Forecast Step Research Articles

Implementing augmented deep Machine learning for effective shallow water table management and forecasting

This study addresses the gap in understanding and forecasting shallow water table depth (WTD), a critical factor in groundwater resource management and agricultural productivity. Despite the importance of accurately forecasting WTD for sustainable water resource management, current methods frequently struggle to capture the complexities and dynamics of WTD fluctuations. In response, this research, which was conducted in Québec, Canada, leverages machine learning techniques—namely, extreme learning machines (ELMs) and long short-term memory (LSTM) networks, augmented by the Holt-Winters (HW) state-space method—to develop a comprehensive analysis and forecasting approach for shallow WTD. The datasets were recorded by 8 sensors with hourly temporal resolutions from June to September, covering the growing season. The objective was to increase forecast accuracy by employing a detailed structural analysis of WTD time series data, selecting appropriate forecast steps, and fine-tuning model inputs through statistical tests and model-agnostic interpretation methods. The performance was evaluated via various metrics, including the correlation coefficient (R), root mean square error (RMSE), mean absolute relative error (MARE), and Theil’s U accuracy and quality coefficients, across short- to long-term forecasts (1-, 12-, 24-, 48-, and 72-hour ahead). Integration of HW with the ELM and LSTM models markedly improved the forecasting capabilities, particularly for the LSTM model, which achieved high accuracy of R = 0.988 for 1-hour forecasts and low error rates (RMSE = 0.648 cm, MARE = 0.007, UI = 0.005, and UII = 0.010), although accuracy decreased for longer forecast horizons, resulting in the lowest accuracy for 72-hour forecasts, with R = 0.638, RMSE = 4.550 cm, MARE = 0.051, UI = 0.036, and UII = 0.071. Similarly, the ELM model showed promising results in short-term forecasts when coupled with HW (R = 0.988, RMSE = 0.676 cm, MARE = 0.007, UI = 0.005, and UII = 0.010) but experienced a decrease in performance accuracy over more extended forecast periods (R = 0.707, RMSE = 5.559 cm, MARE = 0.053, UI = 0.045, and UII = 0.089). Although the ELM model presented a negligible strong correlation in some forecast steps, the LSTM model offered consistently higher forecast accuracy and quality across all assessed horizons. The study demonstrates the superiority of the LSTM model in consistently providing more accurate forecasts, highlighting the importance of integrating HW to capture complex temporal patterns in hydrological forecasting. This advancement in forecasting WTD has substantial implications for enhancing groundwater resource management and agricultural decision-making, significantly contributing to sustainable water resource utilization and supporting agricultural productivity through informed data-driven practices.

Hybrid covariance super-resolution data assimilation

The super-resolution data assimilation (SRDA) enhances a low-resolution (LR) model with a Neural Network (NN) that has learned the differences between high and low-resolution models offline and performs data assimilation in high-resolution (HR). The method enhances the accuracy of the EnKF-LR system for a minor computational overhead. However, performance quickly saturates when the ensemble size is increased due to the error introduced by the NN. We therefore combine the SRDA with the mixed-resolution data assimilation method (MRDA) into a method called “Hybrid covariance super-resolution data assimilation” (Hybrid SRDA). The forecast step runs an ensemble at two resolutions (high and low). The assimilation is done in the HR space by performing super-resolution on the LR members with the NN. The assimilation uses the hybrid covariance that combines the emulated and dynamical HR members. The scheme is extensively tested with a quasi-geostrophic model in twin experiments, with the LR grid being twice coarser than the HR. The Hybrid SRDA outperforms the SRDA, the MRDA, and the EnKF-HR at a given computational cost. The benefit is the largest compared to the EnKF-HR for small ensembles. However, even with larger computational resources, using a mix of high and low-resolution members is worth it. Besides, the Hybrid SRDA, the EnKF-HR, and the SRDA, unlike the MRDA, prevent the smoothing of dynamical structures of the background error covariance matrix. The Hybrid SRDA method is also attractive because it is customizable to available resources.

