Power transformer is one of the most crucial devices in power grid. It is significant to determine incipient faults of power transformers fast and accurately. Input features play critical roles in fault diagnosis accuracy. In order to further improve the fault diagnosis performance of power transformers, a random forest feature selection method coupled with optimized kernel extreme learning machine is presented in this study. Firstly, the random forest feature selection approach is adopted to rank 42 related input features derived from gas concentration, gas ratio and energy-weighted dissolved gas analysis. Afterwards, a kernel extreme learning machine tuned by the Aquila optimization algorithm is implemented to adjust crucial parameters and select the optimal feature subsets. The diagnosis accuracy is used to assess the fault diagnosis capability of concerned feature subsets. Finally, the optimal feature subsets are applied to establish fault diagnosis model. According to the experimental results based on two public datasets and comparison with 5 conventional approaches, it can be seen that the average accuracy of the proposed method is up to 94.5%, which is superior to that of other conventional approaches. Fault diagnosis performances verify that the optimum feature subset obtained by the presented method can dramatically improve power transformers fault diagnosis accuracy.

The imbalance of dissolved gas analysis (DGA) data will lead to over-fitting, weak generalization and poor recognition performance for fault diagnosis models based on deep learning. To handle this problem, a novel transformer fault diagnosis method based on improved auxiliary classifier generative adversarial network (ACGAN) under imbalanced data is proposed in this paper, which meets both the requirements of balancing DGA data and supplying accurate diagnosis results. The generator combines one-dimensional convolutional neural networks (1D-CNN) and long short-term memories (LSTM), which can deeply extract the features from DGA samples and be greatly beneficial to ACGAN’s data balancing and fault diagnosis. The discriminator adopts multilayer perceptron networks (MLP), which prevents the discriminator from losing important features of DGA data when the network is too complex and the number of layers is too large. The experimental results suggest that the presented approach can effectively improve the adverse effects of DGA data imbalance on the deep learning models, enhance fault diagnosis performance and supply desirable diagnosis accuracy up to 99.46%. Furthermore, the comparison results indicate the fault diagnosis performance of the proposed approach is superior to that of other conventional methods. Therefore, the method presented in this study has excellent and reliable fault diagnosis performance for various unbalanced datasets. In addition, the proposed approach can also solve the problems of insufficient and imbalanced fault data in other practical application fields.

In order to further improve the accuracy of photovoltaic (PV) power prediction and the stability of power system, a short-term PV power prediction model based on hierarchical clustering of K-means++ algorithm and deep learning hybrid model is proposed in this paper. First, hierarchical clustering of the K-means++ algorithm is used to cluster historical data into different weather scenes according to different seasons. Second, a hybrid model combining convolutional neural network (CNN), squeeze-and-excitation attention mechanism (SEAM), and bidirectional long short-term memory (BILSTM) neural network is constructed to capture long-term dependencies in time series, and the improved pelican optimization algorithm (IPOA) is used to optimize the hyperparameters in the prediction model. Finally, an example for modeling analysis is conducted by using the actual output and meteorological data of a PV power station in the Ili region of Xinjiang, China. The effectiveness and accuracy of the proposed model are verified by comparing with LSTM, BILSTM, CNN-BILSTM, and POA-CNN-SEAM-BILSTM models, and the superiority of IPOA is verified by comparing with particle swarm optimization and whale optimization algorithm. The results show that the proposed model can obtain better results under different weather scenes in different seasons, and the prediction accuracy of the model optimized by IPOA is further improved.

To address the issue of traditional static evaluation models being unable to comprehensively analyze the performance of ultra-supercritical coal-fired units under varying loads, we propose a dynamic comprehensive evaluation model based on the improved Criteria Importance Through Inter-criteria Correlation (CRITIC) method and entropy weight method (EWM). The comprehensive performance evaluation index system of ultra-supercritical coal fired units is constructed by examining the boiler performance, turbine performance, plant power performance, environmental performance, and flexible performance of coal-powered units. The CRITIC and EWM methods are used to calculate the weights of the indicators, which are then combined with the static evaluation results. Using a dynamic comprehensive evaluation model, we analyze ultra-supercritical coal-fired units, taking into account time weight. This allows us to obtain the comprehensive dynamic real-time evaluation value of the units under different loads. The research indicates that the weight of the evaluation index is changed when using the dynamic comprehensive evaluation model of the improved CRITIC and EWM. The index with lower weight is increased by 6.2%, while the index with higher weight is decreased by 0.22%. This alteration in weight range can provide a more objective reflection of the relationship between evaluation indicators. This model offers significant advantages in improving evaluation accuracy, weight balance distribution, and generality.

Aiming at the problem that the energy consumption of the boiler system varies greatly under the flexible peaking requirements of coal-fired units, an energy consumption prediction model for the boiler system is established based on a Least-Squares Support Vector Machine (LSSVM). First, the Mean Impact Value (MIV) algorithm is used to simplify the input characteristics of the model and determine the key operating parameters that affect energy consumption. Secondly, the Snow Ablation Optimizer (SAO) with tent map, adaptive t-distribution, and the opposites learning mechanism is introduced to determine the parameters in the prediction model. On this basis, based on the operation data of an ultra-supercritical coal-fired unit in Xinjiang, China, the boiler energy consumption dataset under variable load is established based on the theory of fuel specific consumption. The proposed prediction model is used to predict and analyze the boiler energy consumption, and a comparison is made with other common prediction methods. The results show that compared with the LSSVM, BP, and ELM prediction models, the average Relative Root Mean Squared Errors (aRRMSE) of the LSSVM model using ISAO are reduced by 2.13%, 18.12%, and 40.3%, respectively. The prediction model established in this paper has good accuracy. It can predict the energy consumption distribution of the boiler system of the ultra-supercritical coal-fired unit under variable load more accurately.