up the reference signals around the noise sources. Furthermore, a coherence-based-selection algorithm is proposed to select the reference signals with high coherence. The proposed method improves the quality of the reference signals and the noise reduction performance of the ANC system. Numerical simulations and real-time experiments are conducted to validate the effectiveness of the proposed algorithm.

The growing integration of information and communication technology exposes more exploitable cyber-attack vectors to electric power systems. Measurement manipulation is an impactive type of cyber-attacks that can be launched to cause massive damages by interfering decision-making of control centers. Existing data-driven measurement manipulation detection models often bear criticisms due to the deficient mining of spatial-temporal correlations. In this paper, based on the Gramian Angular Summation Field (GASF) algorithm, various types of phasor time-series measured at different locations are transformed into a series of generalized multichannel images. From these generalized images, Convolutional Neural Networks (CNNs) are trained to discover suspicious patterns. GASF can visualize the temporal features of a single time-series and facilitate collaborative CNNs to figure out spatial-temporal correlations between different measurements. The improvements of the proposed framework in the detection performance are demonstrated through case studies.

Fast and accurate transient stability assessment (TSA) is significant for power system. Topological information is often ignored by some machine learning methods, which may have a profound influence for TSA. For the purpose of improving the assessment accuracy and the adaptability to topology changes for online TSA, graph convolutional network (GCN) is introduced. GCN can aggregate node features and topological information. A jumping knowledge layer is added to prevent overfitting by concatenating the output of every GCN layer. Steady-state features of power system that are got through state estimation from the SCADA system are employed as the input. Tests on IEEE 39-Bus system indicate that the GCN model has a higher performance than some other methods. Several system topology changes are considered, and good adaptability to different topologies is indicated.

The goal of optimal dynamic topology evolution is to establish a distribution network which accommodates a high penetration of renewable energy with least cost. In this paper, the evolution process is formulated as a bi-level optimization model. The upper-level model optimizes the location and size of renewable distributed generation (RDG) as well as flexible resources and aims to minimize the total cost of evolution. Besides, out of simplification, the upper-level model discretizes the continues evolution process into several stages and becomes an integer programming solved by an improved genetic algorithm (GA). The lower-level model is a linear programming after the linearization of nonconvex constraints and minimizes the operation cost of the distribution system. In order to handle the uncertainty in distribution system, an hourly simulation is performed in the lower-level model which is solved by SeDuMi. Finally, the model is applied to a 13-bus distribution network to show its effectiveness.

Urban noise pollution is an omnipresent but often neglected threat to public health that must be addressed urgently. Passive noise control measures, which are less effective at reducing low-frequency noise and are often bulky and may impede airflow. As evidenced in automobiles, active control of cabin noise has resulted in lighter cars due to reduced passive insulation. Despite its long history and recent popularisation by consumer head- phones, the implementation of active noise control in the built environment is still rare. To date, active noise control (ANC) has been demonstrated, at source, in construction machines and, in the transmission path, in noise barriers. Recent demand for naturally-ventilated buildings has also spurred the development of active control solutions at the receiving end, such as on windows. The ten questions aim to demystify the principles of ANC and highlight areas in which environmental noise can be actively mitigated. Since the implementation of active control in the built environment usually involves multiple stakeholders, operational concerns are addressed. To conclude, research gaps are identified that would enable increased adoption of ANC in the built environment. There is also renewed interest in applying intelligent ANC to tackle environmentally complex applications, such as varying noise levels in the earcup of ANC headphones, particularly with the advent of the low-cost, low-power, highly-efficient embedded electronics; advancing speaker technology; and new impetus from digital signal processing and artificial intelligence Algorithms. Ministry of National Development (MND) National Research Foundation (NRF) Published version This research is supported by the Singapore Ministry of National Development and the National Research Foundation, Prime Minister’s Office under the Cities of Tomorrow (CoT) Research Programme (CoT Award No. COT-V4-2019-1). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of the Singapore Ministry of National Development and National Research Foundation, Prime Minister’s Office, Singapore.

