Long-term Forecasting Model Research Articles

An Integrated Artificial Intelligence Approach for Building Energy Demand Forecasting

Buildings are complex assets, characterized by environments and uses that change over time, variable occupancies, and long life cycles. They have high operational costs, mostly due to their energy requirements, and account for 30% to 40% of global greenhouse gas emissions. Consequently, substantial effort has been made to forecast their energy needs, with the scope of optimizing their economic and environmental impact. In this regard, the available literature focuses mainly on short-term modeling through the implementation of sets of physics-based equations (i.e., white-box), functional relationships between input and output variables (i.e., black-box), or a combination of both (i.e., grey-box). On the other hand, more research is required on long-term forecast models with the aim of reducing the energy needs. Within this context, this article presents an original automatic procedure for forecasting the energy needs of buildings in short- and long-term time horizons. This is accomplished by scaling an unknown facility from a similar facility that is already known and by executing a black-box approach based on machine learning algorithms. The proposed method is implemented in real case studies in Italy, predicting the energy needs (i.e., heating, cooling, and electricity) of Sant’Anna Hospital in Ferrara using the historical data of Ca’ Foncello Hospital in Treviso. The results show an adjusted coefficient of determination above 0.7 and an average error below 10% for all the energy vectors, demonstrating a feasible forecast performance with a low training set-to-test set ratio.

VTNet: A multi-domain information fusion model for long-term multi-variate time series forecasting with application in irrigation water level

Time series forecasting is intricately tied to production and life, garnering widespread attention over an extended period. Enhancing the performance of long-term multivariate time series forecasting (MTSF) poses a highly challenging task, as it requires mining complicated and obscure temporal patterns in many aspects. For this reason, this paper proposes a long-term forecasting model based on multi-domain fusion (VTNet) to adaptively capture and refine multi-scale intra- and inter-variate dependencies. In contrast to previous techniques, we devise a dual-stream learning architecture. Firstly, the fast Fourier Transform (FFT) is adopted to extract frequency domain information. The original sequences are then transformed into 2D visual features in the temporal-frequency domain, and a 2D-TBlock is designed for multi-scale dynamic learning. Secondly, a combination of convolution and recurrent networks continues to explore the local temporal features and preserve the global trend. Finally, multi-modal circulant fusion is applied to achieve a more comprehensive and enriched feature fusion representation, further promoting overall performance. Extensive experiments are conducted on 9 public benchmark datasets and the real-world irrigation water level to showcase VTNet’s promoted performance and generalization. Moreover, VTNet yields 46.93% and 25.36% relative improvements for water level forecasting, revealing its potential application value in water-saving planning and extreme event early warning.

Optimized Hybrid Deep Learning for Real-Time Pandemic Data Forecasting: Long and Short-Term Perspectives

Background: With new variants of COVID-19 causing challenges, we need to focus on integrating multiple deep-learning frameworks to develop intelligent healthcare systems for early detection and diagnosis. Objective: This article suggests three hybrid deep learning models, namely CNN-LSTM, CNN-Bi- LSTM, and CNN-GRU, to address the pressing need for an intelligent healthcare system. These models are designed to capture spatial and temporal patterns in COVID-19 data, thereby improving the accuracy and timeliness of predictions. An output forecasting framework integrates these models, and an optimization algorithm automatically selects the hyperparameters for the 13 baselines and the three proposed hybrid models. Methods: Real-time time series data from the five most affected countries were used to test the effectiveness of the proposed models. Baseline models were compared, and optimization algorithms were employed to improve forecasting capabilities. Results: CNN-GRU and CNN-LSTM are the top short- and long-term forecasting models. CNNGRU had the best performance with the lowest SMAPE and MAPE values for long-term forecasting in India at 3.07% and 3.17%, respectively, and impressive results for short-term forecasting with SMAPE and MAPE values of 1.46% and 1.47%. Conclusion: Hybrid deep learning models, like CNN-GRU, can aid in early COVID-19 assessment and diagnosis. They detect patterns in data for effective governmental strategies and forecasting. This helps manage and mitigate the pandemic faster and more accurately.

