This paper presents a hybrid deep learning and adaptive residual framework for accurate solar power forecasting and PV system reliability enhancement using real-world operational data acquired from two grid-connected PV power plants (1 MW and 100 MW) located in Chhattisgarh, India. Initially, three deep learning models—Long Short-Term Memory (LSTM), Convolutional Neural Network-based BiLSTM (CNN-BiLSTM), and Gated Recurrent Unit (GRU)—are developed and evaluated for short-term solar power forecasting. The models incorporate temporal feature engineering, lagged inputs, rolling statistical features, and meteorological parameters to capture the nonlinear dynamics of PV generation Validation on both PV sites demonstrates that the GRU model consistently provides the best generalization performance, achieving R² values of 0.996 and 0.995 for respective PV sites, with low normalized forecasting errors (NRMSE of 0.0152 and 0.0143, respectively), thereby outperforming the LSTM and CNN–BiLSTM models across PV plants of different capacities. Building upon the forecasting results, a data-driven fault detection approach is implemented using residual analysis. An adaptive thresholding technique based on rolling statistical measures is proposed to dynamically identify anomalies under varying operating conditions. The proposed framework employs realistic stochastic fault simulation incorporating gradual soiling, partial shading, combined soiling–shading faults, inverter degradation, severe module degradation, sensor drift, measurement noise, and overlapping fault conditions to emulate practical PV operating environments. An ensemble bagging classifier is subsequently employed to classify seven operating states comprising healthy operation and six fault categories. Under these realistic conditions, the proposed classifier achieves an overall classification accuracy of 95.95% with a mean F1-score of 0.946, demonstrating robust fault discrimination despite measurement uncertainty and heterogeneous fault characteristics. Furthermore, a predictive maintenance framework is introduced to prioritize maintenance actions into healthy operation, cleaning recommended, inspection required, urgent maintenance, and component replacement recommended based on adaptive residual severity. Overall, the proposed integrated forecasting, fault diagnosis, and predictive maintenance framework provides a reliable and scalable solution for intelligent condition monitoring and maintenance planning of utility-scale PV systems.
The sustainable valorization of underutilized fruit as a source of high-value bioactive compounds is gaining significant attention in the food, nutraceutical, and pharmaceutical sectors. This study investigates the ultrasound-assisted extraction (UAE) for the recovery of gallic acid and phenolic content from Fragrant Manjack fruit and optimization using the Box-Behnken design – response surface methodology. The effects of extraction time (5–25 min), ultrasonic power (130–520 W), duty cycle (30–70 • Optimization of UAE from Fragrant Manjack fruits using response surface methodology. • Ultrasound-assisted extraction enhanced the recovery of gallic acid and total phenolics. • Syringic acid, gallic acid, protocatechuic acid, and caffeic acid were polyphenols identified by LC-MS. • Mathematical models were designed using extraction time, US power, duty cycle, SSR. • FESEM images showed surface morphology before after UAE.
Zika virus primarily spreads through the bites of Aedes aegypti mosquitoes and may also be transmitted from an infected woman to her fetus, through human contact, and from humans to mosquitoes. Traditional non-fractional models and bilinear incidence rates fail to account for memory-dependent dynamics and the psychological effects resulting from delayed immune responses and interventions. Therefore, we propose a Caputo fractional-order model that incorporates saturated incidence, multiple transmission pathways, and vertical transmission of the Zika virus. The existence and uniqueness of solutions are established by fixed-point theory. We explore the equilibrium states with their stability. The model is validated on weekly data from the 2018 Zika outbreak in Brazil. Numerical simulations performed using the Adams predictor-corrector method to show the effect of memory, vertical transmission, and multiple transmission routes on the spread of disease. Local and global sensitivity analyses identify the key parameter that influences the reproduction number. Numerical simulations and sensitivity reveal that transmission rate, medical treatment, and vector control are key sensitive parameters. We introduce a fractional-order optimal control model that incorporates prevention, treatment, and vector control. An efficiency and cost-effective analysis of seven different policies revealed that the policy with all three controls is the most effective, with the highest efficiency index of 92.19% . In comparison, a policy focused exclusively on prevention measures is the most cost-effective. These insights will help public health authorities develop effective strategies to combat Zika outbreaks.
Perovskite solar cells (PSCs) have gained attention in photovoltaic research due to their high power conversion efficiencies (PCEs) and low production costs. However, the utility of most efficient PSCs, those incorporating halide perovskites, is limited by their high toxicity and poor long-term stability. To address these issues, chalcogenide perovskites have emerged as a promising alternative. Chalcogenide perovskites are non-toxic and offer excellent long-term durability, improved chemical stability, and superior light absorption. This work analyses single-junction solar cells using BaZrS3 and CaZrSe3 absorbers and explores a multijunction design combining both, aiming to overcome the limited photon absorption in single-layer PSCs that hinders scalability and efficiency in next-generation solar applications. A configuration comprising a double perovskite active layer has been introduced and modeled utilizing the SCAPS-1D tool. The cells designed with BaZrS3 as absorber layer, TiO₂ as the electron transport layer and CuGaO2 as the hole transport layer achieved a PCE of 22.83
In the current study, the authors develop an eco-epidemic model that incorporates fear effects, prey refuge, and nonlinear harvesting terms, and considers the predator biomass to be affected by disease. This study analyses the positivity, boundedness, and stability properties of the system and, by deriving the basic reproduction number, also known as the disease invasion number and the predator invasion number, quantifies the critical thresholds governing disease transmission and predator persistence. This research also investigates the occurrence of transcritical and Hopf bifurcations at different equilibria and determines the direction and stability of the Hopf bifurcation. To identify the effect of the movement of prey-predator species, a diffusion term is incorporated to establish the conditions for Turing instability, leading to the emergence of pattern formation. Turing patterns like labyrinthine stripes, stripe-spot mixtures, and isolated spots are observed as the diffusion parameter of the infected species varies. Also, this study observes that augmenting the levels of fear and refuge reduces the infected population and contributes to the destabilisation of the system. Furthermore, our investigation reveals that an increasing refuge level leads the system to exhibit stable, periodic, period-doubling, and chaotic behaviours. To verify and validate the analytical findings, numerical simulations have been performed.