Wind tunnel wall interference correction for transonic airfoils with data-reduced ensemble Kalman filter

The uncertain factors in wind tunnel experiments, especially the interference effects, will cause the flow around the test model to differ from that under the free flow condition. The wind tunnel wall interference effects are challenging to be eliminated, especially in transonic flow regions where classical linear correction methods are ineffective. To address this issue, an efficient wind tunnel wall interference correction framework based on data assimilation is proposed for steady and shock buffet flows around the airfoils in the transonic region in this work. The ensemble Kalman filter (EnKF) is used to find the combination of inflow conditions, which minimizes the differences between the pressure coefficients measured in the wind tunnel experiment and those estimated by computational fluid dynamics (CFD). The polynomial chaos expansion method is used to propagate uncertainty in the forecast step of the EnKF in order to improve efficiency. Typical wind tunnel wall interference correction problems are studied, including two steady cases: the transonic flow around the Royal Aircraft Establishment 2822 (RAE2822) and the National Aerospace Laboratory 7301 (NLR7301) airfoils, and one unsteady case: the transonic shock buffet on the National Aeronautics and Space Administration supercritical (phase 2)-0714 (NASA SC(2)-0714) airfoil. It is demonstrated that with the corrected Mach number and angle of attack, the estimated CFD results are in better agreement with the measured experimental data than traditional linear methods and the computational efficiency is improved compared to the original EnKF.

Accurate deep learning-based filtering for chaotic dynamics by identifying instabilities without an ensemble.

We investigate the ability to discover data assimilation (DA) schemes meant for chaotic dynamics with deep learning. The focus is on learning the analysis step of sequential DA, from state trajectories and their observations, using a simple residual convolutional neural network, while assuming the dynamics to be known. Experiments are performed with the Lorenz 96 dynamics, which display spatiotemporal chaos and for which solid benchmarks for DA performance exist. The accuracy of the states obtained from the learned analysis approaches that of the best possibly tuned ensemble Kalman filter and is far better than that of variational DA alternatives. Critically, this can be achieved while propagating even just a single state in the forecast step. We investigate the reason for achieving ensemble filtering accuracy without an ensemble. We diagnose that the analysis scheme actually identifies key dynamical perturbations, mildly aligned with the unstable subspace, from the forecast state alone, without any ensemble-based covariances representation. This reveals that the analysis scheme has learned some multiplicative ergodic theorem associated to the DA process seen as a non-autonomous random dynamical system.

A decomposition-ensemble-integration framework for carbon price forecasting

Accurate carbon price forecasting is crucial for effective carbon trading policies that mitigate climate change. The volatility and uncertainty of carbon prices pose significant challenges. Traditional frameworks that use a single model for all intrinsic mode functions (IMFs) fail to leverage the diverse strengths of different models. In this paper, we propose DecEnsInt, a novel framework that utilizes an ensemble of models tailored to distinct frequency domains. By using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to decompose historical prices into IMFs and optimizing ensemble parameters through Sequential Least Squares Programming (SLSQP) for each frequency domain, DecEnsInt enhances forecasting performance. Extensive experiments on four Emissions Trading System (ETS) datasets, evaluated using RMSE, MAPE, MAE and R2, demonstrate DecEnsInt’s superiority, achieving relative improvements in key metrics by 10.58% to 17.72% over runner-up ensemble methods. DecEnsInt also achieves better robustness and stability across different forecasting steps, proving its effectiveness in handling the complexities of carbon price forecasting.