A novel hybrid deep learning time series forecasting model based on long-short-term patterns

Time series forecasting constitutes a cornerstone in decision-making processes across diverse domains, ranging from finance and economics to healthcare and environmental science. However, traditional long-term forecasting models often overlook the nuanced characteristics of short-term patterns, which exhibit rapid fluctuations within brief intervals. These short-term patterns, encapsulating frequency domain features, encode valuable information about transient phenomena and periodic variations that can significantly impact predictive accuracy. In the paper has been proposed a novel approach to long-term forecasting of multi-dimensional time series data using convolutional and recurrent neural networks. The proposed model harnesses the power of advanced neural network architectures, integrating both short and long-term time series patterns trends while concurrently leveraging Fourier transform analysis to extract frequency domain patterns. By synthesizing information from short-term fluctuations and long-term trends, the proposed model offers a comprehensive framework for robust forecasting, enhancing predictive accuracy across diverse applications. Through extensive simulations, the results of experiments demonstrate the superior performance of proposed model compared to conventional forecasting techniques, thereby showcasing its potential to revolutionize time series forecasting methodologies.

Learning from Polar Representation: An Extreme-Adaptive Model for Long-Term Time Series Forecasting

In the hydrology field, time series forecasting is crucial for efficient water resource management, improving flood and drought control and increasing the safety and quality of life for the general population. However, predicting long-term streamflow is a complex task due to the presence of extreme events. It requires the capture of long-range dependencies and the modeling of rare but important extreme values. Existing approaches often struggle to tackle these dual challenges simultaneously. In this paper, we specifically delve into these issues and propose Distance-weighted Auto-regularized Neural network (DAN), a novel extreme-adaptive model for long-range forecasting of stremflow enhanced by polar representation learning. DAN utilizes a distance-weighted multi-loss mechanism and stackable blocks to dynamically refine indicator sequences from exogenous data, while also being able to handle uni-variate time-series by employing Gaussian Mixture probability modeling to improve robustness to severe events. We also introduce Kruskal-Wallis sampling and gate control vectors to handle imbalanced extreme data. On four real-life hydrologic streamflow datasets, we demonstrate that DAN significantly outperforms both state-of-the-art hydrologic time series prediction methods and general methods designed for long-term time series prediction.

A comparative study of Machine Learning and Deep Learning methods for flood forecasting in the Far-North region, Cameroon

Flood crises are the consequence of climate change and global warming, which lead to an increase in the frequency and intensity of heavy rainfall. Floods are, and remain, natural disasters that result in huge loss of lives and material damage. Flood risks threaten all countries of the globe in general. The Far-North region of Cameroon has suffered of flood crises on several occasions, resulting in significant loss of human lives, infrastructural and socio-economic damage, with the destruction of homes, crops and grazing areas, and the halting of economic activities. The models used for flood forecasting in this region are generally physical-based, and produce unsatisfactory results. The use of artificial intelligence based methods for flood forecasting in order to limit its consequences is a way to be explored in the Far-North region of Cameroon. The aims of the present research work is to design and compare the performance of Machine Learning and Deep Learning based models such as one dimensional Convolutional Neural Network, Long and Short Term Memory and Multi Layer Perceptron for short-term and long-term flood forecasting in the Far-North region of Cameroon. The models designed take as input the temperature and rainfall time series recorded in this region. Performance criteria used for evaluating models are Nash–Sutcliffe Efficiency, Percent Bias, Coefficient of Determination and Root Mean Squared Error. As the results of the design and performance comparison of the models, the best model for short-term flood forecasting is the LSTM model , and the best model for long-term flood forecasting is still the LSTM model. The best models obtained from the comparisons have satisfactory performance and good generalization capabilities, as reflected by the performance criteria. The results of our research work can be used for implementation of floods warning systems and for definition of an effective and efficient flood risk management policies in order to make the Far-North region of Cameroon more resilient to flood crises.