With the continuous growing of wind power grid connected capacity, the requirements for the accuracy of wind farm equivalent models in power system analysis are increasing. However, traditional equivalence methods have disadvantages such as irrational number of clusters and poor adaptability of clustering results, which makes the methods unable to meet the needs in certain fields. Therefore, this paper proposes an equivalence method for wind farm based on clustering of output power time series data. This method uses the DTW algorithm to measure the similarity between the time series output power from each wind turbine generator. Then the DBSCAN algorithm is used to achieve a reasonable number of clusters and to identify the outliers. Finally, simulation verification is carried out under the conditions of wind speed disturbance and three-phase short-circuit fault, and the results are compared with those of the single-machine equivalent model and the k-means multi-machine equivalent model. Simulation results show the effectiveness of the proposed equivalence method, and the established model can reflect the operating characteristics of wind farm more accurately.

The gradual adaptation and possibility of divergence hinder the active noise control system from being applied to a wider range of applications. Selective active noise control has been proposed to rapidly reduce noise by selecting a pre-trained control filter for different primary noise detected without an error microphone. For stationary noise, considerable noise reduction performance with a short selection period is obtained. For non-stationary noise, more restrictive requirements are imposed on instant convergence, as it leads to faster tracking and better noise reduction performance. To speed up a selective filtered active noise control system, empirical wavelet transform is introduced here to accurately and instantaneously extract the frequency information of primary noise. The boundary of the first intrinsic mode function of random noises is extracted as the instant signal feature. Primary noise is attenuated immediately by picking the optimal pre-trained control filter labeled by the nearest boundary. The storage requirement for a pre-trained control filter library is reduced. Instant control is obtained, and the instability caused by output saturation is overcome. With more concentrated energy distribution, better noise reduction performance is achieved by the proposed algorithm compared to conventional and selective active noise control algorithms. Simulation results validate these advantages of the proposed algorithm. NRF (Natl Research Foundation, S’pore) Published version

The gradual adaptation and possibility of divergence have been the two main obstacles in the efficient implementation of conventional adaptive active noise control (ANC) to a wider range of applications. Selective ANC (SANC) has been proposed to rapidly reduce noise by selecting a pre-trained control filter for different primary noise detected, and improve the robustness of the system. For stationary noise, considerable noise reduction performance and system stability are obtained by SANC. However, for non-stationary noise, in order to track the variability of the signal, frequency-band-match and selection have to be conducted constantly, which results in high computational burden. To confront this problem, empirical wavelet transform (EWT) is introduced to simplify the matching and selection of SANC in this paper. This EWT based SANC (SANC_EWT) algorithm extracts the first mode of random noises, and attenuates the noise immediately by picking the optimal pre-trained control filter labeled by the first boundary. Therefore, computational complexity is reduced drastically. Simulation results show that convergence could be reached rapidly. Better noise reduction performance is achieved by SANC_EWT compared to conventional FxLMS and SANC algorithms.

Active noise control is an efficient approach in dealing with unwanted acoustic disturbances. However, most of the active noise control algorithms aim to control the signal of the error sensor leading to local noise attenuation only around the error microphones. One of the approaches to enlarge the quiet zone is by restraining the derivative of the sound field around the error microphone to zero. It achieves noise cancellation not only at the error microphone but also in its vicinity. This paper proposes a time-domain adaptive algorithm to implement the spatial derivative constraint, which avoids the complex analytic acoustic calculations. Furthermore, the proposed algorithm does not require extra microphones to acquire the sound field information during control. Numerical simulations are carried out to validate the effectiveness of the proposed method.

