Multi-step Forecasting Research Articles

Multi-step forecasting of the Philippine storm frequencies using Poisson neural network

The paper aims to forecast the Philippine storm frequencies using nonlinear Poisson autoregressive model with exogenous variables. The nonlinearity is defined by its kernel which is an artificial neural network (ANN) with one hidden layer and two output neurons, and is trained to simultaneously forecast two semesters ahead for a given input. Furthermore, the covariates studied were the Average Sea Surface Temperatures in the NINO3.4 region (5∘− 5∘S,170∘− 120∘W) and the Average Sea Surface Temperatures in the eastern pole (0∘− 10∘S,90∘− 110∘E) of the Dipole Mode Index. The data, taken from the Japan Meteorological Agency’s Regional Specialized Meteorological Center with time points running from 1950 to October 2021, is modeled at a semester-level granularity. The estimation is done using maximum likelihood estimation by minimizing the negative log-likelihood function. Furthermore, Bayesian hyper-parameter optimization was used to tune the model across different activation functions, number of hidden neurons, training optimizers, and train and validation splits. Lastly, the best model from the hyper-parameter optimization was then compared to univariate Poisson autoregressive model (and its Bayesian counterpart) and Negative Binomial Autoregressive model. The proposed model captures well the characteristics of the data both in terms of the point forecast and the associated uncertainties.

RAAV Manufacturing: The Challenges of Soft Sensing during Upstream Processing

Recombinant adeno-associated virus (rAAV) is the most effective viral vector technology for directly translating the genomic revolution into medicinal therapies. However, the manufacturing of rAAV viral vectors remains challenging in the upstream processing with low rAAV yield in large-scale production and high cost, limiting the generalization of rAAV-based treatments. This situation can be improved by real-time monitoring of critical process parameters (CPP) that affect critical quality attributes (CQA). To achieve this aim, soft sensing combined with predictive modeling is an important strategy that can be used for optimizing the upstream process of rAAV production by monitoring critical process variables in real time. However, the development of soft sensors for rAAV production as a fast and low-cost monitoring approach is not an easy task. This review article describes four challenges and critically discusses the possible solutions that can enable the application of soft sensors for rAAV production monitoring. The challenges from a data scientist's perspective are (i) a predictor variable (soft-sensor inputs) set without AAV viral titer, (ii) multi-step forecasting, (iii) multiple process phases, and (iv) soft-sensor development composed of the mechanistic model.

Accurate one step and multistep forecasting of very short-term PV power using LSTM-TCN model

Accurate PV power forecasting is becoming a mandatory task to integrate the PV plant into the electrical grid, scheduling and guaranteeing the safety of the power grid. In this paper, a novel model to forecast the PV power using LSTM-TCN has been proposed. It consists of a combination between Long Short Term Memory and Temporal Convolutional Network models. LSTM is used to extract the temporal features from input data, then combined with TCN to build the connection between features and outputs. The proposed model has been tested using a dataset that includes historical time series of measured PV power. The accuracy of this model is then compared to LSTM and TCN models in different seasons, time periods forecast, cloudy, clear, and intermittent days. For one step forecasting, the results show that our proposed model outperforms the LSTM and TCN model. It has carried out a reduction of 8.47%, 14.26% for the autumn season, 6.91%,15.18 for the winter season, 10.22%,14.26% for spring season and 14.26%, 14.23% for the summer season on the Mean Absolute Error compared with LSTM, TCN. For multistep forecasting, LSTM-TCN surpassed all compared models in different time periods forecast from 2 steps to 7 steps PV power forecasting.

Variational Bayesian Network with Information Interpretability Filtering for Air Quality Forecasting

Air quality plays a vital role in people’s health, and air quality forecasting can assist in decision making for government planning and sustainable development. In contrast, it is challenging to multi-step forecast accurately due to its complex and nonlinear caused by both temporal and spatial dimensions. Deep models, with their ability to model strong nonlinearities, have become the primary methods for air quality forecasting. However, because of the lack of mechanism-based analysis, uninterpretability forecasting makes decisions risky, especially when the government makes decisions. This paper proposes an interpretable variational Bayesian deep learning model with information self-screening for PM2.5 forecasting. Firstly, based on factors related to PM2.5 concentration, e.g., temperature, humidity, wind speed, spatial distribution, etc., an interpretable multivariate data screening structure for PM2.5 forecasting was established to catch as much helpful information as possible. Secondly, the self-screening layer was implanted in the deep learning network to optimize the selection of input variables. Further, following implantation of the screening layer, a variational Bayesian gated recurrent unit (GRU) network was constructed to overcome the complex distribution of PM2.5 and achieve accurate multi-step forecasting. The high accuracy of the proposed method is verified by PM2.5 data in Beijing, China, which provides an effective way, with multiple factors for PM2.5 forecasting determined using deep learning technology.

