Total real and reactive power losses in electrical power systems are an inevitable phenomenon and occur due to several factors, such as conductor resistance, transformer impedance, line reactance, equipment losses, and phase unbalance. Minimizing them is crucial to the system’s efficiency. In this study, an artificial neural network, specifically a Multi-layer Perceptron, was employed to predict total real and reactive power losses in electrical systems. The network is composed of three layers: an input layer consisting of the variables loading factor, real and reactive power generated on the slack bus, a hidden layer, and an output layer representing the total real and reactive power losses. The training method used was backpropagation, adjusting the weights based on the desired output. The results obtained, using datasets from IEEE systems with 14, 30, and 57 buses, showed satisfactory performance, with a mean squared error of around 10⁻⁴ and a coefficient of determination (R²) of 0.998. In validation with 20% of the data that was not part of the training, the network demonstrated effectiveness, with a mean squared error around 10⁻³. This indicates that the network was able to accurately predict total power losses based on loads, generating estimates close to the desired values.

This paper introduces a parameterization technique to obtain complete P-V curves for electrical power systems under various conditions, including contingencies (N-1, N-1 severe, and N-2) as well as normal operating conditions N-0 (pre-contingency). The distinct feature of this methodology lies in the utilization of the new (θ_k, V_k) plane for constructing the P-V curve. In this new framework, the solution trajectory exhibits a linear characteristic in the vicinity of the maximum loading point (MLP), effectively mitigating any numerical challenges. Another crucial aspect is the point denoted as the center of the set of lines, point O (λ⁰, V_k⁰). This point can be adjusted along the ordinate axis, further enhancing the method’s ability to achieve a smoother ‘flat start’. The ‘flat start’ approach employed in the proposed methodology enables the determination of the loading margin for any system subjected to both simple and severe contingencies. Additionally, it facilitates the use of larger steps in acquiring complete P-V curves with a reduced number of iterations. A procedure presented is to not update the Jacobian matrix at each iteration but only when the system undergoes significant changes. This technique reduced the CPU Ratio by 29.74% in obtaining the P-V curve for all scenarios.

The thermogravimetric analysis (TGA) data of wood present a very similar layout, which makes it difficult to obtain information on components or types of biomass with traditional techniques, which are generally based on the decrease in weight of the sample at specific temperature values. In this work, artificial neural networks are applied as an innovative technique to identify species or differences in components more effectively. This work evaluates the use of TGA data markers (similar to genetic markers) to obtain information on biomass. That is multiple values of the percentage of residual weight with respect to the initial one at specific temperatures of the TGA data. These networks achieve classification by automatically adjusting weights based on training patterns with input marker data. By replicating the curve for the same sample and having unique characteristics, TGA becomes a valuable tool to identify species and characterize biomass composition. The application of artificial intelligence techniques makes it possible to provide detailed information about the components and improve the accuracy of sample classification. The results demonstrate that the neural network successfully classified 95% of wood samples from eight different species and accurately determined the percentage composition with 98% precision.

The application of mathematical models using biotic and abiotic factors for the efficient use of fertilizers to obtain maximum economic productivity can be an important tool to minimize the cost of soybean (Glycine max (L.) Merr.) grain yield. In this sense, using Artificial Neural Networks (ANN) is an important tool in studies involving optimization. This study aimed to estimate soybean yield in Luiziana, Paraná state, Brazil, by considering two growing seasons and an Artificial Neural Network (ANN) as a function of the morphological and nutritional parameters of the plants. Results reveal a well-trained network, with a margin of error of approximately 10-5, thus acting as a tool to estimate soybean data. For the phases, model validation and network test, i.e., data that were not part of the training (validation), the errors averaged 10-3. These results indicate that our approach is adequate for optimizing soybean yield estimates in the area studied.

This work presents the application of an artificial neural network to estimate the total real power losses of an electrical power system. The network used is the Multilayer Perceptron composed of 3 neurons in the input layer (loading factor, real and reactive power generated in the slack bus), 10 neurons in the intermediate layer and 1 neuron in the output layer, representing the total real power losses. The training used is the backpropagation, which uses the desired output (target) to adjust the weights. From the results (IEEE systems of 14 and 30 buses), the network performed well, with mean squared error around 10 ^-4 , and R-value at 0.99. For validation and testing, with 20% of samples, the network proved to be efficient, with MSE also around 10 ^-4 . In this context, the network was able to estimate the total real power losses in function of loading, showing that the obtained output was very close to the one desired.

This paper presents the features of proposed methods that allow the tracing of the P-V curves. With small modifications it becomes possible to use the traditional methods such as the Newton's method, that makes use of the Jacobian matrix, which becomes singular in the maximum loading point (nose point); and of the standard fast decoupled load flow (version XB and BX), believed not to be adapted for the computation of the points below and near the point of maximum loading. These methods are compared to each other with the purpose of pointing out its features. The results obtained for the IEEE systems (14, 30, 57 and 118 buses) are also compared with the ones of a continuation power flow method and of an optimal power flow. It was observed that in all the methods tested the maximum loading points found were always the same.