The multichannel implementation of the auxiliary-filter-based virtual-sensing (AF-VS) technique for active noise control applications is revisited and realized in the paper. Frequency-domain analysis based on random primary noise proves that the multichannel virtual-sensing active noise control (MVANC) technique can achieve optimal control at the desired virtual locations even if the signals at the physical and virtual microphones are not causally related. Further analysis on a number of sensor-actuator configurations shows that the MVANC technique achieves optimal control at the desired locations as long as the number of secondary sources does not exceed that of the physical error microphones. Furthermore, the simulations with measured transfer functions and real-time experiments conducted on a four-channel system validate the frequency domain analyses NRF (Natl Research Foundation, S’pore) Published version

Online dynamic security assessment is one of the important applications of online situation awareness for power systems, providing essential information for secure operation and preventive control. Different from the existing methods which are fault-dependent or requiring for transient-state data, this paper proposes a novel deep learning model for online power system multi-fault dynamic security assessment driven by hybrid information of anticipated faults and static operating points. The proposed model is universal and suitable for different fault situations. In addition, this model avoids the acquisition of transient information, thus reducing the time-consuming process of online evaluation. For the discrete nominal features in the hybrid information, one-hot encoding is introduced to preprocess the discrete nominal features. Both the performance of the dynamic security assessment model and the principal component analysis are promoted. Furthermore, an improved adaptive synthetic sampling algorithm is proposed and applied to alleviate the problem of power system data imbalance. Finally, the feasibility and effectiveness of the proposed method is verified by testing on the anticipated faults in the New England 39-bus system and the IEEE 54-machine 118-bus system.

Economic dispatch (ED) is of cardinal significance for the power system operation. It is mathematically a typical complex nonlinear multivariable strongly coupled optimization problem with equality and inequality constraints, especially considering the valve-point effects. In order to effectively solve the problem, a simple yet very young and efficient population-based algorithm named across neighborhood search (ANS) is implemented in this paper. In ANS, a group of individuals collaboratively navigate through the search space for obtaining the optimal solution by simultaneously searching the neighborhoods of multiple superior solutions. Four benchmark test cases with diverse complexities and characteristics are firstly employed to comprehensively verify the feasibility and effectiveness of ANS. The experimental and comparison results fully demonstrate the superiority of ANS in terms of the final solution quality, convergence speed, robustness, and statistics. In addition, the sensitivities of ANS to variations of population size and across-search degree are studied. Furthermore, ANS is applied to a practical provincial power grid of China. All the comparison results consistently indicate that ANS is highly competitive and can be used as a promising alternative for ED problems.

Parameter extraction of solar photovoltaic (PV) models is a typical complex nonlinear multivariable strongly coupled optimization problem. In this paper, an improved whale optimization algorithm (WOA), referred to as IWOA, is proposed to accurately extract the parameters of different PV models. The original WOA has good local exploitation ability, but it is likely to stagnate and suffer from premature convergence when dealing with complex multimodal problems. To conquer this concerning shortcoming, IWOA develops two prey searching strategies to effectively balance the local exploitation and global exploration, and thereby enhance the performance of WOA. Three benchmark test PV models including single diode, double diode and PV module models, and two practical PV power station models with more modules in the Guizhou Power Grid of China are employed to verify the performance of IWOA. The experimental and comparison results comprehensively demonstrate that IWOA is significantly better than the original WOA and three advanced variants of WOA, and is also highly competitive with the reported results of some recently-developed parameter extraction methods.

