Short term wind power forecasting using adaptive neuro-fuzzy inference systems

A new intelligent method based on combination of VMD and ELM for short term wind power forecasting

In complex industrial processes (CIPs), due to technical and economic limitations, key performance indicators (KPIs), especially the chemical content-related KPIs, are often difficult to measure in real time, which hinders the propagation of advanced process control technologies. This paper presents a soft sensor-based online KPI inference scheme by a state transition algorithm (STA)-optimized adaptive pre-sparse neuro-fuzzy inference system model, called STA-APSNFIS. It introduces a pre-sparse neural network to the traditional adaptive neuro-fuzzy inference system (ANFIS) model to establish an adaptive pre-sparse neuro-fuzzy inference system (APSNFIS) model to alleviate the adverse effects of data redundancy and noise interference in the detectable process monitoring data, which can effectively reduce the complexity of neuro-fuzzy inference system (NFIS) and speed up its convergence. Successively, to avoid being trapped at a local optimum, the STA-based optimization algorithm is adopted to replace the traditional gradient-based optimization approach to achieve an optimal APSNFIS model. Extensive validation and comparative experiments on nonlinear numeric simulation systems, benchmark Tenessee Eastman (TE) process and a real industrial bauxite flotation process demonstrated that the proposed STA-APSNFIS performed favorably against traditional ANFIS model as well as its variants, e.g., PSO-ANFIS, GA-ANFIS, and some other soft sensor-based KPI inference models.

Short-term wind power forecasting is considered a critical challenge in smart grid management due to the nonlinear, unstable, and multi-scale noise characteristics of wind signals. Although recent advances in hybrid deep learning have improved the accuracy of short-term wind power forecasting, many state-of-the-art models usually consider signal denoising, residual decomposition, and attention mechanisms as independent modules without providing a unified solution. This paper proposes an end-to-end solution, W-HiTS-Attention (Wavelet Transform, N-stacked Hierarchical Interpolation for Time Series, Attention), which coherently integrates wavelet denoising, hierarchical residual learning from N-HiTS (Neural Hierarchical Interpolation), and an in-block self-attention mechanism. The proposed solution outperforms 21 benchmarks in accuracy, including state-of-the-art baselines such as N-BEATS, N-HiTS, TCN, Informer, Autoformer, LSTM, BiLSTM, GRU, and Prophet, achieving an RMSE of 55.56 W and an R2 of 0.9918. Furthermore, the results show that the proposed solution is efficient in terms of parameter count (0.033M), latency (0.0036 ms/sample), and training time, making it promising for low-latency inference in resource-constrained environments. The results show that the coherent integration of frequency preprocessing, hierarchical residual forecasting, and attention-based temporal refinement provides a robust, explainable, and deployable solution for short-term wind power forecasting.

Wind power generation is highly sensitive to meteorological conditions, leading to strong volatility and limited controllability. Accurate short-term wind power forecasting is therefore critical for maintaining power system stability and improving the utilization of renewable energy. In this work, we propose an enhanced forecasting model termed Particle Swarm Optimization Residual Structure Convolutional Neural Network Long Short Term Memory (PR-CNN-LSTM), which extends the conventional CNN-LSTM architecture. The model first employs CNN-LSTM to capture local patterns and temporal dependencies across different forecasting horizons. Residual structures are then introduced into both the convolutional and recurrent modules to stabilize deep network training and enhance representation learning. Furthermore, particle swarm optimization is incorporated to dynamically tune key model hyperparameters, enabling adaptive selection of an optimal parameter configuration. Extensive experiments demonstrate that PR-CNN-LSTM consistently outperforms representative baseline models under all evaluated forecasting scenarios. In single-step forecasting, the proposed model achieves an R 2 exceeding 0.97, indicating strong accuracy and robustness. These results suggest that the proposed approach provides effective technical support for smart grid scheduling, power balance management, and the large-scale integration of wind energy.