Effective short-term forecasts of Saudi stock price trends using technical indicators and large-scale multivariate time series.

Forecasting the stock market trend and movement is a challenging task due to multiple factors, including the stock's natural volatility and nonlinearity. It concerns discovering the market's hidden patterns with respect to time to enable proactive decision-making and better futuristic insights. Recurrent neural network-based methods have been a prime candidate for solving complex and nonlinear sequences, including the task of modeling multivariate time series forecasts. Due to the lack of comprehensive and reference work in short-term forecasts for the Saudi stock price and trends, this article introduces a comprehensive and accurate forecasting methodology tailored to the Saudi stock market. Two steps were configured to render effective short-term forecasts. First, a custom-built feature engineering streamline was constructed to preprocess the raw stock data and enable financial-related technical indicators, followed by a stride-based sliding window to produce multivariate time series data ready for the modeling phase. Second, a well-architected Gated Recurrent Unit (GRU) model was constructed and carefully calibrated to yield accurate multi-step forecasts, which was trained using the recently published historical multivariate time-series data from the primary Saudi stock market index (TASI index), in addition to being benchmarked against a suitable baseline model, namely Vector Autoregression Moving-Average with Exogenous Regressors (VARMAX). The output predictions from the proposed GRU model and the VARMAX model were evaluated using a set of regression-based metrics to assess and interpret the model precision. The empirical results demonstrate that the proposed methodology yields outstanding short-term forecasts of the Saudi stock price trends price compared to existing efforts related to this work.

Order and disorder in the evolution of online knowledge community: an investigation of the chaotic behavior in social tagging systems with evidence of stack overflow

PurposeOnline question-and-answer (Q&A) communities serve as important channels for knowledge diffusion. The purpose of this study is to investigate the dynamic development process of online knowledge systems and explore the final or progressive state of system development. By measuring the nonlinear characteristics of knowledge systems from the perspective of complexity science, the authors aim to enrich the perspective and method of the research on the dynamics of knowledge systems, and to deeply understand the behavior rules of knowledge systems.Design/methodology/approachThe authors collected data from the programming-related Q&A site Stack Overflow for a ten-year period (2008–2017) and included 48,373 tags in the analyses. The number of tags is taken as the time series, the correlation dimension and the maximum Lyapunov index are used to examine the chaos of the system and the Volterra series multistep forecast method is used to predict the system state.FindingsThere are strange attractors in the system, the whole system is complex but bounded and its evolution is bound to approach a relatively stable range. Empirical analyses indicate that chaos exists in the process of knowledge sharing in this social labeling system, and the period of change over time is about one week.Originality/valueThis study contributes to revealing the evolutionary cycle of knowledge stock in online knowledge systems and further indicates how this dynamic evolution can help in the setting of platform mechanics and resource inputs.

Multistep Ahead Prediction of Vehicle Lateral Dynamics Based on Echo State Model

Lateral dynamics are critical for high-level autonomous driving, and multistep forecasting can significantly improve vehicle safety. However, the dynamics prediction method is rarely announced due to the difficulty of accurate prediction. In contrast to conventional time-series prediction, this article proposes a novel method for predicting a dynamical system with implicit parameters. Specifically, we construct an echo state model (ESM) incorporating a rational data organization and an efficient neural network structure. We observe the fading information property (FIP) of dynamic vehicle parameters and use temporal data to construct equivalent states for implicit parameters. Finally, a gated recurrent unit (GRU) neural network is designed and trained to accomplish the complex mapping of a vehicle dynamical system. Its performance is compared to that of several benchmark networks. The proposed method is capable of performing four-step-ahead prediction with high accuracy in both dry and slippery road conditions, implying the increased potential for some safe and comfortable driving applications of autonomous vehicles.