The performance of a solid oxide fuel cell (SOFC) is tightly related to relevant parameters associated with the internal multi-physicochemical processes. Accurate identification of these parameters is considerably important for modelling the voltage versus current (V-I) characteristic of SOFCs. In this paper, an improved teaching-learning based algorithm (TLBO) referred to as RTLBO is proposed to identify the exact values for these parameters. The parameter identification of SOFCs is transformed into a minimization optimization problem. The mean square error (MSE) between the measured output voltage and the calculated output voltage is used as the objective function. TLBO has been shown to be competitive with other population-based algorithms. However, its convergence rate is relatively slow especially for complex optimization problems. Inspired by the ranking mechanism in the actual scenarios of teaching-learning process, a ranking based learner selection method is proposed and integrated into both the teacher and learner phases of RTLBO. In RTLBO, poor learners are more likely to be eliminated from the current class in the ranking based teacher phase and good learners are more likely to be chosen to interact with others in the ranking based learner phase, which hence can improve the overall performance of the class quickly. The experimental results on a 5-kW SOFC stack comprehensively demonstrate that RTLBO is able to achieve a better trade-off between the exploration and exploitation compared with twelve advanced TLBO variants and eight popular advanced non-TLBO based methods. In addition, the sensitivity of RTLBO to variations of population size is empirically investigated.

Fault section diagnosis (FSD) of power systems plays an important role in power system operation. In order to quickly and accurately diagnose the fault section or sections after the occurrence of an event, a novel variant of brain storm optimization (BSO) in objective space algorithm, referred to as BCBSO (binary coded BSO), is proposed in this paper. The FSD problem is transformed into a 0–1 integer programming problem. The difference between the reported alarms and the expected states of protective relays and circuit breakers is used as the objective function. In BCBSO, each population individual is directly encoded as a binary vector and thereby the transcoding process can be avoided when solving the 0–1 integer programming problem. In addition, logical operations instead of floating operations are employed for binary strings, making the evolutionary process more convenient. In order to verify the performance of BCBSO, three test systems, i.e., the typical 4-substation power system, IEEE 118-bus system, and a practical power grid in Jilin province of China with different fault scenarios including single fault and multiple faults with failed and/or malfunctioned protective devices are employed. Six popular metaheuristic methods including ABC, BBO, DE, GA, PSO, and BSO are utilized to validate the effectiveness of BCBSO. The experimental results comprehensively demonstrate the superiority of BCBSO in terms of successful rate, diagnosis error, robustness, computation efficiency, convergence speed, and statistics. In addition, the effect of population size is investigated as well.

Dynamic economic dispatch (DED), mathematically, is a typical highly complex nonlinear multivariable strongly coupled optimization problem with equality and inequality constraints, especially considering valve-point effects. In this paper, a hybrid method named BBOSB by combining biogeography-based optimization (BBO) with brain storm optimization (BSO) is proposed. BBO has good local exploitation ability due to its information sharing mechanism. But it is likely to suffer from premature convergence when dealing with complex multimodal problems. Quite the opposite, BSO possesses excellent global exploration ability owing to its grouping evolution strategy which, however, also can drag its global searching process. In such contexts, the hybrid BBOSB method is able to fully take advantages of both BBO and BSO to conquer premature convergence and to accelerate the global searching process simultaneously. The experimental and comparison results on four non-convex benchmark DED test cases with valve-point effects and a practical provincial power system of China comprehensively demonstrate that BBOSB is highly competitive and can be used as a promising alternative for DED problems. In addition, the effect of population size on the optimization performance is investigated as well.

Power system economic dispatch (ED), mathematically, is a typical complex nonlinear multivariable strongly coupled optimization problem with equality and inequality constraints, especially considering the valve-point effects. In this paper, a novel variant of competitive swarm optimizer (CSO) referred to as OLCSO is proposed to solve both convex and non-convex ED problems. In the canonical CSO, the loser particle in each pair updates its position by learning from the winner particle. Such a learning strategy is easy to implement, but is not efficient enough because even a bad particle can contain useful information while a good particle may have misleading features. To solve this issue, an orthogonal learning (OL) strategy based on orthogonal experimental design is developed for CSO to quickly discover more useful information that contained in the winner and the loser particles. The OL strategy can easily construct a more promising solution to provide a more systematic search strategy for OLCSO. A set of 24 numerical benchmark functions and three ED cases with diverse complexities and characteristics are used to comprehensively verify the performance of OLCSO. The experimental results and comparison results consistently indicate that OLCSO can be used as a competitive alternative for ED problems.

