Mid-term Forecasts Research Articles

Forecasting U.S. Fossil Energy Consumption: Advancing Accuracy with Multilayer Perceptron and Residual Learning

Accurate midterm forecasts of fossil fuel consumption are essential for effective energy planning, economic management, and resource allocation. While machine learning models have demonstrated their efficacy in handling large-scale nonlinear datasets, many, including Multilayer Perceptrons (MLPs), suffer from performance degradation with increased depth. Fortunately, recent studies have revealed that Residual Networks (ResNets) can mitigate or even overcome this challenge. In this paper, we propose a Weighted Residual Network based on MLP to enhance predictive performance. We employ the Adam algorithm for model training and utilize the Gridsearch algorithm for hyperparameter tuning. In the application section, we develop predictive models using three case studies: natural gas, petroleum, and total fossil fuel consumption. We validate the effectiveness of our proposed model and compare it with ten other machine learning models. Our findings demonstrate that our proposed model consistently outperforms others in all three cases, underscoring its superior performance in midterm forecasting of fossil fuel consumption.

The McNish and Lincoln Solar-Activity Predictions: The Method and Its Performance

The McNish and Lincoln (ML) method, introduced in 1949, was one of the first attempts to produce midterm forecasts of solar activity, up to 12 months ahead. However, it has been poorly described and evaluated in the past literature, in particular its actual operational implementation by NOAA. Here, we reconstruct the exact formulation of the method, as it was applied since the early 1970s, and we provide a full mathematical derivation of the prediction errors. For bench-marking the method, we also produce monthly predictions over the past 190 years, from 1833 (Cycle 8) to 2023 (Cycle 25), and develop statistics of the differences between the predictions and the observed 13-month smoothed sunspot-number (SSN) time series, according to the phase in the solar cycle. Our analysis shows that the ML method is heavily constrained because it is primarily based on the mean of all past cycles, which imposes a fixed amplitude and length and suffers from a temporal smearing that grows towards the end of the solar cycle. We find that predictions are completely unreliable in the first 12 months of the cycle, and over the last two years preceding the ending minimum (around 130 months), and beyond this minimum. By contrast, in the course of the cycle (months 18 to 65), ML predictions prove to be reliable over a time range of up to 50 months (4.2 years), thus much longer than the 12-month conventional range used so far. However, we find that predictions then suffer from systematic under-(over-)estimates for cycles that have a higher (lower) amplitude than the base mean cycle. Overall, we conclude that although the ML method provides valid prediction errors, it suffers from strong limitations, with very little room for improvement, as it indifferently merges all past cycles into a single fixed statistics.

A univariate time series methodology based on sequence-to-sequence learning for short to midterm wind power production

The biggest wind farm of Turkey is placed at Manisa which is located in the Aegean Region. Electricity is a non-storable commodity for that reason, it is very important to have a strong forecast and model of the potential electricity production to plan the electricity loads. In this study, the aim is to model and forecast electricity production of the wind farms located at Manisa by using a univariate model based on sequence-to-sequence learning. The forecasting range of the study is from short term to midterm. The strength of the proposed model is that; it only needs its own lagged value to make forecasts. The empirical evidences show that the model has high coefficient of variation (R2) in short term and moderate R2 in the midterm forecast. Although in the mid-range forecasts R2 slightly decreases mean squared error and mean absolute error shows that the model is accurate also in the midterm forecasts. The proposed model is not only strong in hourly electricity production forecasts but with a slight modification also in forecasting the minimum, maximum and average electricity production for a fixed range. This study concludes with two fresh and intriguing future research ideas.