Multi-step forecast of PM2.5 and PM10 concentrations using convolutional neural network integrated with spatial–temporal attention and residual learning

Accurate and reliable forecasting of PM2.5 and PM10 concentrations is important to the public to reasonably avoid air pollution and for the governmental policy responses. However, the prediction of PM2.5 and PM10 concentrations has great uncertainty and instability because of the dynamics of atmospheric flows, making it difficult for a single model to efficiently extract the spatial–temporal dependences. This paper reports a robust forecasting system to achieve accurate multi-step ahead forecasting of PM2.5 and PM10 concentrations. First, correlation analysis is adopted to screen the spatial information on pollution and meteorology that may facilitate the prediction of concentrations in a target city. Then, a spatial–temporal attention mechanism is used to assign weights to original inputs from both space and time dimensions to enhance the essential information. Subsequently, the residual-based convolutional neural network with feature extraction capabilities is employed to model the refined inputs. Finally, five accuracy metrics and two additional statistical tests are applied to comprehensively assess the performance of the proposed forecasting system. In addition, experimental studies of three major cities in the Yangtze River Delta urban agglomeration region indicate that the forecasting system outperforms various prevalent baseline models in terms of accuracy and stability. Quantitatively, the proposed STA-ResCNN model reduces root mean square error by 5.595 %-15.247 % and 6.827 %-16.906 % for the average of 1–4 h ahead predictions in three major cities of PM2.5 and PM10, respectively, compared to baseline models. The applicability and generalization of the proposed forecasting system are further verified by the extended applications in the other 23 cities in the entire region. The results prove that the forecasting system is promising in the early warning, regional prevention, and control of air pollution.

A novel three-stage hybrid learning paradigm based on a multi-decomposition strategy, optimized relevance vector machine, and error correction for multi-step forecasting of precious metal prices

Accurate forecasting of precious metal prices will greatly assist investors, account managers, and mineral enterprises in making sound decisions and market assessments, while further improvement in the accuracy of precious metal price forecasting is challenging due to its non-linear oscillating characteristics. This paper proposes a novel three-stage hybrid learning paradigm for accurately predicting the prices of three precious metals, i.e., silver, palladium, and platinum. Firstly, the original price series will be decomposed by the complementary ensemble empirical mode decomposition (CEEMD) into several subsequences. Then the subsequences are reprocessed by secondary decomposition (SD) and sequence reconstruction using WPD and permutation entropy to reduce data noise and repetitive modeling. Secondly, all subsequences are fed into the relevance vector machine improved by the African vulture optimization algorithm for prediction to obtain the initial prediction results and the error series. Finally, the error series is further decomposed and predicted by WPD and an optimized relevance vector machine to correct the previously predicted precious metal prices and obtain the final prediction results. The actual precious metal price data are provided as model inputs for multi-step ahead forecasting, and the experimental results show that: the proposed novel learning paradigm achieves MAPE values of 0.2003%, 0.4552%, and 0.2151% for one-step-ahead forecasting of the prices of silver, palladium, and platinum, respectively; the proposed hybrid model performs the best among all the compared models, where hybridization of all the components is essential to improve the prediction accuracy and enables the proposed model to have a higher generalization capability; the results of this paper may provide more scientific insight and reference for precious metals investment and production.

The study of the quality of multi-step time series forecasting

The work is devoted to the study of the quality of multistep forecasting of time series using the electricity consumption data for forecasting. Five models of multistep forecasting have been implemented, with their subsequent training and evaluation of the results obtained. The dataset is an upgraded minute-by-minute measurement of four years of electricity consumption. The dataset has been divided into training, validation, and test samples for training and testing models. The implementation is simplified by using the TensorFlow machine learning library, which allows us to conveniently process and present data; build and train neural networks. The TensorFlow functionality also provides standard metrics used to assess the accuracy of time series forecasting, which made it possible to evaluate the obtained models for forecasting the time series of electricity consumption and highlight the best ofthose considered according to the given indicators. The models are built in such a way that they can be used in studies of the quality of time series forecasting in various areas of human life. The problem of multistep forecasting for twenty fourhours ahead, considered in the paper, has not yet been solved for estimating electricity consumption. Theobtainedforecasting accuracy is comparable to recently published methods for estimating electricity consumption used in other conditions.At the same time, the forecasting accuracy of the constructed models has been improved in comparison with other methods.