Active noise control (ANC) technology is increasingly ubiquitous in wearable audio devices, or hearables. Owing to its low computational complexity, high robustness, and exemplary performance in dealing with dynamic noise, the fixed-coefficient control filter strategy plays a central role in portable ANC implementation. Unlike its traditional adaptive counterpart, the fixed-filter strategy is unable to attain optimal noise reduction for different types of noise. Hence, we propose a selective fixed-filter ANC method based on a simplified two-dimensional convolution neural network (2D CNN), which is implemented on a co-processor (e.g., in a mobile phone), to derive the most suitable control filter for different noise types. To further reduce classification complexity, we designed a lightweight one-dimensional CNN (1D CNN), which can directly classify noise types in time domain. A numerical simulation based on measured paths in headphones demonstrates the proposed algorithm’s efficacy in attenuating real-world non-stationary noise over conventional adaptive algorithms.

Active noise control (ANC) is gaining credence as an effective approach in mitigating low-frequency urban noise. Current ANC algorithms that attenuate noise across the full audio frequency band, often exert control effort excessively at higher frequencies. Moreover, such unrestrained control unavoidably attenuates some critical sounds, such as alarms and warning signals. In practice, excessive control effort in ANC usually results in output-saturation distortion, which affects the adaptive system stability and degrades the residual audio quality. To provide greater flexibility of control in frequency bands of interest without incurring output saturation, this paper proposes two variants of the filtered reference comb-partitioned frequency-domain adaptive filter (FxCFDAF) algorithm, namely the leaky FxCFDAF and drop-out FxCFDAF algorithms, which exert output-effort constraint at the control filter in the frequency domain. In addition, the comb-partitioning approach in the proposed FxCFDAF algorithms is delayless and computationally-efficient, which are essential for practical implementation, unlike the conventional frequency-domain adaptive algorithm. Experimental simulations carried out on measured primary and secondary paths validate the proposed algorithms’ advantage over the conventional FxLMS algorithm in mitigating large-amplitude noise without output saturation.

Accurate transient stability assessment (TSA) is a fundamental requirement for ensuring secure and stable operation of power systems. Tremendous efforts have been made to apply artificial intelligence approaches for TSA with phasor measurement unit data. However, many previous approaches may be failed to provide favorable accuracy due to the shallow architectures and error-prone hand-crafting features. This paper proposed a model for TSA, which is termed multi-branch stacked denoising autoencoder (MSDAE). This model is a unified framework integrating multiple stacked denoising autoencoders (SDAEs), one fusion layer, and one logistic regression (LR) layer. Initially, the SDAEs at the bottom of MSDAE extract features from multiple kinds of measurements respectively. Then, the extracted features are encoded into unified fusion features by the fusion layer. Finally, the LR layer performs TSA by using the fusion features. The depth of the architecture contributes to the remarkable ability for feature learning, while the width of the architecture (i.e., the multiple branches) enables MSDAE to deal with different kinds of measurements by a reasonable mechanism. In this way, MSDAE achieves feature extraction and classification intrinsically and simultaneously, namely, achieves TSA in an end-to-end manner. The results of experiments on IEEE 50-machine system demonstrate the superiority of the proposed model over the prior methods.

Conventional real-time active noise control (ANC) usually employs the adaptive filtered-x least mean square (FxLMS) algorithm to approach optimum coefficients for the control filter. However, lengthy training is usually required, and the perceived noise reduction is not immediately realized. Motivated by the practical implementation, we propose a selective fixed-filter active noise control (SFANC) algorithm, which selects a pre-trained control filter to attenuate the detected primary noise rapidly. On top of improved robustness, the complexity analysis reveals that SFANC appears to be more efficient. The SFANC algorithm chooses the most suitable control filter based on the frequency-band-match approach implemented in a partitioned frequency-domain filter. Through simulations, SFANC is shown to exhibit a satisfactory response time and steady-state noise reduction performance, even for time varying noise and real nonstationary disturbance. NRF (Natl Research Foundation, S’pore) Accepted version

