Good Generalization Performance Research Articles

A lightweight dual-attention network for tomato leaf disease identification

Tomato disease image recognition plays a crucial role in agricultural production. Today, while machine vision methods based on deep learning have achieved some success in disease recognition, they still face several challenges. These include issues such as imbalanced datasets, unclear disease features, small inter-class differences, and large intra-class variations. To address these challenges, this paper proposes a method for classifying and recognizing tomato leaf diseases based on machine vision. First, to enhance the disease feature details in images, a piecewise linear transformation method is used for image enhancement, and oversampling is employed to expand the dataset, compensating for the imbalanced dataset. Next, this paper introduces a convolutional block with a dual attention mechanism called DAC Block, which is used to construct a lightweight model named LDAMNet. The DAC Block innovatively uses Hybrid Channel Attention (HCA) and Coordinate Attention (CSA) to process channel information and spatial information of input images respectively, enhancing the model’s feature extraction capabilities. Additionally, this paper proposes a Robust Cross-Entropy (RCE) loss function that is robust to noisy labels, aimed at reducing the impact of noisy labels on the LDAMNet model during training. Experimental results show that this method achieves an average recognition accuracy of 98.71% on the tomato disease dataset, effectively retaining disease information in images and capturing disease areas. Furthermore, the method also demonstrates strong recognition capabilities on rice crop disease datasets, indicating good generalization performance and the ability to function effectively in disease recognition across different crops. The research findings of this paper provide new ideas and methods for the field of crop disease recognition. However, future research needs to further optimize the model’s structure and computational efficiency, and validate its application effects in more practical scenarios.

Hand Trajectory Recognition by Radar with a Finite-State Machine and a Bi-LSTM

Gesture plays an important role in human–machine interaction. However, the insufficient accuracy and high complexity of gesture recognition have blocked its widespread application. A gesture recognition method that combines state machine and bidirectional long short-term memory (Bi-LSTM) fusion neural network is proposed to improve the accuracy and efficiency. Firstly, gestures with large movements are categorized into simple trajectory gestures and complex trajectory gestures in advance. Afterwards, different recognition methods are applied for the two categories of gestures, and the final result of gesture recognition is obtained by combining the outputs of the two methods. The specific method used is a state machine that recognizes six simple trajectory gestures and a bidirectional LSTM fusion neural network that recognizes four complex trajectory gestures. Finally, the experimental results show that the proposed simple trajectory gesture recognition method has an average accuracy of 99.58%, and the bidirectional LSTM fusion neural network has an average accuracy of 99.47%, which can efficiently and accurately recognize 10 gestures with large movements. In addition, by collecting more gesture data from untrained participants, it was verified that the proposed neural network has good generalization performance and can adapt to the various operating habits of different users.

Self-Supervised Monocular Depth Estimation for Endoscopic Imaging.

Endoscopy holds a pivotal role in the early detection and treatment of diverse diseases, with artificial intelligence (AI)-assisted methods increasingly gaining prominence in disease screening. Among them, the depth estimation from endoscopic sequences is crucial for a spectrum of AI-assisted surgical techniques. However, the development of endoscopic depth estimation algorithms presents a formidable challenge due to the unique environmental intricacies and constraints within the dataset. This paper proposes a self-supervised depth estimation network to comprehensively explore the brightness changes in endoscopic images, and fuse different features at multiple levels to achieve an accurate prediction of endoscopic depth. First, a FlowNet is designed to evaluate the brightness changes of adjacent frames by calculating the multi-scale structural similarity. Second, a feature fusion module is presented to capture multi-scale contextual information. Experiments show that the average accuracy of the algorithm is 97.03% in the Stereo Correspondence and Reconstruction of Endoscopic Data (SCARED dataset). Based on the training parameters of the SCARED dataset, the algorithm achieves superior performance on the other two datasets (EndoSLAM and KVASIR dataset), indicating that the algorithm has good generalization performance.