Enhancing urban mobility: integration of IoT road traffic data and artificial intelligence in smart city environment

<span>Efficient traffic management poses a significant challenge in smart cities, requiring the integration of diverse approaches. This paper presents an artificial intelligence framework that integrates internet of things (IoT) road traffic data to optimize traffic flow in smart city environments. Real-time traffic data is collected using IoT edge sensors, processed using machine learning (support vector machines, logistic regression, k-nearest neighbors) and deep learning long short-term memory (LTSM) algorithms, and utilized to develop accurate short-term and long-term traffic forecasting models. The proposed framework showcases superior performance compared to existing approaches, making it a widely applicable solution for smart city traffic management. By leveraging IoT road traffic data and artificial intelligence (AI) techniques, real-time monitoring, proactive decision-making, and dynamic traffic control can be achieved, leading to optimized traffic flow, reduced congestion, and enhanced urban mobility. This research provides valuable insights into the potential of IoT and AI technologies in addressing urban traffic challenges and lays the foundation for intelligent transportation systems in smart city environments.</span>

출생 성비와 명절 시기를 고려한 월별 한우 도축두수 예측모형 개발

The number of Korean native cattle, Hanwoo, has exceeded 3.5 million, the historical largest one, in May 2022. As a result, the Hanwoo industry are in a state of oversupply and are concerned about the risk of a price collapse. The main objective of this study is to develop medium- and long-term forecasting models for monthly Hanwoo supply as a way to reduce the risk of imbalance between supply and demand. To construct the forecasting models, the empirical histogram of the month-specific rate for Hanwoo is applied. The monthly data of Hanwoo in 2014-2022 are obtained from the Animal Product Traceability System in Korea. The results are as follows: A forecasting model for the number of monthly Hanwoo at 6-month-old with considering the birth sex ratio outperforms the other models. Also, adjusting the slaughter rate based on holiday season made the forecast more accurate. The forecasting models were evaluated based on multiple accuracy measures such as mean absolute percentage error(MAPE) and root mean square percentage error (RMSPE). The MAPE for the monthly slaughtered Hanwoo females and males were 9.37% and 8.98%, respectively. Therefore, the forecasting model developed in this study is expected to be useful as a tool to reduce the risk faced by those in the Hanwoo industry.

A Short-Term Forecast Method of Maximum Usable Frequency for HF Communication

To further improve the quantitative forecasting ability of high-frequency (HF) communication frequency, we propose a maximum usable frequency (MUF) adaptive forecasting method. This method has four new features: 1) The Volterra filter is introduced to this method, which is easy to implement due to the simple structure. 2) This method can forecast M(3000)F2 based on past measurements without using any previous or forecast data of geomagnetic and solar activity parameters. 3) The spatial characteristics of the M(3000)F2 can be obtained by the improved reconstruction method based on geomagnetic coordinates. 4)The high prediction accuracy can be obtained by the training data set on the 27-day (one solar rotation period). Through the statistical analysis of the measurements and predictions, the results show that the forecast values are in good agreement consistent with the measurement during the geomagnetic quiet or the disturbance periods. This method is proven to have good adaptability and usability and has a further performance improvement compared with ITU and long-term forecast models. And the root-mean-square error of MUF forecasts decreased by 1.9MHz and 0.9MHz on average, and the relative root-mean-square error decreased by 12.6% and 2.9% on average.