With the continuous development of photovoltaic and energy storage hybrid system (PV-ES hybrid system), how to configure the power and capacity of energy storage system (ESS) has become a hot issue. In this paper, the energy storage allocation method of PV-ES hybrid system considering economy and power stability is proposed. The comparative benefit-cost ratio evaluation index of economy is defined based on the cost and benefit analysis theory. And by analyzing the characteristics of PV power on long and short time scale, the probability index of suppressing power fluctuations is defined in a statistic way. Combining the above two indexes, an aggregative index is constructed to establish an energy storage allocation model, and the power and capacity of energy storage can be determined. Finally, case study verifies the feasibility and effectiveness of the proposed method.

Cyber-attackers could stealthily interfere with the normal operation of power systems by maliciously tampering the measurements transmitted from field devices to control centers. Although some data-driven methods capable of identifying abnormal measurements have been proposed, there are some remaining shortcomings such as low accuracy and slow convergence rate. This paper devises a novel framework by combining convolutional neural networks (CNN) and long-short-term memory (LSTM) networks for detecting tampering attacks based on recognizing spatial-temporal correlations between measurements. Besides, optimization such as the attention mechanism, Dropout layers and SVM are applied to improve the performance of the proposed framework. This paper also introduces how to implement the proposed framework in practical cyber-physical systems and expounds the synergistic relationships between the data-driven detector and the traditional bad data detection module. Case studies prove that the proposed framework has better learning performance than existing ones.

Data-driven solutions to detecting power system measurement manipulation rely on thorough understandings of the patterns behind the data. From a probabilistic viewpoint, power system measurements inherently follow multimodal distributions usually with heavy tails. Existing normalization methods, both linear and nonlinear, suffer from some drawbacks when being applied to this kind of distribution. Using inappropriately normalized data to train the detectors may undermine their pattern recognition capability. In this paper, a tradeoff between the linear and nonlinear normalization is investigated to propose a variant of the widely used probability integral transform (PIT) by introducing an approximate linear compensation. When being employed to normalize power system measurements, the socalled desaturated PIT can have a better fidelity towards the measurements collected under infrequent operation conditions. Experiments show that the proposed normalization method is promising in improving the detection performance.

With a large number of renewable energy sources and DC-type loads entering the distribution network, the cost of inverter investment and system operation become high. By introducing a DC part into the traditional distribution network, the cost can be reduced. Also, it will be more convenient for new energy access. Nowadays, the main body of the distribution network is in AC form. Thus, it is impractical for whole grid DC retrofit. To analyze the optimal proportion of the DC part to be employed into the traditional distribution network, firstly, a bilevel optimization model is established to optimize the topology of the AC-DC distribution network. Secondly, in order to solve the bi-level optimization model, an improved GA algorithm, in which the Warshell algorithm is used to improve the convergence performance of the traditional GA algorithm, has been adopted. Finally, based on the bi-level optimization model, a multi-stage planning method of the distribution network is proposed, which reveals the dynamic topology evolution of the distribution network.

This paper revisits the virtual sensing active noise control (VS-ANC) technique and extends it to a general multichannel ANC (MCANC) implementation. A frequency domain analysis shows that the multichannel virtual sensing ANC (VS-MCANC) technique arrives at an optimal control filter to cancel the noise disturbance at the virtual locations and overcomes the spatial correlation and causality constraints between the physical microphone and the virtual microphone. A real-time control of broadband noise with a 4-channel VS-MCANC implemented in a test chamber validates its theoretical analysis and demonstrates its active control effectiveness.