Multi-step carbon emissions forecasting model for industrial process based on a new strategy and machine learning methods

The rising industrial sector capacity increases carbon emissions, necessitating a low-carbon transformation for global sustainability. Accurate multi-step forecasting of industrial carbon emissions is crucial for optimizing production, reducing energy consumption, improving efficiency, and achieving decarbonization. However, current forecasting faces challenges like model uncertainty and low accuracy with multi forecasting steps. To address these issues, this study proposed a multi-step carbon emissions forecasting model for industrial processes. Firstly, a new multi-step forecasting strategy framework was proposed. Secondly, SHapley Additive exPlanations (SHAP) and lagged autocorrelation analysis were used to extract key features affecting industrial carbon emissions. On this basis, the grey model, autoregressive moving average (ARMA) model, and least squares support vector machine (LSSVM) method were used to build forecasting models separately. The weights of the three models were solved using the induced ordered weighted average operator and the Monte Carlo search tree (MCST) method. The final forecasting results were obtained by using the weighted summation method. The data collected from the real industry were used to validate the proposed model. The results showed that the proposed model could accurately forecast industrial carbon emissions within 12 steps and its MAPE is less than 10%. The proposed multi-step industrial carbon emissions forecasting model can provide data support for industrial energy saving, carbon reduction, and green and sustainable development.

GGNet: A novel graph structure for power forecasting in renewable power plants considering temporal lead-lag correlations

Power forecast for each renewable power plant (RPP) in the renewable energy clusters is essential. Though existing graph neural networks (GNN)-based models achieve satisfactory prediction performance by capturing dependencies among distinct RPPs, the static graph structure employed in these models ignores crucial lead-lag correlations among RPPs, resulting from the time difference of the air flow at spatially dispersed RPPs. To address this problem, this paper proposes a novel dynamic graph structure using multiple temporal granularity groups (TGGs) to characterize the lead-lag correlations among RPPs. A granular-based GNN called GGNet is designed to generate an optimal adjacency matrix for the proposed graph structure. Specifically, a two-dimensional convolutional neural network (2D-CNN) is used to quantify the uncertain lead-lag correlations among RPPs; secondly, a gate mechanism is used to calculate a dynamic adjacency matrix; Finally, a graph attention network (GAT) is used to aggregate the information on RPPs based on the well-learned adjacency matrix. Case studies conducted using real-world datasets, with wind power plants and photovoltaic power plants, show our method outperforms state-of-the-art (SoTA) ones with better performance. Compared with the SoTA models, the RMSEN and MAEN of wind power plants for 1–4 h forecast steps decreased on average by 22.925% and 13.18%, respectively; the RMSEN and MAEN of photovoltaic power plants for 1–4 h forecast steps decreased on average by 48.95% and 18.75%, respectively. The results show that the proposed framework can generate improved performance with accuracy and robustness.

The Contribution of Aeolus Wind Observations to ECMWF Sea Surface Wind Forecasts

AbstractAeolus is the first satellite mission focusing on wind profile detection from near the surface to about 30 km in height on a global scale. This study evaluates the contribution of Aeolus winds to sea surface wind forecasts geographically by further analyzing the Observing System Experiments from the European Centre for Medium‐Range Weather Forecasts (ECMWF) with scatterometer winds from the meteorological operational satellites (assimilated into the model) and the Haiyang‐2B satellite (not assimilated into the model). The findings indicate that Aeolus has the ability to reduce the root‐mean‐square difference between scatterometer winds and background forecasts (short‐range) by about 0.05%–0.16% on average for climatic regions, except for the meridional wind component in the tropics. Also, Aeolus can generally reduce zonal biases of the background forecasts, while its beneficial impact on meridional biases mainly occurs in the Northern Hemisphere extratropics and tropics. For medium‐range forecast assessments, as the forecast step extends up to day 5, the positive impact of Aeolus on sea surface wind forecasts becomes more evident and is even greater than 3%, especially for extratropical ocean regions in the Southern Hemisphere. Furthermore, the impact of Aeolus shows seasonal variation, with a substantial positive impact from September 2019 to February 2020 and a negative impact mainly in March, April, and May 2020.

