The LLC converter, known for its soft-switching properties, is widely used in industry. To balance the best steady-state and dynamic performance and to facilitate controller design, the small-signal model is critical. Recently, based on the time domain trajectory under perturbation and the extended description function, Y. -H. Hsieh proposed an accurate modeling method of bang-bang charge-controlled LLC converter. However, it did not consider the impact on the model of adding ramp compensation, which is commonly introduced in industrial applications to improve the stability of charge control. In addition, the modeling required consistent control logic between positive and negative half-cycles, i.e., the symmetric charge control. When dealing with asymmetric charge control, such as current-integrating charge control, it becomes inaccurate. This paper explores the extension of this modeling approach to LLC converters adopting asymmetric charge control with the consideration of ramp compensation. The small signal model is subsequently established, and both simulation and experimental results validate its accuracy.

Multimodal recommendation enhances item representation in recommendation systems by integrating diverse modalities of item information beyond traditional ID-based features. This approach utilizes supplementary item details, including images, text, videos, and audio, to refine the accuracy of item representation and thereby boost the precision of recommendations. Multimodal recommendation has emerged as a vibrant field within the scope of systems that generate suggestions. It offers a powerful approach to address the challenges of data scarcity and the representation of long-tail content, thereby improving the overall quality of recommendations. However, the advancement in multimodal recommendation is currently hindered by two main obstacles. First, the process of extracting multimodal features from pre-trained models using either shallow or deep neural networks often results in insufficient data extraction or sparse recommendation data, leading to suboptimal model performance. Second, a significant portion of previous research has focused on integrating information across modalities, often overlooking the distinct characteristics inherent in different modalities. Addressing these challenges, we introduce a unique methodology titled “modality-disentangled featureextraction via knowledge distillation in multimodal recommendation systems” (MODEST). First, to tackle the aforementioned problems that arise when extracting multimodal features with either deep or shallow neural networks, our approach adopts a teacher–student network framework. In this framework, deep neural networks are utilized to extract representation vectors from text and image data. Feature fusion is then carried out via attention mechanisms, and semantic labels are employed as classification labels to derive three supervised learning loss functions. This process significantly enhances the teacher network’s capacity to extract multimodal features. Subsequently, the knowledge from the teacher network is transferred to the student network through knowledge distillation. The student network makes use of a shallow neural network, and during the inference stage, we rely on the student network. This strategy effectively resolves the issues of data sparsity in deep networks and insufficient information extraction in shallow networks. Second, to more effectively capture the similarities and distinct features among different modalities, we implement a disentangled modality decomposition technique. Through integrated mappings, it separately extracts text and image information from the teacher–student networks and decomposes them into cross-modality common information and cross-modality specific information. By applying the constraints of contrastive learning, we minimize the distance between cross-modality common information and maximize the separation of cross-modality specific information, promoting convergence with the aid of auxiliary loss. This effectively addresses the problem of cross-modality feature alignment. Lastly, we combine the recommendation loss function with the multiple loss constraints we have added to formulate a unified optimization objective function. To underscore the remarkable efficacy of our proposed model, we have executed comprehensive experiments and visualizations on several real-world datasets. The results distinctly show a significant enhancement in our model’s performance, allowing it to achieve a level of competitiveness against other methods.

In recent years, recommendation systems have become increasingly prevalent in various fields, facilitating quick access to the information users need. As a result, many models have been proposed to model user interests, leading to more accurate recommendation lists, superior user experience, and business value. However, characterizing the dynamically changing interests of users is a challenging task. User interests shift over time while maintaining some long-term interests, and at each time, users’ interests are diverse. To investigate the benefits of multidimensional interests for users, this article proposes to characterize user preferences based on their spatiotemporal interests. Utilizing temporal and spatial information is critical for improving recommendation accuracy. To achieve this, we present a novel approach called multilong short-term interest (MLSI) user representation for recommendation. This method extracts long-term and short-term interests of users from their behavioural sequences using decoupled self-supervised learning with different optimizers. Self-attention is then employed to capture the diverse interests of users through their behavioral sequences. Final, long-term and short-term interests, as well as diversified interests, are aggregated to represent user interests. Extensive experiments on real-world datasets show that MLSI not only outperforms state-of-the-art methods but also more effectively characterizes user interests, reflecting an improvement ranging from 5% to 20% across various metrics on multiple datasets.

The Star-connected Cascaded H-Bridge (SCHB) converter is one of the most attractive multilevel topologies for medium/high-voltage STATCOM. Under negative current compensation status, cluster voltage balance control is of great importance and requires extra unbalanced active power transfer pathway. Zero- sequence voltage injection (ZSVI) is a normal method to exchange the active power but the large ZVSI will seriously restrict the negative sequence current range. Hybrid Cascaded STATCOM, with two-level three- phase module is proposed to reduce the value of ZSVI, and the phasor control is designed by calculating the coordinate position, which reduces the calculation complexity and extend the negative sequence current compensation range greatly. Finally, the proposed control strategy within the derived negative sequence current range is verified by the experimental results on a 400V/7.5kVar hybrid cascaded STATCOM.