Evaluation of a novel hybrid lion swarm optimization – AdaBoostRegressor model for forecasting monthly precipitation

Precipitation is essential for crop production, water resource management, and other activities. Since precipitation has unpredictable sequential and seasonal characteristics, it is extremely difficult to forecast precipitation. Hence, a better forecasting model will lead to an early warning that minimizes the risk to life and property. Many approaches, such as statistical, Machine Learning (ML), Deep Learning (DL), and hybrid algorithms, are used to build precipitation forecasting systems. Feature selection is especially important for selecting past values (lags) and to eliminate irrelevant and/or redundant features. The optimization algorithm helps identify correct time series lags because it does not make data assumptions. In this work, a hybrid long-term (1-month ahead) precipitation forecast model (ABR-LSO) is proposed by integrating the AdaBoostRegressor (ABR) with the Lion Swarm Optimization (LSO) algorithm. Time series precipitation data for the period 1989–2019 for Idukki, Kerala, India. The ABR model considered 12 antecedent rainfall records as its input features and predicted the rainfall data for the next time step (t + 1), whereas the ABR-LSO hybrid model optimized the input features with LSO. The efficacy of the forecast models is evaluated using statistical indices like mean absolute error (MAE) and root mean square error (RMSE). It has been noted that in models 1, 2, and 3, the LSO-ABR model produced good MAE above ABR model improvements of 5.08 %, 4 %, and 6.12 %, respectively. Precipitation forecasts will help the administrators devise disaster prevention mechanisms against flooding in the downstream areas.

FEDformer-Based Paddy Quality Assessment Model Affected by Toxin Change in Different Storage Environments.

The storage environment can significantly impact paddy quality, which is vital to human health. Changes in storage can cause growth of fungi that affects grain quality. This study analyzed grain storage monitoring data from over 20 regions and found that five factors are essential in predicting quality changes during storage. The study combined these factors with the FEDformer (Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting) model and k-medoids algorithm to construct a paddy quality change prediction model and a grading evaluation model, which showed the highest accuracy and lowest error in predicting quality changes during paddy storage. The results emphasize the need for monitoring and controlling the storage environment to preserve grain quality and ensure food safety.

Ensemble Kalman filter for GAN-ConvLSTM based long lead-time forecasting

Data-driven machine learning techniques have been increasingly utilized for accelerating nonlinear dynamic system prediction. However, machine learning-based models for long lead-time forecasts remain a significant challenge due to the accumulation of uncertainty along the time dimension in online deployment. To tackle this issue, the ensemble Kalman filter (EnKF) has been introduced to machine learning-based long-term forecast models to reduce the uncertainty of long lead-time forecasts of chaotic dynamic systems. Both the deep convolutional generative adversarial network (DCGAN) and convolutional long short term memory (ConvLSTM) are used for learning the complex nonlinear relationships between the past and future states of dynamic systems. Using an iterative Multi-Input Multi-Output (MIMO) algorithm, the two-hybrid forecast models (DCGAN-EnKF and ConvLSTM-EnKF) are able to yield long lead-time forecasts of dynamic states. The performance of the hybrid models has been demonstrated by one-level and two-level Lorenz 96 models. Our results show that the use of EnKF in ConvLSTM and DCGAN models successfully corrects online model errors and significantly improves the real-time forecasting of dynamic systems for a long lead-time.

Information granules-based long-term forecasting of time series via BPNN under three-way decision framework

As a significant issue in the machine learning field, the long-term forecasting of time series has aroused extensive attention from academia and industry. Specifically, transforming time series into granularity time series (GTS) for forecasting is usually perceived as an efficient way to address the long-term forecasting of time series. Although some long-term forecasting models have been explored to granulate time series information, the extraction of time series trend information is always accompanied by information losses. In order to address this challenge, a Back Propagation neural network (BPNN) time series long-term forecasting model is established via information granules under the framework of three-way decisions (TWD), which is conducive to greatly preserving the trend information of time series. First, significant points in the time series are extracted to initially extract the trend information of time series. Second, a reasonable loss function and conditional probability of TWD are defined, and a TWD-based method is proposed to compress the information and extract the trend information. Third, information granules are used as input for the prediction via BPNN. Notably, the proposed model not only retains the trend information, but also reduces errors due to the segmentation of information particles and extends the application range of forecasting problems with TWD. Finally, experimental outcomes with several data sets reveal that the proposed model owns better performance than existing long-term forecasting models.

