The randomness and volatility of wind power greatly affect the safety and economy of the power systems, and the wake effect of the wind farm aggravates the wind energy loss and the wind power fluctuation. Taking into consideration the wake effect of the wind farm, a new coordinated wind power smoothing control strategy for multi-wind turbines (M-WT) and energy storage systems (ESS) is proposed. The proposed method is based on a multi-agent deep reinforcement learning (MADRL), in which the relationship between output power and wake effect is firstly analyzed, and a power smoothing control model of the M-WT and ESS is established. MADRL is then introduced to optimize the power control of M-WT and ESS. In order to further increase the learning and training efficiency, an improved MADRL algorithm based on the partitioned experience buffer and prioritized experience replay is proposed, where the experience buffer is divided into positive, negative, and neutral experiences, and the experiences are sampled according to experience priority. The effectiveness of the proposed strategy is verified on the SimWindFarm platform. The results show that the proposed control strategy can maximize the economic benefits while further smoothing wind power fluctuations and increasing power generation.
To reduce the influence of the random fluctuation on wind power prediction, a new ultra-short-term wind power prediction model, based on wavelet decomposition (WD), variational mode decomposition (VMD), and least-squares support vector machine (LSSVM), is proposed in this paper. The method is based on the double decomposition and LSSVM, where the wind power sequence is decomposed by WD into low- and high-frequency components, which are further decomposed by VMD to obtain many modal components with tendency and periodicity. Multiple LSSVM prediction models are then established with historical wind power data and weather data as the inputs to obtain the predicted values of the multiple modal components. The final predicted values of wind power are achieved by data fusion of outputs of these LSSVM models. The experimental results show that the MAPE (mean absolute percentage error) of the combined prediction model is 4.66%, which is the best compared with nine benchmark models. This demonstrates the high performance of the proposed WD-VMD-LSSVM model for short-term prediction of wind power.
Kubernetes, known for its versatility in infrastructure management, rapid scalability, and ease of deployment, makes it an excellent platform for edge computing. However, its native scheduling algorithm struggles with load balancing, especially during peak task deployment in edge environments characterized by resource limitations and low latency demands. To address this issue, a proximal policy optimization with the least response time (PPO-LRT) algorithm was proposed in this paper. This deep reinforcement learning approach learns the pod-scheduling process, which can adaptively schedule edge tasks to the most suitable worker nodes with the shortest response time according to the current cluster load and pod state. To evaluate the effectiveness of the proposed algorithm, multiple virtual machines were created, and we built a heterogeneous node cluster. Additionally, we deployed k3s, a Kubernetes distribution suited for edge environments, on the cluster. The load balancing, high load resilience, and average response time during peak task deployment were tested by initiating numerous tasks within a limited time frame. The results validate that the PPO-LRT-based scheduler shows superior performance in cluster load balancing compared to the Kube scheduler. After the deployment of 500 random tasks, several cluster nodes become overwhelmed by using the Kube scheduler, whereas the PPO-LRT-based scheduler evenly allocates the workload across the cluster, reducing the average response time by approximately 31%.
UVB irradiation can cause acute damage such as sunburn, or photoaging and melanoma, all of which are major health threats. This study was designed to investigate the mechanism of skin photoaging induced by UVB radiation in mice through the analysis of the differential expression of miRNAs. A UVB irradiation photoaging model was constructed. HE and Masson special stains were used to examine the modifications in the epidermis and dermis of mice. The miRNA expression profiles of the mouse skin model exposed to UVB radiation and the normal skin of mice were analyzed using miRNA-sequence analysis. GO and Pathway analysis were employed for the prediction of miRNA targets. A total of 23 miRNAs were evaluated for significantly different expressions in comparison to normal skin. Among them, 7 miRNAs were up-regulated and 16 were down-regulated in the skin with photoaging of mice exposed to UVB irradiation. The differential expression of miRNA is related to a variety of signal transduction pathways, among which mmu-miR-195a-5p and mitogen-activated protein kinase (MAPK) signal pathways are crucial. There was a significant differential expression of miRNA in the skin of normal mice in comparison with the skin with photoaging induced by UVB irradiation. Due to time and energy constraints, the specific protein level verification, MAPK pathway exploration, and miR-195a-5p downstream molecular mechanism need to be further studied in the future. UVB-induced skin photoaging can be diagnosed and treated using miRNA.
It is difficult to identify the working conditions of the rotary kilns due to the harsh environment in the kilns. The flame images of the firing zone in the kilns contain a lot of working condition information, but the flame image data sample size is too small to be used to fully extract the key features. In order to solve this problem, a method combining transfer learning and attention mechanism is proposed to extract key features of flame images, in which the deep residual network is used as the backbone network, the coordinate attention module is introduced to capture the position information and channel information on the branch of feature graphs, and the features of flame images obtained are further screened to improve the extraction ability. At the same time, migration learning is performed by the pre-trained ImageNet data set, and feature migration and parameter sharing are realized to cope with the training difficulty of a small sample data size. Moreover, an activation function Mish is introduced to reduce the loss of effective information. The experimental results show that, compared with traditional methods, the working condition recognition accuracy of rotary kilns is improved by about 5% with the proposed method.