Innovative second-generation wavelets construction with recurrent neural networks for solar radiation forecasting

A multilayer perceptron neural network approach for optimizing solar irradiance forecasting in Central Africa with meteorological insights

Promoting renewable energy sources, particularly in the solar industry, has the potential to address the energy shortfall in Central Africa. Nevertheless, a difficulty occurs due to the erratic characteristics of solar irradiance data, which is influenced by climatic fluctuations and challenging to regulate. The current investigation focuses on predicting solar irradiance on an inclined surface, taking into consideration the impact of climatic variables such as temperature, wind speed, humidity, and air pressure. The used methodology for this objective is Artificial Neural Network (ANN), and the inquiry is carried out in the metropolitan region of Douala. The data collection device used in this research is the meteorological station located at the IUT of Douala. This station was built as a component of the Douala sustainable city effort, in partnership with the CUD and the IRD. Data was collected at 30-min intervals for a duration of around 2 years, namely from January 17, 2019, to October 30, 2020. The aforementioned data has been saved in a database that underwent pre-processing in Excel and later employed MATLAB for the creation of the artificial neural network model. 80% of the available data was utilized for training the network, 15% was allotted for validation, and the remaining 5% was used for testing. Different combinations of input data were evaluated to ascertain their individual degrees of accuracy. The logistic Sigmoid function, with 50 hidden layer neurons, yielded a correlation coefficient of 98.883% between the observed and estimated sun irradiation. This function is suggested for evaluating the intensities of solar radiation at the place being researched and at other sites that have similar climatic conditions.

Chain-Structure Echo State Network With Stochastic Optimization: Methodology and Application

In this article, a chain-structure echo state network (CESN) with stacked subnetwork modules is newly proposed as a new kind of deep recurrent neural network for multivariate time series prediction. Motivated by the philosophy of "divide and conquer," the related input vectors are first divided into clusters, and the final output results of CESN are then integrated by successively learning the predicted values of each clustered variable. Network structure, mathematical model, training mechanism, and stability analysis are, respectively, studied for the proposed CESN. In the training stage, least-squares regression is first used to pretrain the output weights in a module-by-module way, and stochastic local search (SLS) is developed to fine-tune network weights toward global optima. The loss function of CESN can be effectively reduced by SLS. To avoid overfitting, the optimization process is stopped when the validation error starts to increase. Finally, SLS-CESN is evaluated in chaos prediction benchmarks and real applications. Four different examples are given to verify the effectiveness and robustness of CESN and SLS-CESN.

Cloud Cover Forecast Based on Correlation Analysis on Satellite Images for Short-Term Photovoltaic Power Forecasting

Photovoltaic power generation must be predicted to counter the system instability caused by an increasing number of photovoltaic power-plant connections. In this study, a method for predicting the cloud volume and power generation using satellite images is proposed. Generally, solar irradiance and cloud cover have a high correlation. However, because the predicted solar irradiance is not provided by the Meteorological Administration or a weather site, cloud cover can be used instead of the predicted solar radiation. A lot of information, such as the direction and speed of movement of the cloud is contained in the satellite image. Therefore, the spatio-temporal correlation of the cloud is obtained from satellite images, and this correlation is presented pictorially. When the learning is complete, the current satellite image can be entered at the current time and the cloud value for the desired time can be obtained. In the case of the predictive model, the artificial neural network (ANN) model with the identical hyperparameters or setting values is used for data performance evaluation. Four cases of forecasting models are tested: cloud cover, visible image, infrared image, and a combination of the three variables. According to the result, the multivariable case showed the best performance for all test periods. Among single variable models, cloud cover presented a fair performance for short-term forecasting, and visible image presented a good performance for ultra-short-term forecasting.