With the development of renewable energy and the widespread use of electric vehicles, the AC/DC converter, which plays an important role as an interface to the power grid, has drawn much attention. Compared to the two stage AC/DC converter, the single stage AC/DC converter has the merits of high power density and high efficiency. with introducing the LLC resonant tank, the converter can achieve soft switching operation and wide voltage range. However, in the LLC-type single stage AC/DC converter, the control method is usually the Pulse Frequency Modulation (PFM), which has less outstanding dynamic response faced with sudden changes in load. To improve the dynamic response, this paper proposes a charge control strategy for the single stage AC/DC converter. By introducing charge control as the inner loop, the bandwidth can be extended. The simulation of an LLC-type matrix AC/DC converter is implemented to verify the proposed control strategy. A 1kW prototype is built and the preliminary experiment is conducted. The simulation and the preliminary experimental results are in good agreement with the theoretical analysis.

Streaming Session-Based Recommendation (SSBR) is a challenging problem as user preferences in sessions are continually drifting with sessions generated chronologically. In recent years, some SSBR models have been proposed to address this problem by reservoir technique and Graph Neural Networks (GNN) which help to preserve a representative sketch of the historical data and extract item transition information in sessions. However, there are two critical problems in existing methods: (1) most existing methods only focus on the local session information without exploiting the information of other sessions and users; (2) GNN models in existing SSBR methods are unable to capture the importance of different user features. To address the problems mentioned above, we propose a novel architecture namedGlobalItem andUser embeddingAssistedGraphNeuralNetwork (GIUA-GNN) for combining the global user and item information in an attentional manner with local session information for the recommendation. We also propose a novel architecture of graph neural network which utilizes the attention mechanism for better extracting the importance of different features of user embeddings namedBi-directedAttentionalGraphConvolutionalNetwork (BA-GCN). Extensive experiments on three different sizes of real-world datasets have been conducted to demonstrate the superiority of our model on metrics MRR and Recall.

To achieve low cost and high power density, low dc-link capacitance is preferred in the Cascaded H-Bridge (CHB) STATCOM. However, low dc-link capacitance will result in the increase of dc-link ripples, which will introduce voltage harmonics through modulation. According to the operation principle of the inverter, the output voltage will not exceed the input voltage at any time. Hence, the output voltage range of STATCOM will be affected by large voltage oscillation across the DC link, leading to affect the output capacity of STATCOM. Thus, to guarantee operation range of STATCOM, system capacity should be a key design consideration in the design of low dc-link capacitance, which has not been taken into consideration in the existing design methods. To address this issue, this paper establishes the relationship between the range of compensable current and the dc-link capacitance based on the Kirchhoff's law. Based on this relationship, a new design consideration is proposed to achieve the minimum dc-link capacitance, which is as small as possible under certain capacity of STATCOM. The designed capacitance can satisfy the requirements of system capacity and support the proper operation of STATCOM. Meanwhile, the designed capacitance is small enough to achieve low cost and high power density. A 400V/7kVA STATCOM prototype is built to verify the derived relationship and the proposed design method.

A novel non-isolated three-port converter (TPC) is analyzed and evaluated for a photovoltaic (PV)-battery hybrid power system application. An interleaved bidirectional Buck/Boost converter and a semi-active full-bridge rectifier are connected in a stack configuration to provide three power ports, i.e. a PV input port, a battery storage port and an output load port, simultaneously. Single-stage power conversion between any two of the three ports is achieved, while independent power control of two of the three power ports is realized by adopting PWM plus phase-shift modulation strategy. In addition, soft-switching can be achieved for all active switches. The operation principles, modulation and control strategies are analyzed in detail, design considerations of this TPC are presented as well. Performance of this TPC is further evaluated with experimental results.

A semi-two-stage transformerless PV grid-tied inverter based on H5 is proposed in this paper. By introducing the front-end Boost converter, the PV module and the output of the front-end Boost can be used as two different inputs of the H5 converter respectively. Therefore, the proposed topology can operate in a wide input range and part of the PV module power can be transmitted to the utility grid directly without being transferred through the path of the front-end Boost converter. The efficiency can be higher than that of the conventional two-stage H5 inverter. A modulation strategy for this topology is presented and analyzed in detail. The common-mode (CM) voltage varies with a low frequency which results in a low leakage current. An experimental prototype is built to verify the effectiveness of the proposed topology and modulation strategy.

Colourisation is a process of adding colour to greyscale images. In this study, the authors propose two new colourisation models based on non‐local total variation regularisation in the chromaticity and brightness (CB) colour space and the YIQ colour space. Lagrange multiplier method is used to handle the sphere constraint of chromaticity in the CB colour space. By introducing an extra variable and using the dual version of non‐local total variation, they split the proposed colourisation problems into two subproblems with closed‐form solutions and get two iterative algorithms. Experimental results and comparisons demonstrate that the advantage of the proposed methods is that they can preserve the colour edges better than the closely related existing methods, especially the total variation methods.

A new digital control scheme aiming to improve the transient response of an FPGA-based digitally controlled DC-DC converters is presented in this paper. The proposed approach enhances the transient response by dynamically controlling the ramp of the Digital Pulse Width Modulator (DPWM) unit through applying either linear or nonlinear shift to the conventional ramp-based DPWM. This allows the compensator to reach steady-state faster. The advantages and disadvantages of both techniques are presented and weight against improvement on transient response. Detailed analysis, simulation results and experimental waveforms are presented to verify the concept.