Short- and mid-term forecasts of actual evapotranspiration with deep learning

• Deep learning is applied for nowcasting and forecasting of evapotranspiration. • Weather variables and soil moisture are used as input for the deep learning models. • The models were applied to seven distinct climate zones within the continental U.S. • Results reveal great potential for evapotranspiration forecasting. Evapotranspiration is a key component of the hydrologic cycle. Accurate short-, medium-, and long-term forecasts of actual evapotranspiration (ET a ) are crucial not only for quantifying the impacts of climate change on the water and energy balance, but also for real-time estimation of crop water demand and irrigation water allocation in agriculture. Despite considerable advances in satellite remote sensing technology and the availability of long ground-measured and remotely sensed ET a timeseries, real-time ET a forecasts are deficient. Applying a state-of-the-art deep learning (DL) approach, Long Short-Term Memory (LSTM) models were employed to nowcast (real-time) and forecast (ahead of time) ET a based on (1) major meteorological and ground-measured (i.e., soil moisture) input variables and (2) long ET a timeseries from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard of the NASA Aqua satellite. The conventional LSTM and convolutional LSTM (ConvLSTM) DL models were evaluated for seven distinct climatic zones across the contiguous United States. The employed LSTM and ConvLSTM models were trained and evaluated with data from the National Climate Assessment-Land Data Assimilation System (NCA-LDAS) and with MODIS/Aqua Net Evapotranspiration MYD16A2 product data. The obtained results indicate that when major atmospheric and soil moisture input variables are used for the conventional LSTM models, they yield accurate daily ET a forecasts for short (1, 3, and 7 days) and medium (30 days) time scales, with normalized root mean squared errors (NRMSE) and Nash-Sutcliffe efficiencies (NSE) of less than 10% and greater than 0.77, respectively. At the watershed scale, the univariate ConvLSTM models yielded accurate weekly spatiotemporal ET a forecasts (mean NRMSE less than 6.4% and NSE greater than 0.66) with higher computational efficiency for various climatic conditions. The employed models enable precise forecasts of both the current and future states of ET a , which is crucial for understanding the impact of climate change on rapidly depleting water resources.

Forecasting oil and gold volatilities with sentiment indicators under structural breaks

This paper contributes to the literature on forecasting the realized volatility of oil and gold by (i) utilizing the Infinite Hidden Markov (IHM) switching model within the Heterogeneous Autoregressive (HAR) framework to accommodate structural breaks in the data and (ii) incorporating, for the first time in the literature, various sentiment indicators that proxy for the speculative and hedging tendencies of investors in these markets as predictors in the forecasting models. We show that accounting for structural breaks and incorporating sentiment-related indicators in the forecasting model does not only improve the out-of-sample forecasting performance of volatility models but also has significant economic implications, offering improved risk-adjusted returns for investors, particularly for short-term and mid-term forecasts. We also find evidence of significant cross-market information spilling over across the oil, gold, and stock markets that also contributes to the predictability of short-term market fluctuations due to sentiment-related factors. The results highlight the predictive role of investor sentiment-related factors in improving the forecast accuracy of volatility dynamics in commodities with the potential to also yield economic gains for investors in these markets.

A Missing Data on Covid-19 Forecasts

Mathematical and computational studies of Covid-19 have underestimated the influence that other countries have on their daily records. To visualize this, a Granger causality analysis was implemented in Python to determine if the cases registered in Brazil, Chile, Colombia, Ecuador, Panama, Paraguay, Peru and the USA have any effect on Venezuela, and between all of them. Finally, this paper highlights the need to incorporate causality analysis employing only the cases of Covid-19 to improve mid and long term forecasts.

Balance Model of COVID-19 Epidemic Based on Percentage Growth Rate

The paper examines the possibility of using an alternative approach to predicting statistical indicators of a new COVID-19 virus type epidemic. A systematic review of models for predicting epidemics of new infections in foreign and Russian literature is presented. The accuracy of the SIR model for the spring 2020 wave of COVID-19 epidemic forecast in Russia is analyzed. As an alternative to modeling the epidemic spread using the SIR model, a new CIR discrete stochastic model is proposed based on the balance of the epidemic indicators at the current and past time points. The new model describes the dynamics of the total number of cases (C), the total number of recoveries and deaths (R), and the number of active cases (I). The system parameters are the percentage increase in the C(t) value and the characteristic of the dynamic balance of the epidemiological process, first introduced in this paper. The principle of the dynamic balance of epidemiological process assumes that any process has the property of similarity between the value of the total number of cases in the past and the value of the total number of recoveries and deaths at present. To calculate the values of the dynamic balance characteristic, an integer linear programming problem is used. In general, the dynamic characteristic of the epidemiological process is not constant. An epidemiological process the dynamic characteristic of which is not constant is called non-stationary. To construct mid-term forecasts of indicators of the epidemiological process at intervals of stationarity of the epidemiological process, a special algorithm has been developed. The question of using this algorithm on the intervals of stationarity and non-stationarity is being examined. Examples of the CIR model application for making forecasts of the considered indicators for the epidemic in Russia in May-June 2020 are given.