Intelligent Identification of Surrounding Rock Grades Based on a Self-Developed Rock Drilling Test System

The classification of surrounding rock is crucial for formulating safe tunnel construction plans and support measures. However, the complex geological environment of tunnels presents a challenge in obtaining accurate drilling parameters for rock mass classification. This paper presents the development of a rock drilling testing system, which includes a propulsion speed acquisition system, oil pressure acquisition system, air pressure acquisition system, and an automatic data acquisition system. This system enables real-time, high-precision automatic collection and storage of parameters such as propulsion speed, with data collected twice per second for each parameter. Leveraging the Qingdao Metro Line 6 as a case study, we conducted rock mass drilling and constructed a rock mass classification database. By employing kernel density estimation and Pearson correlation analysis, we quantified the correlation between rock mass classification and the drilling parameters. The results indicated that relying on a single drilling parameter is insufficient for accurately determining rock mass classification. Both impact pressure and rotational pressure showed the strongest correlation with rock mass classification, each with a correlation coefficient below −0.8 (indicating a strong negative correlation). Outlier values of drilling parameters were excluded using the interval method. Based on the remaining data, we established an intelligent rock mass classification model using the random forest algorithm. This model demonstrated good accuracy and generalization performance, with an average accuracy exceeding 0.9. The proposed rock drilling testing system, combined with the intelligent rock mass classification model, forms an integrated system for the intelligent identification of rock mass grades. This system has significant implications for the intelligent and safe construction of drill-and-blast tunnels.

Unsupervised Test-Time Adaptation Learning for Effective Hyperspectral Image Super-Resolution With Unknown Degeneration.

Fusing a low-resolution hyperspectral image (HSI) with a high-resolution (HR) multi-spectral image has provided an effective way for HSI super-resolution (SR). The key lies on inferring the posteriori of the latent (i.e., HR) HSI using an appropriate image prior and the likelihood determined by the degeneration between the latent HSI and the observed images. However, in scenarios with complex imaging environments and various imaging scenes, the prior of HSIs can be prohibitively complicated and the degeneration is often unknown, which causes it difficult to accurately infer the posteriori of each latent HSI. To tackle this problem, we present an unsupervised test-time adaptation learning (UTAL) framework for HSI SR under unknown degeneration. Instead of directly modeling the complicated image prior, it first implicitly learns a content-agnostic prior shared across different images through supervisedly pre-training a mutual-guiding fusion module on extensive synthetic data. Then, it adapts the shared prior to those private characteristics in the latent HSI for posteriori inference through unsupervisedly learning a self-guiding adaptation module and a degeneration estimation network on two observed images in the test phase. Such a two-stage learning scheme models the complicated image prior in a divide-and-conquer manner, which eases the modeling difficulty and improves the prior accuracy. Moreover, the unknown degeneration can be estimated properly. Both of these two advantages empower us to accurately infer the posteriori of the latent HSI, thereby increasing the generalization performance in real applications. Additionally, in order to further mitigate the over-fitting in coping with more challenging cases (e.g., degenerations in both spectral and spatial domains are unknown) and speed up, we propose to meta-train UTAL on extensive synthetic SR tasks and solve it using an alternative optimization strategy such that UTAL learns to produce good generalization performance in real challenging cases with a small number of gradient descent steps. To verify the efficacy of UTAL, we evaluate it on HSI SR tasks with different unknown degenerations as well as some other HSI restoration tasks (e.g., compressive sensing), and report strong results superior to that of existing competitors.

A hybrid tool wear prediction model based on JDA

PurposeAiming at solving the problems of low prediction accuracy and poor generalization caused by the difference in tool wear data distribution and the fixation of single global model parameters, a hybrid prediction modeling method for tool wear based on joint distribution adaptation (JDA) is proposed.Design/methodology/approachFirstly, JDA is exploited to adapt the data features with different data distributions. Then, the adapted data features are identified by the KNN classifier. Finally, according to the tool state classification results, different regression prediction models are assigned to different wear stages to complete the whole tool wear prediction task.FindingsThe results of milling experiments show that the maximum prediction accuracy of this method is 95.13%, and it has good recognition accuracy and generalization performance. Through the application of the tool wear hybrid prediction modeling method, the prediction accuracy and generalization performance of the model are improved and the tool monitoring is realized.Originality/valueThe research results can provide solutions and a theoretical basis for the application of tool wear monitoring technology in practical industrial applications.