Ultra-Short-Term Forecasting of Photo-Voltaic Power via RBF Neural Network

With the fast expansion of renewable energy systems during recent years, the stability and quality of smart grids using solar energy have been challenged because of the intermittency and fluctuations. Hence, forecasting photo-voltaic (PV) power generation is essential in facilitating planning and managing electricity generation and distribution. In this paper, the ultra-short-term forecasting method for solar PV power generation is investigated. Subsequently, we proposed a radial basis function (RBF)-based neural network. Additionally, to improve the network generalization ability and reduce the training time, the numbers of hidden layer neurons are limited. The input of neural network is selected as the one with higher Spearman correlation among the predicted power features. The data are normalized and the expansion parameter of RBF neurons are adjusted continuously in order to reduce the calculation errors and improve the forecasting accuracy. Numerous simulations are carried out to evaluate the performance of the proposed forecasting method. The mean absolute percentage error (MAPE) of the testing set is within 10%, which show that the power values of the following 15 min. can be predicted accurately. The simulation results verify that our method shows better performance than other existing works.

Deep Learning Method Based on Gated Recurrent Unit and Variational Mode Decomposition for Short-Term Wind Power Interval Prediction

Wind power interval prediction (WPIP) plays an increasingly important role in evaluations of the uncertainty of wind power and becomes necessary for managing and planning power systems. However, the intermittent and fluctuating characteristics of wind power mean that high-quality prediction intervals (PIs) production is a challenging problem. In this article, we propose a novel hybrid model for the WPIP based on the gated recurrent unit (GRU) neural networks and variational mode decomposition (VMD). In the hybrid model, VMD is employed to decompose complex wind power data into simplified modes. Basic GRU prediction models, comprising a GRU input layer, multiple fully connected layers, and a rank-ordered terminal layer, are then trained for each mode to produce PIs, which are combined to obtain final PIs. In addition, an adaptive optimization method based on constructed intervals (CIs) is proposed to build high-quality training labels for supervised learning with the hybrid model. Several numerical experiments were implemented to validate the effectiveness of the proposed method. The results indicate that the proposed method performs better than the traditional interval prediction models with much higher quality PIs, and it requires less training time.

Multi-Step Solar Irradiance Forecasting and Domain Adaptation of Deep Neural Networks

The problem of forecasting hourly solar irradiance over a multi-step horizon is dealt with by using three kinds of predictor structures. Two approaches are introduced: Multi-Model (MM) and Multi-Output (MO). Model parameters are identified for two kinds of neural networks, namely the traditional feed-forward (FF) and a class of recurrent networks, those with long short-term memory (LSTM) hidden neurons, which is relatively new for solar radiation forecasting. The performances of the considered approaches are rigorously assessed by appropriate indices and compared with standard benchmarks: the clear sky irradiance and two persistent predictors. Experimental results on a relatively long time series of global solar irradiance show that all the networks architectures perform in a similar way, guaranteeing a slower decrease of forecasting ability on horizons up to several hours, in comparison to the benchmark predictors. The domain adaptation of the neural predictors is investigated evaluating their accuracy on other irradiance time series, with different geographical conditions. The performances of FF and LSTM models are still good and similar between them, suggesting the possibility of adopting a unique predictor at the regional level. Some conceptual and computational differences between the network architectures are also discussed.

Performance Evaluation of Probabilistic Methods Based on Bootstrap and Quantile Regression to Quantify PV Power Point Forecast Uncertainty

This paper presents two probabilistic approaches based on bootstrap method and quantile regression (QR) method to estimate the uncertainty associated with solar photovoltaic (PV) power point forecasts. Solar PV output power forecasts are obtained using a hybrid intelligent model, which is composed of a data filtering technique based on wavelet transform (WT) and a soft computing model (SCM) based on radial basis function neural network (RBFNN) that is optimized by particle swarm optimization (PSO) algorithm. The point forecast capability of the proposed hybrid WT+RBFNN+PSO intelligent model is examined and compared with other hybrid models as well as individual SCM. The performance of the proposed bootstrap method in the form of probabilistic forecasts is compared with the QR method by generating different prediction intervals (PIs). Numerical tests using real data demonstrate that the point forecasts obtained from the proposed hybrid intelligent model can be effectively used to quantify PV power uncertainty. The performance of these two uncertainty quantification methods is assessed through reliability.