Mid-Epidemic Forecasts of COVID-19 Cases and Deaths: A Bivariate Model Applied to the UK.

Background The evolution of the COVID-19 epidemic has been accompanied by efforts to provide comparable international data on new cases and deaths. There is also accumulating evidence on the epidemiological parameters underlying COVID-19. Hence, there is potential for epidemic models providing mid-term forecasts of the epidemic trajectory using such information. The effectiveness of lockdown or lockdown relaxation can also be assessed by modelling later epidemic stages, possibly using a multiphase epidemic model. Methods Commonly applied methods to analyse epidemic trajectories or make forecasts include phenomenological growth models (e.g., the Richards family of densities) and variants of the susceptible-infected-recovered (SIR) compartment model. Here, we focus on a practical forecasting approach, applied to interim UK COVID data, using a bivariate Reynolds model (for cases and deaths), with implementation based on Bayesian inference. We show the utility of informative priors in developing and estimating the model and compare error densities (Poisson-gamma, Poisson-lognormal, and Poisson-log-Student) for overdispersed data on new cases and deaths. We use cross validation to assess medium-term forecasts. We also consider the longer-term postlockdown epidemic profile to assess epidemic containment, using a two-phase model. Results Fit to interim mid-epidemic data show better fit to training data and better cross-validation performance for a Poisson-log-Student model. Estimation of longer-term epidemic data after lockdown relaxation, characterised by protracted slow downturn and then upturn in cases, casts doubt on effective containment. Conclusions Many applications of phenomenological models have been to complete epidemics. However, evaluation of such models based simply on their fit to observed data may give only a partial picture, and cross validation against actual trends is also valuable. Similarly, it may be preferable to model incidence rather than cumulative data, although this raises questions about suitable error densities for modelling often erratic fluctuations. Hence, there may be utility in evaluating alternative error assumptions.

Short to Long-Term Forecasting of River Flows by Heuristic Optimization Algorithms Hybridized with ANFIS

Accurate forecast of short-term to long-term streamflow prediction is of great importance for water resources management. However, with the advent of novel hybrid machine learning methods, it remains unclear whether these hybrid models can outperform the traditional streamflow forecast models. Therefore, in this study, we trained and tested the performance of several evolutionary algorithms, including Fire-Fly Algorithm(FFA), Genetic Algorithm (GA), Grey Wolf Optimization (GWO), Particle Swarm Optimization (PSO), and Differential Evolution (DE) hybridized with ANFIS. Three forecast horizons, short-term (Daily), mid-term (Weekly and Monthly) and long-term (Annual) with fifteen input-output combinations, a total of 90 models, were developed and tested. A Monte Carlo Simulation (MCS) framework is used for uncertainty analysis. Daily inflow to the Karun III dam, located in the southeast of Iran, for the period of June 2005 to December 2016 were used. Results indicated that: 1) All developed hybrid algorithms significantly outperformed the traditional ANFIS model performance for all prediction horizons. The best hybrid models were ANFIS-GWO1, ANFIS-GWO7 and ANFIS-GWO11 such that the values of R2, RMSE, NSE, and RAE were improved by 12%, 10%, 18.5% and 14.3% for the short-term forecasts, 15%, 13%, 20% and 21.1% for the mid-term forecasts, and 10.3%, 7.5%, 10.5% and 14% for the long-term forecasts; 2) Uncertainty analysis indicates that nearly all hybrid models have significantly reduced uncertainty levels compared to the traditional ANFIS model; and 3) A simple explicit equation based on the hybrid ANFIS results was provided for streamflow forecasting, which is a major advantage compared to the classical blackbox machine learning models.