A multi-step wind power group forecasting seq2seq architecture with spatial–temporal feature fusion and numerical weather prediction correction

Accurate total wind power forecasting is essential to wind farms. Conventional methods predict each wind turbine separately, which fails to exploit the spatial correlation between turbines and causes error accumulation through summation. Additionally, numerical weather prediction (NWP) with limited spatial–temporal resolutions is difficult to assist in power forecasting effectively, resulting in decreased performance as the forecasting steps increase. To address these drawbacks, a novel “N encoder-1 decoder” multi-channel fusion group forecasting sequence-to-sequence (seq2seq) architecture with wind resource quality assisted spatial attention (WRQASA) and NWP correction (NWPC) is proposed in this paper. The encoders in the architecture extract the temporal features of each turbine, followed by WRQASA adaptively adjusting their impact on the decoder. NWPC constructs micro-scale meteorological data based on its large-scale counterpart. The decoder takes the output from WRQASA and NWPC, and predicts the total power of the turbine group. The model is evaluated on a real-world wind farm. Ablation studies revealed the effectiveness of the novel architecture, WRQASA and NWPC. In 12, 24, 36 and 72-h multi-step forecasting scenarios, the new method outperformed conventional methods, such as seq2seq, light gradient boosting machine and support vector regression, achieving a more effective prediction of total power.

Wind Power Forecasting in the presence of data scarcity: A very short-term conditional probabilistic modeling framework

The uncertainty of wind power (WP) poses a significant challenge to power systems with a high percentage of WP. Accurate WP forecasting is an important approach to mitigate this issue. With the increasing demand for electricity, some wind farms (WFs) have been built or expanded. Such WFs lack abundant historical data for establishing forecasting models, i.e., data scarcity. This paper proposes a novel method for very short-term probabilistic forecasting based on transfer learning (TL) and conditional screening for WP. Utilizing numerical weather prediction as predictive features to address deterministic forecast data scarcity. Considering the data orientation, combining bidirectional long short-term memory with attention mechanisms to enhance prediction accuracy. Based on TL and conditional screening, it achieves data completion and singular value filtering to enhance the comprehensive performance of probabilistic forecasting errors. Using real world data, the proposed WP forecasting model reduces MAPE by at least 0.87% and RMSE by at least 0.6483 compared to the benchmark at different forecast steps. The comprehensive performance of probabilistic forecasting improves by at least 1.9134 at different forecast steps and confidence levels. The results indicate that the proposed forecasting model for data scarcity WFs is feasible, providing new insights for data scarcity WF forecasting.

Using synthetic case studies to explore the spread and calibration of ensemble atmospheric dispersion forecasts

Abstract. Ensemble predictions of atmospheric dispersion that account for the meteorological uncertainties in a weather forecast are constructed by propagating the individual members of an ensemble numerical weather prediction forecast through an atmospheric dispersion model. Two event scenarios involving hypothetical atmospheric releases are considered: a near-surface radiological release from a nuclear power plant accident and a large eruption of an Icelandic volcano releasing volcanic ash into the upper air. Simulations were run twice-daily in real time over a 4-month period to create a large dataset of cases for this study. The performance of the ensemble predictions is measured against retrospective simulations using a sequence of meteorological fields analysed against observations. The focus of this paper is on comparing the spread of the ensemble members against forecast errors and on the calibration of probabilistic forecasts derived from the ensemble distribution. Results show good overall performance by the dispersion ensembles in both studies but with simulations for the upper-air ash release generally performing better than those for the near-surface release of radiological material. The near-surface results demonstrate a sensitivity to the release location, with good performance in areas dominated by the synoptic-scale meteorology and generally poorer performance at some other sites where, we speculate, the global-scale meteorological ensemble used in this study has difficulty in adequately capturing the uncertainty from local- and regional-scale influences on the boundary layer. The ensemble tends to be under-spread, or over-confident, for the radiological case in general, especially at earlier forecast steps. The limited ensemble size of 18 members may also affect its ability to fully resolve peak values or adequately sample outlier regions. Probability forecasts of threshold exceedances show a reasonable degree of calibration, though the over-confident nature of the ensemble means that it tends to be too keen on using the extreme forecast probabilities. Ensemble forecasts for the volcanic ash study demonstrate an appropriate degree of spread and are generally well-calibrated, particularly for ash concentration forecasts in the troposphere. The ensemble is slightly over-spread, or under-confident, within the troposphere at the first output time step T + 6, thought to be attributable to a known deficiency in the ensemble perturbation scheme in use at the time of this study, but improves with probability forecasts becoming well-calibrated here by the end of the period. Conversely, an increasing tendency towards over-confident forecasts is seen in the stratosphere, which again mirrors an expectation for ensemble spread to fall away at higher altitudes in the meteorological ensemble. Results in the volcanic ash case are also broadly similar between the three different eruption scenarios considered in the study, suggesting that good ensemble performance might apply to a wide range of eruptions with different heights and mass eruption rates.