Long-term electricity forecasting for the industrial sector in western China under the carbon peaking and carbon neutral targets

To reflect the future electricity demand variations in the industrial sector of western China under the “carbon peaking” and “carbon neutral” strategies, that the traditional methods of electricity consumption forecasting are no longer effective, this paper proposes a new hierarchy of electricity demand influencing factors based on the transition path of low-carbon and CO2 emission intensity constraints of the industrial sector. Then, the feedback equations of the influencing factors based on CO2 emission intensity is constructed, and a long-term electricity forecast model based on system dynamics is established. Finally, taking Ningxia Province as an example, the electricity demand and carbon emissions of the industrial sector are predicted under different policy constraint scenarios. The results show: (1) In the baseline scenario, electricity demand will tend to saturate with about 140.1TWh in 2030, and the CO2 emission level of electricity demand will peak at 72.52 million tons in 2029. (2) Compared with the system dynamics model without considering CO2 emission intensity, the mean absolute percentage error is reduced about 3.16 % by inputting the influence factors and regression coefficients designed in this paper. (3) Compared with other forecasting methods, the model proposed in this paper has the lowest mean absolute percentage error of 2.75 %. The model improves the accuracy and provides a reference for planning departments to develop energy transition plans, where the prediction results reflect the reality of the industrial sector in western China. Correspondingly, this model has reference value for relevant institutions in forecasting the electricity demand under the low-carbon development path in industrial sectors of different regions.

SARIMA-Based Medium- and Long-Term Load Forecasting

The operation and planning of power systems depend heavily on M-LTLF, which is complicated and nonlinear, making it challenging for conventional medium- and long-term forecasting models to produce reliable results. The SARIMA model is chosen for M-LTLF in this study, and the model’s parameters are tuned. This study takes the electricity consumption data of the whole Yunnan as the research object. Among them, the electricity consumption data from 2008 to 2018 is used as a training sample for fitting and analysis, and the electricity consumption of the whole province is predicted from 2019 to 2020. The end results demonstrate the viability and efficacy of the SARIMA model for M-LTLF.

An attention-based LSTM model for long-term runoff forecasting and factor recognition

With advances in artificial intelligence, machine learning-based models such as long short-term memory (LSTM) models have shown much promise in forecasting long-term runoff by mapping pathways between large-scale climate patterns and catchment runoff responses without considering physical processes. The recognition of key factors plays a vital role and thus affects the performance of the model. However, there is no conclusion on which recognition algorithm is the most suitable. To address this issue, an LSTM model combined with two attention mechanisms both in the input and hidden layers, namely AT-LSTM, is proposed for long-term runoff forecasting at Yichang and Pingshan stations in China. The added attention mechanisms automatically assign weights to 130 climate phenomenon indexes, avoiding the use of subjectively set recognition algorithms. Results show that the AT-LSTM model outperforms the Pearson’s correlation based LSTM model in terms of four evaluation metrics for monthly runoff forecasting. Further, the set indirect runoff prediction method verifies that the AT-LSTM model also performs effectively in precipitation and potential evapotranspiration forecasting, and the indirect runoff prediction is inferior to the AT-LSTM model to establish a direct link between climate factors and runoff. Finally, four key factors related to runoff are identified by the attention mechanism and their impacts on runoff are analyzed on intra- and inter-annual scales. The proposed AT-LSTM model can effectively improve the accuracy of long-term forecasting and identify the dynamic influence of input factors.