Improving real-time reservoir operation during flood season by making the most of streamflow forecasts

Streamflow forecasts play an important role in real-time reservoir operation, especially during the flood season. However, nowcast, short-term, and mid-term forecasts have significant differences in their uncertainties, leading to a gap between these forecasts and their combined use. To make the most of these forecasts, this paper defines a reliable forecast horizon (RFH) and proposes a model based on the three-stage-based operation rule (Three-OR), in which the streamflow forecasts are jointly used with different levels of uncertainty. To evaluate the effectiveness and efficiency of the proposed model, it is applied to the Wan’an Reservoir located at the midstream of the Ganjiang River in China with the objectives of flood control and water utilization (i.e., hydropower generation and water supply). The results reveal the following: 1) both flood control and water utilization can be improved when compared with the current operation rule; 2) floods with a higher peak and larger volume gain more improvements using Three-OR; 3) a smaller difference in the forecast error produces better operation performance in water utilization; and 4) the extension of the RFH is beneficial for reservoir operation, as it can help to improve flood control if the flood is large; otherwise, it can help improve water utilization. In summary, the proposed model based on Three-OR can make the best use of streamflow information, thus improving the real-time reservoir operation during the flood season.

Forecasting Long-Term Electricity Demand Time Series Using Artificial Neural Networks

Accurate forecast of electricity load are increasingly important. We present a method to forecast long-term weather-dependent hourly electricity load using artificial neural networks. The fully connected dense artificial neural networks with 5 hidden layers and 1,024 hidden nodes per layer are trained using historic data from 2006 to 2015. Input parameters comprise calendrical information, annual peak loads and weather data. The results are benchmarked against the method to forecast electric loads used in the current mid-term adequacy forecasts published by the European Network of Transmission System Operators (entso-e). For validation year 2016, our approach shows a mean absolute percentage error of 2.8%, whereas the common approach as used by entso-e shows an average error of 4.8% using peak load scaling. Further, we conduct forecasts for Germany, Sweden, Spain, and France for scenario year 2025 and assess parameter variations. Our approach can serve to increase prediction accuracy of future electricity loads.

Strategic foresight of staffing needs of innovative development of socio-economic systems

The research has been conducted in order to develop theoretical and methodological approaches to forecasting the future needs of the innovative economy in personnel to ensure the demand for formed human resources capacity. Development of basic techniques and instruments of calculation of need of innovative economy for highly qualified personnel on long-term and medium-term prospects has been based on such methodological approaches as system, activity and competence-based approaches, situational approach, methodology of the interactive, understanding and interpretive sociology, methodology of synergy, network thinking and social management, methodology of the modal analysis. A number of methodological problems of the existing systems of providing innovative economy in a personnel has been revealed, described and systematized: the forecast of balance of a labor resources does not consider changes of working conditions, modernization of production assets and new technologies; planning of target figures of enrollment of students for master programs is performed in a separation from real personnel requirements; lack of criteria system of detection of competence-based requirements of innovative economy; the imminent need of transition from scenario approach of forecasting of needs of economy for professional staff to approach “a threefold spiral”. The article proposes a methodological approach to forecasting the staffing needs of the innovative economy, which makes a provision for creating strategic long-term and mid-term forecasts and their annual adjustment at the operating level using the technologies of strategic foresight. The decomposition has been carried out by the types of forecasts for such characteristics as: level of regulation, purpose, key definitions, methods and approaches, factors, types of information, tools for providing staffing needs at each level, the main results. Such approach will allow to reduce the state costs for staff coaching and to prevent “combustion” of competences unclaimed by future labor market. The model of interaction of the labor market and the sphere of professional education, based on a strategic foresight of personnel requirements, has been designed to provide the independent, reasonable professional choice of youth today for satisfaction and social mobility of the population in future. The implementation of the proposed approach will allow to increase the quality of forecasts of personnel needs of the innovative economy and to create the basis for scientifically sound regulation of the development of general, professional and additional education at the federal and regional levels. This, in turn, will contribute to achieve maximum compliance between the personnel needs of the innovative economy and its formed resource potential.

