Forecasting 24-Hour Ahead Electricity Load Using Time Series Models

Forecasting electricity load is important for various purposes like planning, operation and control. Forecasts can save operating and maintenance costs, increase the reliability of power supply and delivery systems, and correct decisions for future development. This paper compares various time series methods to forecast 24 hours ahead of electricity load. The methods considered are the Holt-Winters smoothing, SARIMA Modeling, LSTM Network, Fbprophet and Tensorflow probability. The performance of each method is evaluated by using the forecasting accuracy criteria namely, the Mean Absolute Error and Root Mean Square Error. The National Renewable Energy Laboratory (NREL) residential energy consumption data are used to train the models. The results of this study show that SARIMA model is superior to the others for 24 hours ahead forecasts. Furthermore, a Bagging technique is used to make the predictions more robust. The obtained results show that by Bagging multiple time-series forecasts we can improve the robustness of the models for 24 hour ahead electricity load forecasting.

Time Series Forecasting Using Various Deep Learning Models

Time Series Forecasting (TSF) is used to predict the target variables at a future time point based on the learning from previous time points. To keep the problem tractable, learning methods use data from a fixed length window in the past as an explicit input. In this paper, we study how the performance of predictive models change as a function of different look-back window sizes and different amounts of time to predict into the future. We also consider the performance of the recent attention-based transformer models, which had good success in the image processing and natural language processing domains. In all, we compare four different deep learning methods (Recurrent Neural Network (RNN), Long Short-term Memory (LSTM), Gated Recurrent Units (GRU), and Transformer) along with a baseline method. The dataset (hourly) we used is the Beijing Air Quality Dataset from the website of University of California, Irvine (UCI), which includes a multivariate time series of many factors measured on an hourly basis for a period of 5 years (2010-14). For each model, we also report on the relationship between the performance and the look-back window sizes and the number of predicted time points into the future. Our experiments suggest that Transformer models have the best performance with the lowest Mean   Absolute Errors (MAE = 14.599, 23.273) and Root Mean Square Errors (RSME = 23.573, 38.131) for most of our single-step and multi-steps predictions. The best size for the look-back window to predict 1 hour into the future appears to be one day, while 2 or 4 days perform the best to predict 3 hours into the future.

Stock Movement Prediction Using Price Factor and Deep Learning

The development of machine learning methods and techniques has opened doors for investigation in many areas such as medicines, economics, finance, etc. One active research area involving machine learning is stock market prediction. This research paper tries to consider multiple techniques and methods for stock movement prediction using historical price or price factors. The paper explores the effectiveness of some deep learning frameworks for forecasting stock. Moreover, an architecture (TimeStock) is proposed which takes the representation of time into account apart from the price information itself. Our model achieves a promising result that shows a potential approach for the stock movement prediction problem.

Decision-Making Strategies on Smart Dairy Farms: A Review

Farm management and operations will drastically change due to access to real-time data, real-time forecasting and tracking of physical items in combination with Internet of Things (IoT) developments to further automate farm operations. Dairy farms have embraced technological innovations and procured vast amounts of permanent data streams during the past decade; however, the integration of this information to improve the whole farm decision-making process does not exist. It is now imperative to develop a system that can collect, integrate, manage, and analyze on-farm and off-farm data in real-time for practical and relevant environmental and economic actions. The developed systems, based on machine learning and artificial intelligence, need to be connected for useful output, a better understanding of the whole farming issue and environmental impact. Evolutionary Computing (EC) can be very effective in finding the optimal combination of sets of some objects and finally, in strategy determination. The system of the future should be able to manage the dairy farm as well as an experienced dairy farm manager with a team of the best agricultural advisors. All these changes should bring resilience and sustainability to dairy farming as well as improving and maintaining good animal welfare and the quality of dairy products. This review aims to provide an insight into the state-of-the-art of big data applications and EC in relation to smart dairy farming and identify the most important research and development challenges to be addressed in the future. Smart dairy farming influences every area of management and its uptake has become a continuing trend.

Scenario and Decision Analysis for Solar Energy in Egypt by 2035 Using Dynamic Bayesian Network

Bayesian networks are now considered to be a promising tool in the field of energy with different applications. In this study, the aim was to indicate the states of a previous constructed Bayesian network related to the solar energy in Egypt and the factors affecting its market share, depending on the followed data distribution type for each factor, and using either the Z-distribution approach or the Chebyshev’s inequality theorem. Later on, the separate and the conditional probabilities of the states of each factor in the Bayesian network were derived, either from the collected and scrapped historical data or from estimations and past studies. Results showed that we could use the constructed model for scenario and decision analysis concerning forecasting the total percentage of the market share of the solar energy in Egypt by 2035 and using it as a stable renewable source for generating any type of energy needed. Also, it proved that whenever the use of the solar energy increases, the total costs decreases. Furthermore, we have identified different scenarios, such as the best, worst, 50/50, and most likely one, in terms of the expected changes in the percentage of the solar energy market share. The best scenario showed an 85% probability that the market share of the solar energy in Egypt will exceed 10% of the total energy market, while the worst scenario showed only a 24% probability that the market share of the solar energy in Egypt will exceed 10% of the total energy market. Furthermore, we applied policy analysis to check the effect of changing the controllable (decision) variable’s states acting as different scenarios, to show how it would affect the target nodes in the model. Additionally, the best environmental and economical scenarios were developed to show how other factors are expected to be, in order to affect the model positively. Additional evidence and derived probabilities were added for the weather dynamic nodes whose states depend on time, during the process of converting the Bayesian network into a dynamic Bayesian network.