Study of Long-Term Energy Storage System Capacity Configuration Based on Improved Grey Forecasting Model

Distributed generation equipment improves renewable energy utilization and economic benefits through an energy storage system (ESS). However, dominated by short-term data, the configuration of long-period ESS capacity is absent based on the dynamic change of load, which leads to a large deviation from the expected return. Considering the system characteristics of lack of data and less information, after introducing the grey theory, we propose a new long-term capacity configuration method for ESS and establish the long-term grey forecasting model (GFM) of user load, improving the basic forecasting model to improve the accuracy of the long-term forecasting model. Then, the scheduling model is established with the maximum economic and social benefits as the optimization objective. Based on the forecast data of the improved grey forecasting model (IGFM), the hierarchical solution method is used to solve the scheduling model. Finally, the parameters are configured based on the service life of the equipment and the expected rate of return. The simulation results show that higher accuracy is realized in the improved prediction model, and the improved algorithm gets higher convergence speed and precision. Apart from that, the nonlinear correlation trend of the EES return rate between the capacity and life cycle is revealed. Compared with the ESS configuration in a short period, this study provides more comprehensive and accurate data support for the capacity configuration of the ESS, reducing the error between the actual return and the expected return significantly.

Intelligent Traffic Management using IoT and Machine Learning

The continuous improvements in the Internet of Things (IoTs) and machine learning (ML) make them the key enabling technologies for intelligent traffic management (ITM).The ability to accurately predict network traffic has been demonstrated as crucial for effective network management and strategic planning. Proactive management of future congestion incidents requires access to reliable long-term forecasting models. Conventional prediction methods often fail to completely capture the spatiotemporal features of the traffic flows because of the complexity of the interdependence between the flows. To this end, we proposed to improve the management of traffic with a novel framework for the predictive modeling of traffic flows. The proposed formwork introduces an improved graph network to capture the positional information in traffic follows. It is also capable of precisely capturing temporal dynamics using an improved bidirectional learning module. An attention mechanism is presented to capture the interactions among spatial and temporal patterns to further empower the predictive power of the model. Proof-of-concept experimentations are conducted on the PeMSD7 dataset, and the results (MAE: 0.197, MSE: 0.13, RMSE: 0.36, ) demonstrate the efficiency of our model over the state-of-the-art.

A long-term forecasting and simulation model for strategic planning of hospital bed capacity

Growing healthcare needs leverage the potential savings of using resources efficiently. To that end, ProMoBed is a comprehensive model that supports strategic planning of bed capacity in inpatient hospitals. The model consists of an extrapolation and simulation component, the former supplying input for the latter. The extrapolation model forecasts admission rates and the average Length of Stay for pathology groups, and corrects for demographic changes. Subsequently, the simulation model emulates the demand for bed capacity, and makes service-level based bed capacity suggestions. Additionally, the model uses the Shapley value principle to disaggregate the effects on demand for inpatient days due to different causes. Results from the extrapolation model are applied to regions in Belgium, showing expected divergence in inpatient day demand evolution.

Short-term Adaptive Forecast Model for TEC over equatorial low latitude region

In today's world, satellite navigation is one of the most widely used navigational aids, which is why accurate positioning is of critical importance. The ionosphere's Total Electron Content (TEC) is a significant contributor to the positional error in signal traveling between a satellite and the earth and reduces the system's positional accuracy. The ionosphere adds an excess delay to the radio wave traversing through it, resulting in position estimation error. Since the delay is a function of the ionospheric parameter TEC, estimation of TEC can help in correcting the errors. The challenge in estimating TEC in the equatorial region is that it exhibits a non-deterministic or unpredictable characteristic due to the complex electrodynamics resulting from Equatorial Ionospheric Anomaly (EIA). In addition, space weather phenomena like geomagnetic storms further disturb the ionosphere causing irregular short-term disturbances. This study explores the inherent VTEC pattern during geomagnetic storm conditions, especially during the high solar activity period, and proposes a short-term adaptive segmented regression model to forecast VTEC. The study is undertaken over six years of high solar activity period in the 24th solar cycle. The forecast model uses both linear and polynomial autoregression coefficients of recent past data. The model's performance is evaluated for extreme conditions, including the high solar activity and geomagnetic storms, which are the most challenging scenario for TEC prediction. The accuracy of the predicted output is calculated in terms of RMSE and correlation coefficient. The results were also compared with IRI 2016 model. The proposed short-term prediction model thus aims to help fill the precision required by long-term forecast models by considering near real-time nonlinear dynamics in the ionosphere.