First results of the NOGGO PERCEPTION: Phase II investigational study of pembrolizumab in combination with chemotherapy in platinum-sensitive recurrent low-grade serous ovarian cancer.

e17550 Background: Low-grade serous ovarian cancer (LGSOC), a rare ovarian malignancy, exhibits a very limited responsiveness to chemotherapy. There is a pressing need for new therapeutic combinations with modern agents to enhance response rates and prognosis in this patient subgroup. Immune checkpoint inhibitors offer a promising pathway, having shown effectiveness in various malignant diseases, including selected cases of ovarian cancer. If our trial should show pembrolizumab effectivity in LGSOC, it would be a signal and impulse for future clinical studies in this rare disease. Methods: This clinical trial is a multi-center, single-arm phase 2 study to evaluate pembrolizumab therapy concomitant to platinum-based chemotherapy (carboplatin plus pegylated liposomal doxorubicin & carboplatin plus gemcitabine) and as maintenance in recurrent LGSOC cases. Eligible for the trial are LGSOC patients with progression or recurrence at least six months after most previous platinum-containing therapy and in good general performance (ECOG 0/1). The primary objective is the 12 months progression free survival (PFS) rate. Secondary end-points include overall survival, response rate (RR), PFS and RR according to Ki67 expression levels, time to first subsequent therapy and its response, safety and quality of life. The trial is planned according to Simon’s two-stage design with total sample size up to 33 patients. The null hypothesis is PFS-rate of at least 12 months of 20%. During the first stage 18 patients were enrolled and since 5 patients showed PFS of 12 months the study has been continued within stage 2 to enrol additional 15 patients. The trial can be claimed as successful, if at least 11 patients show PFS after 12 months. Assuming a true PFS-rate of 40%, this trial has 5% type I error rate and 80% power. Results: Data from 18 patients enrolled into this trial from FPI up to the time of interim analysis have been analyzed. 5 patients had reached the primary endpoint. 6 patients progressed or died before they reached the primary endpoint. At this time, 10 patients are still undergoing treatment and 7 patients may still reach the primary endpoint. Patients have a median age of 60 years (range: 43-81), ECOG score of 0 was recorded in 16 patients (88.9%) and most received one prior line of chemotherapy (61.1%, range: 1-5). 14 patients (77.8%) received carboplatin + PLD and 4 patients (22.2%) received carboplatin + gemcitabine as chemotherapy. 10 patients (55.6%) had at least one SAE. 3 patients (16.7%) had at least one SAE with relation to study treatment. 1 SAE resulted in a fatal outcome but was assessed not be related to study treatment. Conclusions: The interim analysis showed positive results, strengthening the hypothesis that pembrolizumab is an effective agent capable of extending progression-free survival in the studied patient population. Clinical trial information: NCT04575961 .

A Reduced Order Schwarz Method for Nonlinear Multiscale Elliptic Equations Based on Two-Layer Neural Networks

Neural networks are powerful tools for approximating high dimensional data that have been used in many contexts, including solution of partial differential equations (PDEs). We describe a solver for multiscale fully nonlinear elliptic equations that makes use of domain decomposition, an accelerated Schwarz framework, and two-layer neural networks to approximate the boundary-to-boundary map for the subdomains, which is the key step in the Schwarz procedure. Conventionally, the boundary-to-boundary map requires solution of boundary-value elliptic problems on each subdomain. By leveraging the compressibility of multiscale problems, our approach trains the neural network offline to serve as a surrogate for the usual implementation of the boundary-to-boundary map. Our method is applied to a multiscale semilinear elliptic equation and a multiscale $p$-Laplace equation. In both cases we demonstrate significant improvement in efficiency as well as good accuracy and generalization performance.

