Transformers are fundamental and among the most expensive electrical devices in any power transmission and distribution system. Therefore, it is essential to implement powerful maintenance methods to monitor and predict their condition. Due to its many advantages—such as early detection, accurate diagnosis, cost reduction, and rapid response time—dissolved gas analysis (DGA) is regarded as one of the most effective ways to assess a transformer’s condition. In this contribution, we propose a new probabilistic hierarchical intelligent system consisting of five subnetworks of the radial basis functions (RBF) type. Indeed, hierarchical classification minimizes the complexity of the discrimination task by employing a divide-and-conquer strategy, effectively addressing the issue of unbalanced data (a significant disparity between the categories to be predicted). This approach contributes to a more precise and sophisticated diagnosis of transformers. The first subnetwork detects the presence or absence of defects, separating defective samples from healthy ones. The second subnetwork further classifies the defective samples into three categories: electrical, thermal, and cellulosic decomposition. The samples in these categories are then precisely assigned to their respective subcategories by the third, fourth, and fifth subnetworks. To optimize the hyperparameters of the five models, the Linde–Buzo–Gray algorithm is implemented to reduce the number of centers (radial functions) in each subnetwork. Subsequently, a single-layer perceptron is trained to determine the optimal synaptic weights, which connect the intermediate layer to the output layer. The results obtained with our proposed system surpass those achieved with another implemented alternative (a single RBF), with an average sensitivity percentage as high as 96.85%. This superiority is validated by a Student’s t-test, showing a significant difference greater than 5% (p-value < 0.001). These findings demonstrate and highlight the relevance of the proposed hierarchical configuration.

Air gaps are the principal insulating medium for transmission lines. The atmospheric conditions have an important nonlinear impact on the breakdown voltage prediction, such as temperature and humidity. Unlike previous studies that use a single model, this paper introduces a weighted fusion approach, which has not previously been used to test the air gap insulation performance. This novel strategy assigns different importance levels to each model based on its strengths. The experimental breakdown voltage for plane–plane electrodes air gap is in good agreement with the estimated breakdown voltage using three neural network models: Multilayer Perceptron (MLP), Radial Basis Function (RBF), and Support Vector Regression (SVR). The performance of these models is measured using error metrics such as MSE, MAPE, MSPE, and RMSE, with the RBF model showing the best accuracy, reaching MSE values of 1.4139% for humidity and 1.3855% for temperature. This fusion method significantly improves prediction accuracy, reducing MSE to 0.5378% for humidity and 0.6268% for temperature. The results confirm that the proposed approach enhances prediction reliability and helps improve insulation performance in power systems.

Converting energy from solar panels to electric power includes several advantages; it’s considered clean, green, and available. These points make the system a good, attractive, and reliable alternative to electrical power production. Several factors influence the solar system performance, leading to production power loss. Among these are mismatches resulting from partial shading and electrical breakdowns within the PV array, such as open circuits and short circuits. Therefore, it’s necessary to carefully study these faults to address them quickly and ensure the system continues operating without interruption. This study proposes a new hybrid deep learning model organized into three stages. The proposed system’s strength lies in the introduced techniques based on Bayesian probabilistic reasoning. Stage 1 is a compression stage. It consists of two Variational Auto-encoders (VAE) with unsupervised learning: the first VAE receives the current, voltage, and power signals post-processed by the Discrete Wavelet Transform (DWT) and returns 1000 descriptors instead of 4692, while the second VAE receives the first VAE outputs and returns 200 descriptors. Stage 2 consists of a Cramer-Singer Multi-Class Support Vector Machine (CS-MSVM) type classifier. It gets the second VAE output and returns the liking score for three default categories plus the regular class. Stage 3 is a probabilistic decision-level stage; it post-processes the CS-MSVM outputs to produce probability outputs instead of an all-or-nothing decision. The present system was evaluated and validated using statistical metrics, demonstrating its superiority and effectiveness compared to other intelligent techniques also tested in this work. Additionally, noisy data were incorporated to study environmental phenomena under realistic conditions.