The impact of Aeolus winds on near-surface wind forecasts over tropical ocean and high-latitude regions

Abstract. To detect global wind profiles and improve numerical weather prediction (NWP), the European Space Agency (ESA) launched the Aeolus satellite carrying a spaceborne Doppler wind lidar in 2018. After the successful launch, the European Centre for Medium-Range Weather Forecasts (ECMWF) performed the observing system experiments (OSEs) to evaluate the contribution of Aeolus data to NWP. This study aims to assess the impact of Aeolus wind assimilation in the ECMWF model on near-surface (10 m height) wind forecasts over tropical ocean regions by taking buoy measurements for reference and over high-latitude regions by taking weather station data for reference for the year 2020. The assessments were conducted mainly through inter-comparison analysis. The results show that Aeolus data assimilation has a limited impact on sea surface wind forecasts for tropical regions when compared with buoy measurements. For the high-latitude regions in the Northern Hemisphere, Aeolus is able to improve near-surface wind forecasts. This positive impact is more evident as the forecast time step is extended, during the first half year of 2020 and during the winter months. In addition, the v component tends to benefit more from the Aeolus observations than the u component. For the Southern Hemisphere, a few error reductions are observed but exist randomly. Overall, this in situ data-based assessment expands our understanding of the role of Aeolus data assimilation with the global NWP model in predicting near-surface wind for tropical oceans and high-latitude regions.

Physically constrained covariance inflation from location uncertainty

Abstract. Motivated by the concept of “location uncertainty”, initially introduced in Mémin (2014), a scheme is sought to perturb the “location” of a state variable at every forecast time step. Further considering Brenier's theorem (Brenier, 1991), asserting that the difference of two positive density fields on the same domain can be represented by a transportation map, we demonstrate that the perturbations consistently define a stochastic partial differential equation (SPDE) from the original PDE. It ensues that certain quantities, up to the user, are conserved at every time step. Remarkably, derivations following both the SALT (stochastic advection by Lie transport; Holm, 2015) and LU (location uncertainty; Mémin, 2014; Resseguier et al., 2017a) settings can be recovered from this perturbation scheme. Still, it offers broader applicability since it does not explicitly rely on Lagrangian mechanics or Newton's laws of force. For illustration, a stochastic version of the thermal shallow water equation is presented.

A contrastive learning-based framework for wind power forecast

The feature representation of wind power sequences is crucial in the modeling of short-tern wind power forecast, but the existing feature representation methodsmostly depend on the end-to-end model based on supervised learning, ignoring the superiority of self-supervised learning in the feature distribution of similar and non-similar sequences in the feature space. Therefore, this paper applies contrastive learning that is a type of self-supervised learning for the feature representation of wind power sequences, and proposes a two-stage framework for the wind power forecast (CLFNet). Specifically, the proposed framework is composed of the pre-training stage and the regression stage. The pre-training stage uses contrastive learning to supply a support for the optimization of the parameters of the network architectures, allowing the trend consistency sequences to obtain more correlated feature representations; the regression stage takes a projection layer, connected after the well-trained network architecture in the pre-training stage, to output wind power, and then fine-tunes the parameter of this network architecture using MSE loss. The proposed framework is applied for various network architectures including LSTM, CNN, and Transformer, and the results of RMSE, MAE, R2 confirm that the proposed framework is widely suitable for various basis network architectures and can outperform the initial basis network architectures by a mean increase rate of 8.94%, 8.5%, and 8.5%, respectively, among forecast steps. In addition, compared the proposed method with state-of-the-art schemes in the wind power forecasting system, the results show that the proposed framework can generate improved performance with accuracy and robustness.