MERA: Meta-Learning Based Runtime Adaptation for Industrial Wireless Sensor-Actuator Networks

IEEE 802.15.4-based industrial wireless sensor-actuator networks (WSANs) have been widely deployed to connect sensors, actuators, and controllers in industrial facilities. Configuring an industrial WSAN to meet the application-specified quality of service (QoS) requirements is a complex process, which involves theoretical computation, simulation, and field testing, among other tasks. Since industrial wireless networks become increasingly hierarchical, heterogeneous, and complex, many research efforts have been made to apply wireless simulations and advanced machine learning techniques for network configuration. Unfortunately, our study shows that the network configuration model generated by the state-of-the-art method decays quickly over time. To address this issue, we develop a ME ta-learning based R untime A daptation (MERA) method that efficiently adapts network configuration models for industrial WSANs at runtime. Under MERA, the parameters of the network configuration model are explicitly trained such that a small number of optimization steps with only a few new measurements will produce good generalization performance after the network condition changes. We also develop a data sampling method to reduce the measurements required by MERA at runtime without sacrificing its performance. Experimental results show that MERA achieves higher prediction accuracy with less physical measurements, less computation time, and longer adaptation intervals compared to a state-of-the-art baseline.

A physics-guided deep learning model for predicting the magneto-induced mechanical properties of magnetorheological elastomer: Small experimental data-driven

Magnetorheological elastomer (MRE) is a novel intelligent material, which shows excellent potential in vibration control applications. Previous researches have fully demonstrated that the magneto-induced shear storage modulus of MRE largely determines the vibration control effect. However, both existing theoretical and experimental ways to measure the magneto-induced shear storage modulus of MRE face their own shortage. Therefore, a novel physics-guided deep learning model is proposed to efficient predict the magneto-induced mechanical properties of MRE based on Magnetic Dipole theory and data-driven methods. A small database is built by collecting the magneto-induced shear storage modulus of MRE with different material ratios tested on a special shear rheometer. The proposed model trained with small training samples and its prediction results fit well with experimental values (average R2 of 0.99) which is superior to existing constitutive models. The training only takes 25 s, which significantly shortens the time compared to the experiment. Furthermore, the proposed model effectively predicts the magneto-induced storage modulus of MRE and has good generalization and superior transfer performance.

Human Activity Recognition Based on Deep Learning Regardless of Sensor Orientation

In recent years, the continuous progress of wireless communication and sensor technology has enabled sensors to be better integrated into mobile devices. Therefore, sensor-based Human Activity Recognition (HAR) has attracted widespread attention among researchers, especially in the fields of wearable technology and ubiquitous computing. In these applications, mobile devices’ built-in accelerometers and gyroscopes have been typically used for human activity recognition. However, devices such as smartphones were placed in users’ pockets and not fixed to their bodies, and the resulting changes in the orientation of the sensors due to users’ habits or external forces can lead to a decrease in the accuracy of activity recognition. Unfortunately, there is currently a lack of publicly available datasets specifically designed to address the issue of device angle change. The contributions of this study are as follows. First, we constructed a dataset with eight different sensor placement angles using accelerometers and gyroscopes as a prerequisite for the subsequent research. Second, we introduced the Madgwick algorithm to extract quaternion mode features and alleviate the impact of angle changes on recognition performance by fusing raw accelerometer data and quaternion mode features. The resulting study provides a comprehensive analysis. On the one hand, we fine-tuned ResNet and tested its stability on our dataset, achieving a recognition accuracy of 97.13%. We included two independent experiments, one for user-related scenarios and the other for user-independent scenarios. In addition, we validated our research results on two publicly available datasets, demonstrating that our method has good generalization performance.