Forecasting Energy Consumption Using Hybrid CNN and LSTM Auto-Encoder Network with Hyperband Optimization

Abstract: Power consumption in the modern world has significantly increased due to the rapid growth of the human population and technological advancement. The precise forecasting of rising electricity usage is a requirement for planning strategies, boosting profits, cutting down on power waste, and ensuring the energy demand management system runs steadily. Modern developments in the area of electricity consumption prediction offer great techniques for capturing chaotic trends in energy consumption and even surpassing forecasting models that have been around for a long time. But to increase prediction accuracy, some significant drawbacks in the current consumption prediction models must be addressed. In this paper, we present a hybrid deep learning design-based Auto-Encoder (AE) network model comprising of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network for multi-step forecasting. The input dataset is de-noised within the framework using Fast Fourier Transform (FFT) and outliers are removed using Local Outlier Factor (LOF) algorithm, before learning in an unsupervised manner using the auto-encoder network. And, hyperband optimization is applied for tuning the hyperparameters. The outcomes of the numerical experiments demonstrate that the proposed architecture yields excellent prediction capabilities and great potential for generalization.

Long-Term Flooding Maps Forecasting System Using Series Machine Learning and Numerical Weather Prediction System

Accurate real-time forecasts of inundation depth and area during typhoon flooding is crucial to disaster emergency response. The development of an inundation forecasting model has been recognized as essential to manage disaster risk. In the past, most researchers used multiple single-point forecasts to obtain surface flooding depth forecasts with spatial interpolation. In this study, a forecasting model (QPF-RIF) integrating a hydrodynamic model (SOBEK), support vector machine–multi-step forecast (SVM-MSF), and a self-organizing map (SOM) were proposed. The task of this model was divided into four parts: hydrodynamic simulation, point forecasting, inundation database clustering, and spatial expansion. First, the SOBEK model was used in simulating inundation hydrodynamics to construct the flooding maps database. Second, the SVM-MSF yields water level (inundation volume) forecasted with a 1 to 72 h lead time. Third, the SOM clustered the previous flooding maps database into several groups representing different flooding characteristics. Finally, a spatial expansion module produced inundation maps based on forecasting information from forecasting flood volume and flood causative factors. To demonstrate the effectiveness of the proposed forecasting model, we presented an application to the Yilan River basin in Taiwan. Our forecasting results indicated that the proposed model yields accurate flood inundation maps (less than 1 cm error) for a 1 h lead time. For long-term forecasting (46 h to 72 h ahead), the model controlled the error of the forecast results within 7 cm. In the testing events, the model forecasted an average of 83% of the flooding area in the long term. This flood inundation forecasting model is expected to be useful in providing early flood warning information for disaster emergency response.

Hourly solar irradiance forecasting based on encoder–decoder model using series decomposition and dynamic error compensation

Accurate solar irradiance prediction is crucial for harnessing solar energy resources. However, the pattern of irradiance sequence is intricate due to its nonlinear and non-stationary characteristics. In this paper, a deep hybrid model based on encoder–decoder is proposed to cope with the complex pattern for hourly irradiance forecasting. The hybrid deep model integrates complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), encoder–decoder module, and dynamic error compensation (DEC) architecture. The CEEMDAN is implemented to reduce the nonlinear and non-stationarity of the irradiance sequence. The encoder–decoder integrates temporal convolutional networks (TCN), long short-term memory networks (LSTM), and multi-layer perceptron (MLP) for temporal features extraction and multi-step prediction. The DEC architecture dynamically updates the model based on adjacent error information to mine the predictable components of error information. Furthermore, a new loss function is further proposed for multi-objective optimization to balance the performance of multi-step forecasting. In the hourly irradiance forecasting experiments on the three public datasets, the root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R) of the proposed model are observed to be in a range of 30.693-34.433 W/m2, 19.398-22.900 W/m2, and 0.9872-0.9902, respectively. Compared to the benchmark models (including MLP, LSTM, and TCN), the RMSE and MAE reduce by 10.76%–22.00% and 5.47%–20.40%, respectively. The experimental results indicate that the proposed model shows accurate and robust forecasting performance and is a reliable alternative to hourly irradiance forecasting.