Intelligent PV Panels Fault Diagnosis Method Based on NARX Network and Linguistic Fuzzy Rule-Based Systems

The expanding use of photovoltaic (PV) systems as an alternative green source for electricity presents many challenges, one of which is the timely diagnosis of faults to maintain the quality and high productivity of such systems. In recent years, various studies have been conducted on the fault diagnosis of PV systems. However, very few instances of fault diagnostic techniques could be implemented on integrated circuits, and these techniques require costly and complex hardware. This work presents a novel and effective, yet small and implementable, fault diagnosis algorithm based on an artificial intelligent nonlinear autoregressive exogenous (NARX) neural network and Sugeno fuzzy inference. The algorithm uses Sugeno fuzzy inference to isolate and classify faults that may occur in a PV system. The fuzzy inference requires the actual sensed PV system output power, the predicted PV system output power, and the sensed surrounding conditions. An artificial intelligent NARX-based neural network is used to obtain the predicted PV system output power. The actual output power of the PV system and the surrounding conditions are obtained in real-time using sensors. The algorithm is proven to be implementable on a low-cost microcontroller. The obtained results indicate that the fault diagnosis algorithm can detect multiple faults such as open and short circuit degradation, faulty maximum power point tracking (MPPT), and conditions of partial shading (PS) that may affect the PV system. Moreover, radiation and temperature, among other non-linear associations of patterns between predictors, can be captured by the proposed algorithm to determine the accurate point of the maximum power for the PV system.

Toward Better PV Panel’s Output Power Prediction; a Module Based on Nonlinear Autoregressive Neural Network with Exogenous Inputs

Much work has been carried out for modeling the output power of photovoltaic panels. Using artificial neural networks (ANNS), one could efficiently model the output power of heterogeneous photovoltaic (HPV) panels. However, due to the existing different types of artificial neural network implementations, it has become hard to choose the best approach to use for a specific application. This raises the need for studies that develop models using the different neural networks types and compare the efficiency of these different types for that specific application. In this work, two neural network types, namely, the nonlinear autoregressive network with exogenous inputs (NARX) and the deep feed-forward (DFF) neural network, have been developed and compared for modeling the maximum output power of HPV panels. Both neural networks have four exogenous inputs and two outputs. Matlab/Simulink is used in evaluating the proposed two models under a variety of atmospheric conditions. A comprehensive evaluation, including a Diebold-Mariano (DM) test, is applied to verify the ability of the proposed networks. Moreover, the work further investigates the two developed neural networks using their actual implementation on a low-cost microcontroller. Both neural networks have performed very well; however, the NARX model performance is much better compared with DFF. Using the NARX network, a prediction of PV output power could be obtained, with half the execution time required to obtain the same prediction with the DFF neural network, and with accuracy of ±0.18 W.

Day-Ahead Forecasting of Hourly Photovoltaic Power Based on Robust Multilayer Perception

Photovoltaic (PV) modules convert renewable and sustainable solar energy into electricity. However, the uncertainty of PV power production brings challenges for the grid operation. To facilitate the management and scheduling of PV power plants, forecasting is an essential technique. In this paper, a robust multilayer perception (MLP) neural network was developed for day-ahead forecasting of hourly PV power. A generic MLP is usually trained by minimizing the mean squared loss. The mean squared error is sensitive to a few particularly large errors that can lead to a poor estimator. To tackle the problem, the pseudo-Huber loss function, which combines the best properties of squared loss and absolute loss, was adopted in this paper. The effectiveness and efficiency of the proposed method was verified by benchmarking against a generic MLP network with real PV data. Numerical experiments illustrated that the proposed method performed better than the generic MLP network in terms of root mean squared error (RMSE) and mean absolute error (MAE).