Data assimilation in operator algebras

We develop an algebraic framework for sequential data assimilation of partially observed dynamical systems. In this framework, Bayesian data assimilation is embedded in a nonabelian operator algebra, which provides a representation of observables by multiplication operators and probability densities by density operators (quantum states). In the algebraic approach, the forecast step of data assimilation is represented by a quantum operation induced by the Koopman operator of the dynamical system. Moreover, the analysis step is described by a quantum effect, which generalizes the Bayesian observational update rule. Projecting this formulation to finite-dimensional matrix algebras leads to computational schemes that are i) automatically positivity-preserving and ii) amenable to consistent data-driven approximation using kernel methods for machine learning. Moreover, these methods are natural candidates for implementation on quantum computers. Applications to the Lorenz 96 multiscale system and the El Niño Southern Oscillation in a climate model show promising results in terms of forecast skill and uncertainty quantification.

An improved Wavenet network for multi-step-ahead wind energy forecasting

Accurate multi-step-ahead wind speed (WS) and wind power (WP) forecasting are critical to the scheduling, planning, and maintenance of wind farms. Previous forecasting methods tend to focus on improving forecast accuracy by integrating different models and disaggregating data while neglecting the forecasting ability of basic models. In addition, traditional multi-step-ahead output strategies have limitations that constrain the forecasting capability of models. To overcome the above challenges, this study proposes a novel forecasting model called ED-Wavenet-TF. It adopts two Wavenet networks as Encoder and Decoder connected by the multi-head self-attention mechanism. And, teacher forcing is used as the multi-step-ahead output strategy for WS and WP forecasting. In the training phase, ED-Wavenet-TF uses a portion of the actual data to correct the errors at the intermediate forecasting steps, while in the forecasting phase, it runs through an inference loop to make forecasts. In this study, two WS datasets and two WP datasets are used to validate the performance of ED-Wavenet-TF with univariate input. The results show that compared with Wavenet, the symmetric mean absolute percentage error of ED-Wavenet-TF at four forecasting steps is lower by at least 4.8577% on average for the WS datasets and 8.9463% on average for the WP datasets. The advantages of ED-Wavenert-TF over ten comparable models are confirmed by four evaluation indicators and the Harvey, Leybourne, and Newbold statistical hypothesis test. Moreover, ED-Wavenet-TF is extended to make multi-step-ahead forecasts with multivariate inputs, whose effectiveness is demonstrated on another open WS dataset.

Real-time identification of severe heat loads over external interface of lightweight thermal protection system

The lightweight thermal protection system (LTPS) is typically employed on the exterior surface of high-speed aircraft to prevent the penetration of high-temperature heat flux. The sequential function specification method (SFSM) is exploited in this study to solve the inverse issue in anticipation of the external-surface heating conditions of a lightweight thermal protection system by entering temperature data at a single measurement station. The results show that the inversion accuracy is significantly dependent on the temperature acquisition position as well as the forecast time step. Adding noise to the temperature data reveals that temperature noise aggravates the volatility of the inverse heat flux, necessitating an increase in the forecast time step to enhance the inverse accuracy. The experiment of a light thermal protection system demonstrates that the approach can efficiently and correctly recover the boundary heat flux by employing the inner wall temperature of the thermal protection structure.