Impact of Data Loss on Multi-Step Forecast of Traffic Flow in Urban Roads Using K-Nearest Neighbors

Data-driven models have recently proved to be a very powerful tool to extract relevant information from different kinds of datasets. However, datasets are often subject to multiple anomalies, including the loss of important parts of entries. In the context of intelligent transportation, we examine in this paper the impact of data loss on the behavior of one of the frequently used approaches to address this kind of problems in the literature, namely, the k-nearest neighbors model. The method designed herein is set to perform multi-step traffic flow forecasts in urban roads. In our study, we deploy non-prepossessed real data recorded by seven inductive loop detectors and delivered by the Traffic Management Center (VMZ) of Bremen (Germany). Firstly, we measure the performance of the model on a complete dataset of 11 weeks. The same dataset is then used to artificially create 50 incomplete datasets with different gap sizes and completeness levels. Afterwards, in order to reconstruct these datasets, we propose three computationally-low techniques, which proved through empirical testing to be efficient in reproducing missing entries. Thereafter, the performance of the E-KNN model is assessed under the original dataset, incomplete and filled-in datasets. Although the accuracy of E-KNN under incomplete and reconstructed datasets depends on gap lengths and completeness levels, under original dataset, the model proves to deliver six-step forecasts with an accuracy of 83% on average over 3 weeks of the test set, which also translates to a less than one car per minute error.

Multistep short-term wind speed forecasting using transformer

Wind power can effectively alleviate the energy crisis. However, its integration into the grid affects power quality and power grid stability. Accurate wind speed prediction is a key factor in the efficient use of wind power. Because of its intermittent and nonstationary nature, wind speed forecasting is difficult, and is the topic of much research, especially long-time multistep forecasts. In this paper, the multistep wind speed prediction problem is regarded as a sequence-to-sequence mapping problem, and a multistep wind speed prediction model based on a transformer is proposed. This model is based on an encoder–decoder architecture, where the encoder generates representations of historical wind speed sequences of any length, the decoder generates arbitrarily long future wind speed sequences, and the encoder and decoder are associated by an attention mechanism. At the same time, the encoder and decoder of Transformer are completely based on a multi-head attention mechanism. For easy modeling, a 1-dimensional original wind speed sequence is transformed to a 16-dimensional sequence by ensemble empirical mode decomposition (EEMD), and the multidimensional wind speed data are directly modeled with Transformer. We trained the model with very large-scale (19 years of data) wind speed data averaged at 10-minute intervals, and performed the evaluation over one-year wind speed data. Results show that our one-step forecast model achieved an average mean absolute error (MAE) and root mean square error (RMSE) of 0.167 and 0.221, respectively. To the best of our knowledge, our 3-, 6-, 12-, and 24-hour multistep forecast model achieves a new state of the art in wind speed forecasting, with respective MAEs of 0.243, 0.290, 0.362, and 0.453, and RMSEs of 0.326, 0.401, 0.513, and 0.651. It is believed that performance can be further improved with better model parameter optimization.

Multi-step forecasting of multivariate time series using multi-attention collaborative network

Multi-step forecasting of multivariate time series plays a critical role in many fields, such as disaster warning and financial analysis. While attention-based recurrent neural networks (RNNs) achieved encouraging performance, two limitations exist in current models: i) Existing approaches merely focus on variables’ interactions, and ignore the negative noise of non-predictive variables, ii) These methods cannot model the difference in the temporal importance of the target and the non-predictive series to prediction. To tackle these challenges, we propose a triangle structured Multi-Attention Collaborative Network (MACN), which includes a backbone network T-net with attention-based encoder–decoder framework, and an auxiliary hierarchical network NP-net. NP-net focuses on non-predictive variables, capturing the most relevant variables and temporal dependencies through the proposed variables-distillation attention network (VDN) and long short-term memory network (LSTM). T-net executes on target variable, and its encoder and decoder are both connected to NP-net, thereby using the output of NP-net to assist learning and decision-making. Specifically we design a knowledge-enhanced LSTM (KeLSTM) as the encoder and decoder of T-net. In the coding stage, KeLSTM refines the output of NP-net to strengthen the latent semantics of the target variable. In the decoding stage, KeLSTM captures subtle differences between the target and the non-predictive variables’ contribution to prediction, and improves model’s predictive ability by alleviating such conflicts. Experiments on three real-world datasets demonstrate that MACN outperforms different types of state-of-the-art methods.