Accurate short-term wind power forecast technique plays a crucial role in maintaining the safety and economic efficiency of smart grids. Although numerous studies have employed various methods to forecast wind power, there re... | Find, read and cite all the research you need on Tech Science Press

The wind energy is the most encouraging and developed strategy among the sustainable energy source assets but its intermittency makes it difficult for schedule management. The wind power has been integrated into the electric grid due to its irregularity and uncertainty. Therefore, endeavors ought to be made in predicting the wind behavior and its relating electric generation. The wind power forecasting is required to cope with the stated challenges. For short term wind power forecasting, a hybrid intelligent technique is proposed in this paper. The proposed methodology is the hybridization of wavelet transform (WT) and particle swarm optimization (PSO) with adaptive-network-based fuzzy inference system (ANFIS). The average hourly time series data of one year has been taken from the wind park located at Agar in USA, which is first decomposed and then applied to the optimized ANFIS model using particle swarm optimization to predict the wind power output. The proposed hybrid Wavelet-Neuro-Fuzzy-PSO (WNFP) approaches as compared to other intelligent approaches show its superiority with respect to the wind power prediction.

In this paper, we compare and analyze the prediction performance of existing ANFIS (Adaptive Neuro Fuzzy Inference System) models. The ANFIS model is a fusion of neural networks with learning ability, adaptability, and computational ability, and a fuzzy system with human-like thinking ability and reasoning ability. In this paper, we analyze the prediction performance of ANFIS models using different input space partitioning methods. ANFIS 1 uses SC (Subtractive Clustering), a clustering method based on the density of data, to calculate the center of the cluster by calculating potential values inversely proportional to the distances of the different inputs. ANFIS 2 uses Fuzzy C-Means (FCM) clustering, which is a clustering method based on degree of belonging to data, and classifies each data belonging to each cluster according to degree of affiliation. ANFIS 3 is a method of using Context based Fuzzy C-Means (CFCM) clustering, which is a clustering method considering characteristics of input and output space based on FCM clustering. It creates clusters for each context generated according to characteristics of output space, to classify the data. To evaluate these ANFIS models, we use short term electricity price forecasting data and Boston housing data. The prediction performance is based on the Root Mean Squared Error (RMSE) method. Experimental results show that the prediction performance of the ANFIS model is different depending on the characteristics of each data. Furthermore, we confirmed that the prediction performance of ANFIS model using CFCM clustering method is constant.

The importance of the very short term wind power forecasting has increased in the last years because of the large-scale integration of the wind power plants in the electricity grid. In addition to the other functionalities of the Wind Power Monitoring and Forecast Center (RITM) of Turkey, a very short term wind power forecasting module has been implemented and verified for 14 wind power plants (WPPs) in real time. Wind power monitoring system of RITM has been used as data provider to the forecasting module and results are presented to the clients via graphical user interfaces. The results show that autoregressive model (AR) and autoregressive with exogenous variable model (ARX) beat persistence model after 20 minutes for all the forecasted horizon between 0-6 hours.

Incorporating wind energy on a large scale into power systems presents challenges for the operation and control of the grid. To enhance the safety of power grid operation, accurate short-term forecasting of wind power is necessary, as it minimizes the impact of randomness. Considering the uncertainty and prediction issues associated with wind power, this paper introduces a CNN–GRU ultra-short-term wind power prediction model. This model relies on multichannel signals, including data such as wind speed, wind direction, climate conditions, and historical power outputs collected from wind farms. These data types contribute to the formation of a comprehensive multichannel signal for wind power. Following that, the CNN method extracts both global and partial features from these signals. Concurrently, features are extracted from past power outputs based on their time series. These features are then combined with the ones obtained from the convolution process. Subsequently, these combined features are input into a fully connected network. This step is crucial for blending multichannel information and generating forecast results. To validate the model, it was tested using data from a wind farm located in a specific region of Sichuan Province. According to our experimental results, the model demonstrates a high level of accuracy in computation and robust generalization ability.