Short- to Mid-term Day-Ahead Electricity Price Forecasting Using Futures

Due to the liberalization of markets, the change in the energy mix and the surrounding energy laws, electricity research is a dynamically altering field with steadily changing challenges. One challenge especially for investment decisions is to provide reliable short to mid-term forecasts despite high variation in the time series of electricity prices. This paper tackles this issue in a promising and novel approach. By combining the precision of econometric autoregressive models in the short-run with the expectations of market participants reflected in future prices for the short- and mid-run we show that the forecasting performance can be vastly increased while maintaining hourly precision. We investigate the day-ahead electricity price of the EPEX Spot for Germany and Austria and setup a model which incorporates the Phelix future of the EEX for Germany and Austria. The model can be considered as an AR24-X model with one distinct model for each hour of the day. We are able to show that future data contains relevant price information for future time periods of the day-ahead electricity price. We show that relying only on deterministic external regressors can provide stability for forecast horizons of multiple weeks. By implementing a fast and efficient lasso estimation approach we demonstrate that our model can outperform several other models in the literature.

Congressional Midterm Forecasts: A Trump Economic Difference?

Abstract Among certain politicians, pundits and people, the expectation was that President Trump would make an economic difference in the 2018 congressional elections. In particular, the belief was that his economic appeal, coupled with his economic policies, would favor Republican candidates. However, an application of the classic referendum model for forecasting congressional outcomes shows no detectable Trump economic effect. That is, the economic conditions prevailing prior to the election, measured in multiple ways, worked as usual, helping to predict the actual Republican 40 seat loss rather closely. Put another way, any incumbent administration, faced with this set of numbers, would likely have experienced the same outcome. In sum, the Trump presence, in its economic manifestations, did not positively impact the Republican party fortune at midterm time.

Time-series-based hybrid mathematical modelling method adapted to forecast automotive and medical waste generation: Case study of Lithuania.

The aim of the study was to create a hybrid forecasting method that could produce higher accuracy forecasts than previously used 'pure' time series methods. Mentioned methods were already tested with total automotive waste, hazardous automotive waste, and total medical waste generation, but demonstrated at least a 6% error rate in different cases and efforts were made to decrease it even more. Newly developed hybrid models used a random start generation method to incorporate different time-series advantages and it helped to increase the accuracy of forecasts by 3%-4% in hazardous automotive waste and total medical waste generation cases; the new model did not increase the accuracy of total automotive waste generation forecasts. Developed models' abilities to forecast short- and mid-term forecasts were tested using prediction horizon.

Dynamics of firm financial evolution and bankruptcy prediction

The optimal forecasting horizon of bankruptcy prediction models is usually one year. Beyond this point, their accuracy decreases as the horizon recedes. However, the ability of models to provide good mid-term forecasts is an essential characteristic for financial institutions due to prudential reasons. This is why we have studied a method of improving their forecasts up to a 5-year horizon. For this purpose, we propose to quantize how firm financial health changes over time, typify these changes and design models that fit each type. Our results show that, whatever the modeling technique used to design prediction models, model accuracy can be significantly improved when the horizon exceeds two years. They also show that when our method is used in combination with ensemble-based models, model accuracy is always improved whatever the forecasting horizon, when compared to traditional models used by financial institutions. The method we propose in this article appears to be a reliable solution that makes it possible to solve a real problem most models are unable to overcome, and it can therefore help financial companies comply with the current recommendations made by the Basel Committee on Banking Supervision. It also provides the scientific community (which is interested in designing reliable failure models) with insights about how the evolution of firms’ financial situations over time can be modeled and efficiently used to make forecasts.

Short- and Mid-Term Forecasting of Baseload Electricity Prices in the U.K.: The Impact of Intra-Day Price Relationships and Market Fundamentals

In this paper we investigate whether considering the fine structure of half-hourly electricity prices, the market closing prices of fundamentals (natural gas, coal and ${\rm CO}_{2}$ ) and the system-wide demand can lead to significantly more accurate short- and mid-term forecasts of APX U.K. baseload prices. We evaluate the predictive accuracy of a number of univariate and multivariate time series models over a three-year out-of-sample forecasting period and compare it against that of a benchmark autoregressive model. We find that in the short-term, up to a few business days ahead, a disaggregated model which independently predicts the intra-day prices and then takes their average to yield baseload price forecasts is the best performer. However, in the mid-term, factor models which explore the correlation structure of intra-day prices lead to significantly (as measured by the Diebold-Mariano test) better baseload price forecasts. At the same time, we observe that the inclusion of fundamental variables—especially natural gas prices (in the short-term) and coal prices (in the mid-term)—provides significant gains. The ${\rm CO}_{2}$ prices, on the other hand, generally do not improve the price forecasts at all, at least in the time period considered in this study (April 2009–December 2013).