Three-dimensional inversion for short-offset transient electromagnetic data based on 3D U-Net

Abstract The short-offset transient electromagnetic (SOTEM) method carries out surveys in the near source region; the strong signal makes it suitable for deep detection with high precision. When the underground structure is complex, three-dimensional (3D) inversion of SOTEM data is necessary to meet the need of high-precision detection. Currently, difficulties faced by the conventional 3D inversion methods include high computational complexity, and the influence of the initial model. Deep learning (DL), as a completely nonlinear algorithm, can predict the underground structure from the measured data. DL is completely data-driven and does not use traditional misfit optimization methods. In this study, an efficient method is proposed to conduct 3D inversion for SOTEM data, which trains a 3D U-Net based on massive data to establish a mapping from SOTEM data to geoelectric models. After the training is completed, we input the new SOTEM data into the trained network, and the corresponding geoelectric model can be obtained. Although the training is a time-consuming process, prediction for new data can be completed in seconds. The inversion results for simulated data indicate that the 3D U-Net has good generalization performance and anti-noise ability. The inversion performance of 3D U-Net on the double-anomaly model improves by 51.1% compared to a 3D fully convolutional network (FCN). The inversion results of the 3D U-Net on the field data successfully delineated the aquiferous collapse column. The inversion results for simulated and field data demonstrate that the proposed method can achieve accurate 3D inversion for large volume of data while greatly saving computational time.

Distance metric learning-based multi-granularity neighborhood rough sets for attribute reduction

Attribute reduction is a hot research topic in data mining, in which rough set theory-based attribute reduction methods have been widely focused. The neighborhood rough set (NRS) model has good generalization performance and practicality in uncertainty reasoning, so it is often used for attribute reduction in recent years. Calculating the distance between samples in different attribute spaces is a key step in the NRS-based attribute reduction methods, which directly affects the performance of the reduction algorithm. However, the NRS model uses a fixed computational paradigm in calculating the distance between samples and does not consider the influence of labels in the attribute spaces on the distance calculation, which is not conducive to improving the performance of the reduction algorithm. Distance metric learning takes full account of the labels information in the multi-dimensional attribute space, and it can learn the distance between samples by taking into account the integrated principle that samples with the same label are closer and samples with different labels are further away, which helps to reduce the classification uncertainty. Inspired by this, this paper firstly incorporates distance metric learning into the NRS model from the perspective of multi-dimensional attribute space and proposes a distance metric learning-based multi-granularity neighborhood rough set (DmlMNRS) model. The related properties of the DmlMNRS model are also introduced and proved. Then, the DmlMNRS-based attribute reduction criterion and the significance of the attributes are defined. A DmlMNRS-based heuristic attribute reduction (DMNHAR) algorithm is designed based on this. Finally, experiments are performed on fifteen publicly datasets, and the experimental results show that the proposed algorithm has better robustness and classification performance.

GAITBOOST: Boosting gait recognition performance with BAT-extreme learning machines algorithm

Gait analysis is a widely used technique for passive human identification and tracking, with potential applications in security and surveillance systems. However, existing gait recognition methods face challenges in handling changing angles and uncertain features. In this paper, we propose a novel gait recognition approach that leverages real-time spatio-temporal gait features, including step length, gait cycle, height, cadence, swing ratio, and foot length. We apply the Extreme Learning Machines (ELM) algorithm for classification, which has been shown to be effective in various applications due to its fast-learning speed and good generalization performance. To further enhance the recognition rate, we introduce an evolutionary BAT-optimized ELM algorithm that addresses the instability issue in ELM. The proposed BAT-ELM algorithm can optimize the hidden nodes and weights of ELM, which leads to improved efficiency in recognizing gait from multiple view angles ranging from 0° to 180°. Our comprehensive analysis of the proposed approach indicates that it outperforms other reported algorithms in terms of recognition rate and efficiency. Our work demonstrates the effectiveness of combining real-time spatio-temporal gait features with the BAT-ELM algorithm for gait recognition. The proposed approach has potential applications in various fields, including security and surveillance systems, healthcare, and robotics. Our findings highlight the importance of leveraging evolutionary algorithms to optimize machine learning models and achieve better performance in complex recognition tasks.