Multichannel active noise control (MCANC) is widely regarded as an effective solution to achieve a significantly large noise-cancellation area in a complicated acoustic field. However, the computational complexity of MCANC algorithms, such as the multichannel filter-x least mean square (McFxLMS) algorithm, grows exponentially with an increased channel count. Many modified algorithms have been proposed to alleviate the complexity but at the expense of noise reduction performance. Till now, the trade-off between computational complexity and noise reduction performance has limited the practical implementation of MCANC. The block coordinate descent McFxLMS (BCD McFxLMS) algorithm proposed in this paper substantially reduces the computation cost of an MCANC system, while maintaining the same noise reduction performance as the conventional McFxLMS algorithm. Furthermore, a momentum mechanism is integrated into the BCD McFxLMS in practice to improve the convergence speed. The simulation and experimental results validate the effectiveness of proposed algorithms when dealing with noise in a realistic environment.

The feedforward active noise control (FF ANC) technique has been widely used to cancel the broadband noise in many practical applications. However, it fails to cope with the uncorrelated narrow-band disturbance picked up by the error sensor, which is independent of the reference signal picked up by the reference sensor. Hence, the alternative switching hybrid ANC has been proposed in the paper to attenuate this kind of uncorrelated narrow-band disturbance effectively. Furthermore, compared with the conventional hybrid ANC algorithms, the proposed algorithm alternatively switches to do updating between the feedforward (FF) and feedback (FB) ANC. The algorithm, thus, saves the computations and achieves a faster convergence speed. The numerical simulations and real-time experiments in the paper validate the effectiveness of the proposed algorithm, which exhibits better convergence and noise reduction performance than other hybrid active control algorithms.

This paper presents a new strategy in order to model the charging power demand due to large-scale plug-in electric vehicles (PEVs) as realistic a fashion as possible and analyze their impact on the residential power distribution system. The strategy takes the charging willingness of PEV users into consideration, and accounts for the difference in charging frequencies among users. A detailed classification, derived from the historical data on users’ driving patterns, on PEV users is conducted in order to ensure that users in the same user set have the same charging properties. Seven probability load models for PEV charging are established for these user sets, and each model accounts the inherent randomness in the usages and recharges of PEVs. After the consideration of charging willingness, the charging demand differs among weekdays. The aggregated charging demand from a user set on each weekday is calculated based on the Law of Large Numbers, and the total charging demand from all PEVs on each weekday can be obtained by accumulating the aggregated charging demand of the user sets with charging willingness. The strategy can ensure a high utilization of the battery capacity, and the aggregated charging demand resulted is more rational and credible. The proposed charging load modeling strategy is finally applied on the electric load profile on a winter day in Manitoba.

This paper presents an advanced variant of teaching learning based algorithm (TLBO) called either-or teaching learning based algorithm (EOTLBO) to extract accurate and reliable parameters of solar photovoltaic (PV) models. EOTLBO synergizes three enhanced strategies to accelerate the convergence rate and boost the search efficiency of TLBO. (i) A median learner based teacher phase excluding the mean position used in the original TLBO is designed to avoid infeasible and inefficient learners and to form a rational and advisable moving mechanism around the teacher. (ii) A higher-achieving learner based learner phase using three sorted learners is devised to directionally guide the target learner to jump out of local optima and to move towards a more promising region. (iii) A chaotic map based either-or teaching-learning strategy is developed to give each dimension of each learner a chance to go through either the median learner based teacher phase or the higher-achieving learner based learner phase. EOTLBO is applied to three PV cells/modules including seven cases. Compared with the original TLBO, four non-TLBO variants, and five TLBO variants, experimental results verify the superior performance of EOTLBO in terms of both the quality of final solutions and the convergence speed on all cases. In addition, the current-voltage characteristics yielded by EOTLBO agree well with the measured data independently of different PV models at different operating conditions.