Achieving Deepwater Success

Guest editorial Despite some uncertainty in the global economy, the long-term outlook for energy demand remains strong and the industry will continue to be challenged to meet hydrocarbon supply over the next decade. In its latest outlook, the International Energy Agency estimates that approximately 32% of oil production needed by the end of this decade has yet to be discovered or developed, and by 2035 it will be closer to 50%. In response to this challenge, the exploration and production (E&P) industry has become more ambitious in searching out new frontiers, with some notable successes over the past few years, particularly in deep water. More than half of all oil and gas reserves discovered worldwide over the past 10 years have been offshore and the majority of large finds have been in water depths of more than 500 m (1,640 ft). We are reminded that 10 years ago pre-salt reserves offshore Brazil were still not proven, and that deepwater discoveries in areas such as French Guiana, Tanzania, and Ghana were not even on our industry’s radar screens. It is estimated that approximately 200 new deepwater fields will enter production over the next 4 years. By 2020, production from deepwater fields will represent about one-third of total offshore production, representing about 10% of total global oil supply. Deepwater well capex is expected to grow from USD 47 billion to approximately USD 128 billion by 2020, with 71% of this growth taking place in the Atlantic basin—particularly in Brazil, Angola, Nigeria, the US Gulf of Mexico (GOM), and Nor-way. Other countries with solid deepwater activity include Australia, Egypt, and India. To achieve this significant growth, an increase in the number of floater rigs is forecast, from 200 in 2008 to more than 400 by 2020—an average of more than 20 new rigs per year. This estimate is consistent with mid-term forecasts of new deepwater rigs and suggests that the trend will be maintained until the end of the decade. Over the period 2013-2020, deepwater well costs are expected to slightly increase in real terms. While efficiency gains are expected in regions such as the US GOM, thanks to next-generation rigs entering the market and the continued integration trend of new drilling technologies, these will be offset by increasingly complex and deeper wells. Exploration and appraisal activity is set to continue along the same trends for the rest of this decade, with further expansion in basins with recent discoveries such as the Mexican GOM, Equatorial Atlantic, east Africa, and the eastern Mediterranean. Activity will also be strong in basins with less-recent discoveries but where technology and/or access drivers have opened additional opportunities, such as Brazil, west Africa, and the US GOM.

PREDICT – Decision support system for load forecasting and inference: A new undertaking for Brazilian power suppliers

One of the most desired aspects for power suppliers is the acquisition/sale of energy for a future demand. However, power consumption forecast is characterized not only by the variables of the power system itself, but also related to social–economic and climatic factors. Hence, it is imperative for the power suppliers to project and correlate these parameters. This paper presents a study of power load forecast for power suppliers, considering the applicability of wavelets, time series analysis methods and artificial neural networks, for both mid and long term forecasts. Both the periods of forecast are of major importance for power suppliers to define the future power consumption of a given region. The paper also studies the establishment of correlations among the variables using Bayesian networks. The results obtained are much more effective when compared to those projected by the power suppliers based on specialist information. The research discussed here is implemented on a decision support system, contributing to the decision making for acquisition/sale of energy at a future demand; also providing them with new ways for inference and analyses with the correlation model presented here.

Nokia Siemens Networks company: a case study of service visibility in global procurement

This research concentrates on the management of service visibility, which is an important concept in the field of procurement. The objective of this research is to propose a framework to analyse service visibility and forecasts in global procurement. The company under focus is Nokia Siemens Networks. The results are obtained via the contribution of logistics experts, business managers and process owners in NSN and can be generalised to similar telecommunication companies worldwide. We found the need for improvements both in the current processes and tools used in the procurement practises. Specifically, proactive communication and data sharing are needed to achieve information transparency, while the integration of stand-alone tools would eliminate transaction costs. In addition, long-term planning, mid-term forecasts and short-term demand visibility all support the effective management of service visibility through supplier visibility and accurate forecasts. New research avenues are opened in process co-design.