PermuteDDS: a permutable feature fusion network for drug-drug synergy prediction

MotivationDrug combination therapies have shown promise in clinical cancer treatments. However, it is hard to experimentally identify all drug combinations for synergistic interaction even with high-throughput screening due to the vast space of potential combinations. Although a number of computational methods for drug synergy prediction have proven successful in narrowing down this space, fusing drug pairs and cell line features effectively still lacks study, hindering current algorithms from understanding the complex interaction between drugs and cell lines.ResultsIn this paper, we proposed a Permutable feature fusion network for Drug-Drug Synergy prediction, named PermuteDDS. PermuteDDS takes multiple representations of drugs and cell lines as input and employs a permutable fusion mechanism to combine drug and cell line features. In experiments, PermuteDDS exhibits state-of-the-art performance on two benchmark data sets. Additionally, the results on independent test set grouped by different tissues reveal that PermuteDDS has good generalization performance. We believed that PermuteDDS is an effective and valuable tool for identifying synergistic drug combinations. It is publicly available at https://github.com/littlewei-lazy/PermuteDDS.Scientific contributionFirst, this paper proposes a permutable feature fusion network for predicting drug synergy termed PermuteDDS, which extract diverse information from multiple drug representations and cell line representations. Second, the permutable fusion mechanism combine the drug and cell line features by integrating information of different channels, enabling the utilization of complex relationships between drugs and cell lines. Third, comparative and ablation experiments provide evidence of the efficacy of PermuteDDS in predicting drug-drug synergy.

Research on Monocular Depth Sensing Method Based on Liquid Zoom Imaging

Monocular stereo vision has excellent application prospects in the field of microrobots. On the basis of the geometric model of bifocal imaging, this paper proposes a monocular depth perception method by liquid zoom imaging. Firstly, the configuration of a monocular liquid vision system for depth measurement is presented, and the working mechanism of the system is analyzed through theoretical derivation. Then, to eliminate the influence of optical axis drift induced by the liquid gravity factor on the measurement results, the target image area is used as the calculation feature instead of the image vector length. A target area calculation method based on chain code classification and strip segmentation is proposed. Furthermore, in response to the fluctuation problem of liquid lens focal power caused by factors such as temperature and object distance, a dynamic focal length model of the liquid zoom imaging system is constructed after precise calibration of the focal power function. Finally, a testing experiment is designed to validate the proposed method. The experimental results show that the average error of depth perception methods is 4.30%, and its measurement time is only on the millisecond scale. Meanwhile, the proposed method has good generalization performance.

Seizure detection via deterministic learning feature extraction

Epileptic seizures have a significant impact on the well-being of a large number of individuals worldwide. Utilizing electroencephalographic (EEG) signals for automatic seizure detection proves to be a valuable solution. However, dealing with raw EEG signals is inherently complex, necessitating a preliminary step of feature extraction prior to detection. Traditional feature extraction methods often amalgamate various types of features for seizure detection, as each type typically captures specific properties. In contrast, this paper focuses on detecting seizures by analyzing the system dynamics. The proposed Deterministic Learning Feature Extraction (DLFE) method extracts a single type of nonlinear dynamical feature rooted in the EEG system dynamics. DLFE employs deterministic learning to discern the inherent system dynamics of the EEG under both seizure and normal conditions. Through the feature extraction process, the infinite-dimensional system dynamics are transformed into feature vectors, exhibiting distinct distributions in seizure and normal states. This disparity can be effectively utilized for classification using standard classifiers. The performance of the proposed seizure detection method was assessed using the CHB-MIT and Bonn datasets. The average classification accuracy was found to be 98.63% with a specificity of 99.19% and a sensitivity of 98.06% on CHB-MIT dataset. Compared with the latest similar methods, the accuracy, specificity and sensitivity are improved by 0.31%, 0.21% and 0.05% respectively. Moreover, the performance was achieved with the short-time interval EEG signals within a few channels. The average classification accuracy was found to be 99.90% with a 0.22% improvement on Bonn dataset, which indicates the good generalization performance.