By maliciously manipulating the synchrophasors produced by phasor measurement units in power systems, cyber attackers can mislead the control center into taking wrong actions. From the viewpoint of machine learning, normal and malicious synchrophasors may exhibit different spatial distribution characteristics when mapped into a latent space. Hence, a malicious synchrophasor detector can be acquired by training a classification model with instances derived from historical operational synchrophasor data. However, malicious synchrophasors occur infrequently in practice. It is likely to incur a great deal of effort and may even introduce inevitable experience errors when extracting and labeling a sufficient number of malicious synchrophasors from historical operational data for training. For most existing detectors, if they are directly trained with highly imbalanced datasets, their performances may severely deteriorate. In this paper, a novel type of malicious synchrophasor detector is developed based on a combinatorial use of data rebalancing, Bagging-based ensemble learning, and the widely recognized eXtreme Gradient Boosting (XGBoost) classifier. Experiments show that although fewer malicious instances are provided, the proposed detector is still capable of detecting malicious synchrophasors.

Modeling solar photovoltaic (PV) systems accurately is based on optimal values of unknown model parameters of PV cells and modules. In recent years, the use of metaheuristics for parameter extraction of PV models gains more and more attentions thanks to their efficacy in solving highly nonlinear multimodal optimization problems. This work addresses a novel application of supply-demand-based optimization (SDO) to extract accurate and reliable parameters for PV models. SDO is a very young and efficient metaheuristic inspired by the supply and demand mechanism in economics. Its exploration and exploitation are balanced well by incorporating different dynamic modes of the cobweb model organically. To validate the feasibility and effectiveness of SDO, four PV models with diverse characteristics including RTC France silicon solar cell, PVM 752 GaAs thin film cell, STM6-40/36 monocrystalline module, and STP6-120/36 polycrystalline module are employed. The experimental results comparing with ten state-of-the-art algorithms demonstrate that SDO performs better or highly competitively in terms of accuracy, robustness, and convergence. In addition, the sensitivity of SDO to variation of population size is empirically investigated. The results indicate that SDO with a relatively small population size can extract accurate and reliable parameters for PV models.

Fault section location (FSL) is an important role in facilitating quick repair and restoration of distribution networks. In this chapter, an oppositional brain storm optimization referred to as OBSO is proposed to effectively solve the FSL problem. The FSL problem is transformed into a 0–1 integer programming problem. The difference between the reported overcurrent and expected overcurrent states of the feeder terminal units (FTUs) is used as the objective function. BSO has been shown to be competitive to other population-based algorithms. But its convergence speed is relatively slow. In OBSO, opposition-based learning method is utilized for population initialization and also for generation jumping to accelerate the convergence rate. The effectiveness of OBSO is comprehensively evaluated on different fault scenarios including single and multiple faults with lost and/or distorted fault signals. The experimental results show that OBSO is able to achieve more promising performance.

Extracting accurate and reliable values for involved unknown parameters of solar photovoltaic (PV) cells/modules is considerably significant to the characteristic analysis, fault diagnosis, maximum power point tracking, and efficiency evaluation of PV systems. Solving this problem using metaheuristic algorithms has gained increasing attention recently owing to their versatile and promising applications in highly nonlinear multimodal optimization problems. In this paper, an efficient and effective variant of competitive swarm optimizer (CSO) named WLCSODGM is presented to solve the parameter extraction problem of PV models. CSO is an advanced variant of particle swarm optimization and performs well especially on unimodal optimization problems. However, it is easily trapped in local optima when solving complex multimodal optimization problems such as the one considered here due to its poor exploration. In WLCSODGM, two improved components are introduced to remedy the inadequacy of CSO. On the one hand, a winner-leading search strategy is proposed to favor the exploration and help losers locate more promising regions. On the other hand, a dynamic Gaussian mutation operator with stretchable mutation amplitude and adaptive mutation probability is integrated to further enhance the exploration to help individuals jump out of poor local optima. WLCSODGM is applied to four different PV models and verified using a total number of twelve state-of-the-art algorithms. In addition, the influences of improved components and relevant algorithmic parameters are also experimentally evaluated. Results demonstrate that WLCSODGM is significantly better or highly competitive compared with the other algorithms.