A wavelet neural control scheme for a quadrotor unmanned aerial vehicle

Wavelets are designed to have compact support in both time and frequency, giving them the ability to represent a signal in the two-dimensional time-frequency plane. The Gaussian, the Mexican hat and the Morlet wavelets are crude wavelets that can be used only in continuous decomposition. The Morlet wavelet is complex-valued and suitable for feature extraction using the continuous wavelet transform. Continuous wavelets are favoured when high temporal and spectral resolution is required at all scales. In this paper, considering the properties from the Morlet wavelet and based on the structure of a recurrent high-order neural network model, a novel wavelet neural network structure, here called a recurrent Morlet wavelet neural network, is proposed in order to achieve a better identification of the behaviour of dynamic systems. The effectiveness of our proposal is explored through the design of a decentralized neural backstepping control scheme for a quadrotor unmanned aerial vehicle. The performance of the overall neural identification and control scheme is verified via simulation and real-time results.This article is part of the theme issue 'Redundancy rules: the continuous wavelet transform comes of age'.

Massively parallel WRNN reconstructors for spectrum recovery in astronomical photometrical surveys

The investigation of solar-like oscillations for probing star interiors has enjoyed a tremendous growth in the last decade. Once observations are over, the most notable difficulties in properly identifying the true oscillation frequencies of stars are due to the gaps in the observation time-series and the intrinsic stellar granulation noise. This paper presents an innovative neuro-wavelet reconstructor for the missing data of photometric signals. Firstly, gathered data are transformed using wavelet operators and filters, and this operation removes granulation noise, then we predict missing data by a composite of two neural networks, which together allow a "forward and backward" reconstruction. This resulting error is greatly lower than the absolute a priori measurement error. The devised reconstruction approach gives a signal that is better suited to be Fourier transformed when compared with other existing methods.

Cooperative Strategy for Optimal Management of Smart Grids by Wavelet RNNs and Cloud Computing

Advanced smart grids have several power sources that contribute with their own irregular dynamic to the power production, while load nodes have another dynamic. Several factors have to be considered when using the owned power sources for satisfying the demand, i.e., production rate, battery charge and status, variable cost of externally bought energy, and so on. The objective of this paper is to develop appropriate neural network architectures that automatically and continuously govern power production and dispatch, in order to maximize the overall benefit over a long time. Such a control will improve the fundamental work of a smart grid. For this, status data of several components have to be gathered, and then an estimate of future power production and demand is needed. Hence, the neural network-driven forecasts are apt in this paper for renewable nonprogrammable energy sources. Then, the produced energy as well as the stored one can be supplied to consumers inside a smart grid, by means of digital technology. Among the sought benefits, reduced costs and increasing reliability and transparency are paramount.

Short-term load and wind power forecasting using neural network-based prediction intervals

Electrical power systems are evolving from today's centralized bulk systems to more decentralized systems. Penetrations of renewable energies, such as wind and solar power, significantly increase the level of uncertainty in power systems. Accurate load forecasting becomes more complex, yet more important for management of power systems. Traditional methods for generating point forecasts of load demands cannot properly handle uncertainties in system operations. To quantify potential uncertainties associated with forecasts, this paper implements a neural network (NN)-based method for the construction of prediction intervals (PIs). A newly introduced method, called lower upper bound estimation (LUBE), is applied and extended to develop PIs using NN models. A new problem formulation is proposed, which translates the primary multiobjective problem into a constrained single-objective problem. Compared with the cost function, this new formulation is closer to the primary problem and has fewer parameters. Particle swarm optimization (PSO) integrated with the mutation operator is used to solve the problem. Electrical demands from Singapore and New South Wales (Australia), as well as wind power generation from Capital Wind Farm, are used to validate the PSO-based LUBE method. Comparative results show that the proposed method can construct higher quality PIs for load and wind power generation forecasts in a short time.