Univariate and Multivariate LSTM Models for One Step and Multistep PV Power Forecasting

The energy demand is increasing due to population growth and economic development. To satisfy this energy demand, the use of renewable energy is essential to face global warming and the depletion of fossil fuels. Photovoltaic energy is one of the renewable energy sources, widely used by several countries over the world. The integration of PV energy into the grid brings significant benefits to the economy and environment, however, high penetration of this energy also brings some challenges to the stability of the electrical grid, due to the intermittency of solar energy. To overcome this issue, the use of a forecasting system is one of the solutions to guarantee an effective integration of PV plants in the electrical grid. In this paper, a PV power ultra short term forecasting has been done by using univariate and multivariate LSTM models. Different combinations of input variables of the models and different timesteps forecasting were tested and compared. The main aim of this work is to study the influence of the different combinations of variables on the accuracy of the LSTM models for one-step forecasting and multistep forecasting and comparing the univariate and multivariate LSTM models with MLP and CNN models . The results show that for one step forecasting, the use of a univariate model based on historical data of PV output power is sufficient to get accurate forecasting with 28.98W in MAE compared to multivariate models that can reach 35.39W. Meanwhile, for multistep forecasting, it is mandatory to use a multivariate model that has historical data of meteorological variables and PV output power in the input of LSTM model. Moreover, The LSTM model shows great accuracy compared to MLP and CNN especially in multistep PV power forecasting.

M2STAN: Multi-modal multi-task spatiotemporal attention network for multi-location ultra-short-term wind power multi-step predictions

In recent years, wind power has continued to emerge as a key source of renewable energy. When large-scale wind farm clusters are connected to the grid for power generation, accurate multi-location ultra-short-term wind power predictions carry significant value in terms of ensuring the safety, stability, and economical operation of the power system. However, there are complex temporal and spatial correlations among multiple wind farms in multiple locations, which makes wind power predictions involving wind farm clusters very challenging. The development of artificial intelligence technology, especially graph machine learning, provides new approaches for modeling such spatiotemporal correlations. In addition, compared with single-step forecasting, multi-step forecasting can better reflect the general situation, and thus, it is more widely applicable in reality. To optimize multi-step wind power predictions in multiple locations, this report proposes a Multi-Modal Multi-Task Spatiotemporal Attention Network (M2STAN) model. The developed model employs a graph attention network and a bidirectional gated recurrent unit (Bi-GRU) to model the spatial and temporal dependence, respectively. In addition, the introduction of multi-modal and multi-task learning strategies improves the accuracy and computational efficiency of this predictive model. The results indicate that the proposed method is superior to existing methods, including support vector regression, Bi-GRU, multi-modal multi-task graph spatiotemporal networks, and graph convolutional deep learning architectures in terms of prediction performance.

Application of Big Data Analysis to Agricultural Production, Agricultural Product Marketing, and Influencing Factors in Intelligent Agriculture

Agricultural Internet of things (AIoT) promotes the modernization of traditional agricultural production and marketing model. However, the existing time series prediction methods for agricultural production and agricultural product (AP) marketing cannot adapt well to most real-world scenarios, failing to realize multistep forecast of production and AP marketing data. To solve the problem, this paper explores the big data analysis of agricultural production, AP marketing, and influencing factors in intelligent agriculture. To realize long-, and short-term predictions, a small-sample time series model was set up for AIoT production, and a big-sample time series model was constructed for AP marketing. The data fusion algorithm based on Kalman filter (KF) was adopted to fuse the massive multi-source AP marketing data. The proposed strategy was proved valid through experiments.

A Hybrid Generative Adversarial Network Model for Ultra Short-Term Wind Speed Prediction

To improve the accuracy of ultra-short-term wind speed prediction, a hybrid generative adversarial network model (HGANN) is proposed in this paper. Firstly, to reduce the noise of the wind sequence, the raw wind data are decomposed using complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). Then the decomposed modalities are entered into the HGANN network for prediction. HGANN is a continuous game between the generator and the discriminator, which in turn allows the generator to learn the distribution of the wind data and make predictions about it. Notably, we developed the optimized broad learning system (OBLS) as a generator for the HGANN network, which can improve the generalization ability and error convergence of HGANN. In addition, improved particle swarm optimization (IPSO) was used to optimize the hyperparameters of OBLS. To validate the performance of the HGANN model, experiments were conducted using wind sequences from different regions and at different times. The experimental results show that our model outperforms other cutting-edge benchmark models in single-step and multi-step forecasts. This demonstrates not only the accuracy and robustness of the proposed model but also the applicability of our model to more general environments for wind speed prediction.