Extensions of the Lee–Carter model to project the data‐driven rotation of age‐specific mortality decline and forecast coherent mortality rates

AbstractAccurate forecasts and analyses of mortality rates are essential to many operational practices, such as the designing of pension schemes. Recent studies have focused on the age coherent forecasts, such that long‐term predictions will not diverge infinitely among ages. Although intuitively and biologically reasonable, this age coherence property is lost when the seminal Lee–Carter (LC) model and almost all its existing extensions are employed. Those include the famous Li–Lee (LL) model, which is argued to produce coherent forecasts of mortality rates across populations, but not ages. In this paper, we progressively explore two effective frameworks with three extensions, by applying the LL specification to the single‐population model (LC‐S) and by introducing time‐dependent age patterns of mortality declines in the forecasting steps with fixed (LC‐VF) and variant (LC‐V) intercepts when modeling those patterns. In particular, the recommended LC‐V model is motivated from the intermediate LC‐S and LC‐VF and resolves all remaining issues of both models. Also, the final LC‐V model is easy to implement, incorporates the dynamic and rotating age‐specific patterns of mortality decline, provides age‐coherent forecasts of mortality rates in the long run, and is straightforwardly extensible to multi‐population cases. Using a large sample of 15 countries, we show that the LC‐V model and its multipopulation counterpart can consistently improve the forecasting accuracy of the competing LC, LL and other extensions in both single‐population and multipopulation scenarios. Long‐term analyses up to 2100 further demonstrate the existence of age coherence when forecasting with the proposed LC‐V model.

Model-free dynamic management strategy for low-carbon home energy based on deep reinforcement learning accommodating stochastic environments

This paper presents a model-free dynamic optimal management strategy for a low-carbon home energy management system (HEMS) based on deep reinforcement learning (DRL). The method can ideally handle the uncertainties and dynamics of renewable energy and demand-side load. Firstly, the load model is established by a deep Q network (DQN) algorithm with the advantage of ignoring traditional forecasting steps on stochastic environments such as renewable energy generation, load demand, price, etc. Then multi-agents are established for dynamic management based on the DRL. Through “dynamic acquisition, dynamic decision” mechanism, the proposed model-free strategy achieves real-time energy management that can adaptively respond to stochastic environments. Secondly, considering the constraints of system carbon emissions and carbon trading, the proposed strategy can minimize the energy consumption cost, carbon trading cost, and user satisfaction penalties. Ultimately, the effectiveness of the proposed strategy is verified through case studies. Experimental results demonstrate that the strategy can significantly reduce the overall cost, including a 36.7% reduction in carbon trading. At the same time, user satisfaction penalties are reduced by 50.2%. Further, the agent hyperparameter could also be adjusted to capture the trade-off between cost savings and satisfaction penalties. And compared with the traditional forecast-based management strategy, it overcomes the problem of uncertainties and avoids forecasting errors to better accommodate the stochastic environment.

Multi-Step Ahead Natural Gas Consumption Forecasting Based on a Hybrid Model: Case Studies in The Netherlands and the United Kingdom

With worldwide activities of carbon neutrality, clean energy is playing an important role these days. Natural gas (NG) is one of the most efficient clean energies with less harmful emissions and abundant reservoirs. This work aims at developing a swarm intelligence-based tool for NG forecasting to make more convincing projections of future energy consumption, combining Extreme Gradient Boosting (XGBoost) and the Salp Swarm Algorithm (SSA). The XGBoost is used as the core model in a nonlinear auto-regression procedure to make multi-step ahead forecasting. A cross-validation scheme is adopted to build a nonlinear programming problem for optimizing the most sensitive hyperparameters of the XGBoost, and then the nonlinear optimization is solved by the SSA. Case studies of forecasting the Natural gas consumption (NGC) in the United Kingdom (UK) and Netherlands are presented to illustrate the performance of the proposed hybrid model in comparison with five other intelligence optimization algorithms and two other decision tree-based models (15 hybrid schemes in total) in 6 subcases with different forecasting steps and time lags. The results show that the SSA outperforms the other 5 algorithms in searching the optimal parameters of XGBoost and the hybrid model outperforms all the other 15 hybrid models in all the subcases with average MAPE 4.9828% in NGC forecasting of UK and 9.0547% in NGC forecasting of Netherlands, respectively. Detailed analysis of the performance and properties of the proposed model is also summarized in this work, which indicates it has high potential in NGC forecasting and can be expected to be used in a wider range of applications in the future.