Comparison of ANFIS and NN models—With a study in critical buckling load estimation

Electrochemical-kinetics, MD-simulation and multi-input single-output (MISO) modeling using adaptive neuro-fuzzy inference system (ANFIS) prediction for dexamethasone drug as eco-friendly corrosion inhibitor for mild steel in 2 M HCl electrolyte

In this paper, a hybrid meta-heuristic artificial intelligence (AI) method, named “ANFIS–PSOGA,” is developed for modeling the discharge coefficient of circular side orifices for the first time. For this purpose, the adaptive neuro fuzzy inference system (ANFIS) network is optimized by means of two robust optimization algorithms: the genetic algorithm (GA) and the particle swarm optimization (PSO) algorithm. Furthermore, the Monte Carlo simulations are utilized to enhance the performance of the AI hybrid model. Besides, the k-fold cross-validation method is employed in order to verify the numerical results. Initially, the performance of three fuzzy inference system (FIS) generations of the ANFIS model including grid partitioning, subtractive clustering and fuzzy C-means clustering are evaluated. After that, using the parameters affecting the discharge coefficient, eleven ANFIS models are defined and then the best ANFIS model and the most influencing input parameters are identified through the conduction of a sensitivity analysis. For the superior ANFIS model, mean absolute percentage error, root-mean-square error and the correlation coefficient (R) are obtained 0.021, 0.020 and 0.871, respectively. Also, the Froude number (Fr) is detected as the most effective input parameter. Moreover, the discharge coefficient of the circular side orifices and the flow field within main channels along the side orifices are simulated through the adoption of a computational fluid dynamics (CFD) model. Additionally, the flow field at three different cross sections (beginning, middle and end of orifice) along the side weir is evaluated. For instance, by advancing the flow toward the side orifice, the Froude number is reduced. Besides, at each cross section, the maximum shear stress is estimated to occur in the vicinity of the crest of the circular side orifice. According to the CFD results, the maximum lateral velocity is estimated near the circular side orifice. In addition, the results yielded by the ANFIS, ANFIS–PSO, ANFIS–PSOGA and CFD models are compared and an uncertainty analysis is carried out for the models. Based on the comparison, ANFIS–PSOGA has the best performance in simulating the discharge coefficient. Ultimately, a simple and practical computer code for the best model (ANFIS–PSOGA) is proposed which can be easily used by scholars and engineers in practical usages.

This paper presents the results obtained from a research work investigating the performance of different Adaptive Neuro-Fuzzy Inference System (ANFIS) models developed to predict excitation forces on a dynamically loaded flexible structure. For this purpose, a flexible structure is equipped with ac celeration transducers at each degree of freedom and a force transducer for validation and training. Two types of models were developed; the first type uses current accelerations only while the second type considers both the current accelerations and the historical values. The models are trained using data obtained from applying a random excitation force on the flexible structure. The performance of the developed models is evaluated by analyzing the prediction capabilities based on a normalized prediction error. The frequency domain is considered to analyze the similarity of the frequencies in the predicted and the original force signal. For a selection of the best models, a more advanced performance analysis is carried out. This includes application of the trained models to deterministic and non-deterministic excitation forces with different excitation frequencies and amplitudes. Additionally, the influence of the sampling frequency and sensor location on the model performance is investigated. The results obtained in this paper show that ANFIS models can be used to set up reliable force predictors for dynamical loaded flexible structures, when a certain degree of inaccuracy is accepted. Furthermore, the comparison study points out that the transducer location is crucial for the model performance. However, there exists no general solution for the final selection of models. The findings showed that the model type employing historical values gives better predictions when operating in their trained regions while the models using only current values have generally higher prediction errors in trained regions but are less sensitive to changes of the system dynamics history.

The shear strength (Su) of soils is one of the most widely used parameters for designing structures safely, where Su is found with the unconfined compression test (UCS). Although UCS can be acquired by performing uniaxial compression test it would be extremely helpful to predict the UCS without performing any compression test, namely, using computational methods considering different parameters of soils such as consistency limits, fine grain ratio, liquidity index, and void ratio. The goal of present work is predicting UCS taking into account these soil parameters with the aid of developed Adaptive Neuro-Fuzzy Inference System (ANFIS) model and the Multiple Linear Regression (MLR) analyses. On the other hand, the effect of the size of the training set of designed models on the results is examined, also. For this aim, four different models composed of different training and test set ratios have been constructed and analyzed using ANFIS and MLR. It is concluded that UCS can be predicted using MLR analysis and ANFIS model with best 0.76 and 0.91 values of determination coefficient (R2) around the x = y line respectively, and the effect of the size of the training set of models on ANFIS is more pronounced than MLR models.

ANFIS for analysis friction factor and Nusselt number of pulsating nanofluids flow in the fluted tube under magnetic field