An Adaptive Learning Rate Deep Learning Optimizer Using Long and Short-Term Gradients Based on G–L Fractional-Order Derivative

AbstractDeep learning model is a multi-layered network structure, and the network parameters that evaluate the final performance of the model must be trained by a deep learning optimizer. In comparison to the mainstream optimizers that utilize integer-order derivatives reflecting only local information, fractional-order derivatives optimizers, which can capture global information, are gradually gaining attention. However, relying solely on the long-term estimated gradients computed from fractional-order derivatives while disregarding the influence of recent gradients on the optimization process can sometimes lead to issues such as local optima and slower optimization speeds. In this paper, we design an adaptive learning rate optimizer called AdaGL based on the Grünwald–Letnikov (G–L) fractional-order derivative. It changes the direction and step size of parameter updating dynamically according to the long-term and short-term gradients information, addressing the problem of falling into local minima or saddle points. To be specific, by utilizing the global memory of fractional-order calculus, we replace the gradient of parameter update with G–L fractional-order approximated gradient, making better use of the long-term curvature information in the past. Furthermore, considering that the recent gradient information often impacts the optimization phase significantly, we propose a step size control coefficient to adjust the learning rate in real-time. To compare the performance of the proposed AdaGL with the current advanced optimizers, we conduct several different deep learning tasks, including image classification on CNNs, node classification and graph classification on GNNs, image generation on GANs, and language modeling on LSTM. Extensive experimental results demonstrate that AdaGL achieves stable and fast convergence, excellent accuracy, and good generalization performance.

Short-term Load Forecasting Based on CEEMDAN-LSTM-AdaBoost

High precision load forecasting is of great significance for effectively allocating resources in energy systems and improving energy utilization efficiency. In order to improve the accuracy of load forecasting, this paper proposes a new load forecasting model based on the adaptive noise complete ensemble empirical mode decomposition (CEEMDAN)-long short-term memory neural network (LSTM)-adaptive boosting algorithm (AdaBoost). Firstly, CEEMDAN is used to decompose the original load sequence to obtain a series of eigenmode components, which can reduce the impact caused by the non-stationary nature of data. Then, these eigenmode components are input into LSTM for prediction, and an integrated learning model by AdaBoost is introduced. A strong predictor is constructed through several weak predictors to increase the prediction accuracy. Finally, the prediction result of each component is superimposed and reconstructed to obtain the final prediction result. Compared with other models, the two evaluation indicators of the proposed prediction model have decreased by 35.52% and 76.61% respectively, indicating the good prediction accuracy and generalization performance of the proposed method.

Water level control of nuclear steam generators using intelligent hierarchical autonomous controller

The challenge of water level control in steam generators, particularly at low power levels, has always been a critical aspect of nuclear power plant operation. To address this issue, this paper introduces an IHA controller. This controller employs a CPI controller as the primary controller for direct water level control, coupled with an agent-based controller optimized through a DRL algorithm. The agent dynamically optimizes the parameters of the CPI controller in real-time based on the system’s state, resulting in improved control performance. Firstly, a new observer information is obtained to get the accurate state of the system, and a new reward function is constructed to evaluate the status of the system and guide the agent’s learning process. Secondly, a deep ResNet with good generalization performance is used as the approximator of action value function and policy function. Then, the DDPG algorithm is used to train the agent-based controller, and an advanced controller with good performance is obtained after training. Finally, the popular UTSG model is used to verify the effectiveness of the algorithm. The results demonstrate that the proposed method achieves rise times of 73.9 s, 13.6 s, and 16.4 s at low, medium, and high power levels, respectively. Particularly, at low power levels, the IHA controller can restore the water level to its normal state within 200 s. These performances surpass those of the comparative methods, indicating that the proposed method excels not only in water level tracking but also in anti-interference capabilities. In essence, the IHA controller can autonomously learn the control strategy and reduce its reliance on the expert system, achieving true autonomous control and delivering excellent control performance.