Artificial Neural Network-Based Short-Term Load Forecasting for Mymensingh Area of Bangladesh

Electrical load forecasting is considered to be one of the most indispensable parts of a modern-day electrical power system. To ensure a reliable and efficient supply of electric energy, special emphasis should have been put on the predictive feature of electricity supply. Artificial Neural Network-based approaches have emerged to be a significant area of interest for electric load forecasting research. This paper proposed an Artificial Neural Network model based on the particle swarm optimization algorithm for improved electric load forecasting for Mymensingh, Bangladesh. The forecasting model is developed and simulated on the MATLAB environment with a large number of training datasets. The model is trained based on eight input parameters including historical load and weather data. The predicted load data are then compared with an available dataset for validation. The proposed neural network model is proved to be more reliable in terms of day-wise load forecasting for Mymensingh, Bangladesh.

The Origin, Diffusion and a Comparison of Ordinary Differential Equations Numerical Solutions Used by SIR Model in Order to Predict SARS-CoV-2 in Nordic Countries

SARS-CoV-2 virus is currently one of the most infectious pathogens for humans. It started in China at the end of 2019 and now it is spread in all over the world. The origin and diffusion of the SARS-CoV-2 epidemic, is analysed based on the discussion of viral phylogeny theory. With the aim of understanding the spread of infection in the affected countries, it is crucial to modelize the spread of the virus and simulate its activity. In this paper, the prediction of coronavirus outbreak is done by using SIR model without vital dynamics, applying different numerical technique solving ordinary differential equations (ODEs). We find out that ABM and MRT methods perform better than other techniques and that the activity of the virus will decrease in April but it never cease (for some time the activity will remain low) and the next cycle will start in the middle July 2020 for Norway and Denmark, and October 2020 for Sweden, and September for Finland.

Enhancing Temporal Extrapolation of Wind Speed Using a Hybrid Technique: A Case Study in West Coast of Denmark

The demand for renewable energy is significantly increasing, major investments are being supplied to the wind power generation industry as a leading source of clean energy. The wind energy sector is entirely dependable and driven by the prediction of wind speed, which by the nature of wind is very stochastic and widely random. This s0tudy employs deep multi-fidelity Gaussian process regression, used to predict wind speeds for medium term time horizons. Data of the RUNE experiment in the west coast of Denmark were provided by the Technical University of Denmark, which represent the wind speed across the study area from the period between December 2015 and March 2016. The study aims to investigate the effect of pre-processing the data by denoising the signal using empirical wavelet transform (EWT) and engaging the vector components of wind speed to increase the number of input data layers for data fusion using deep multi-fidelity Gaussian process regression (GPR). The outcomes were compared using root mean square error (RMSE) and the results demonstrated a significant increase in the accuracy of predictions which demonstrated that using vector components of the wind speed as additional predictors exhibits more accurate predictions than strategies that ignore them, reflecting the importance of the inclusion of all sub data and pre-processing signals for wind speed forecasting models.

Increasing the Forecasting Fidelity of Current Collection System Operating Capability by Means of Contact Pressure Simulation Modelling

Current collection quality is one of the limiting factors when increasing trains movement speed in the rail sector. With the movement speed growth, the impact forces on the current collector from the rolling stock and the aerodynamic influence increase, which leads to the spread in the contact pressure values, separation of the current collector head from the contact wire, contact arcing and excessive wear of the contact elements. The upcoming trend in resolving this issue is the use of the automatic control systems providing stabilization of the contact pressure value. The present paper considers the features of the contemporary automatic control systems of the current collector’s pressure; their major disadvantages have been stated. A scheme of current collector pressure automatic control has been proposed, distinguished by a proactive influence on undesirable effects. A mathematical model of contact strips wearing has been presented, obtained in accordance with the provisions of the central composition rotatable design program. The analysis of the obtained dependencies has been carried out. The procedures for determining the optimal current collector pressure on the contact wire and the pressure control principle in the pneumatic drive have been described.

Constructing a Bayesian Network for Solar Energy in Egypt Using Life Cycle Analysis and Machine Learning Algorithms

In an era where machines run and shape our world, the need for a stable, non-ending source of energy emerges. In this study, the focus was on the solar energy in Egypt as a renewable source, the most important factors that could affect the solar energy’s market share throughout its life cycle production were analyzed and filtered, the relationships between them were derived before structuring a Bayesian network. Also, forecasted models were built for multiple factors to predict the states in Egypt by 2035, based on historical data and patterns, to be used as the nodes’ states in the network. 37 factors were found to might have an impact on the use of solar energy and then were deducted to 12 factors that were chosen to be the most effective to the solar energy’s life cycle in Egypt, based on surveying experts and data analysis, some of the factors were found to be recurring in multiple stages. The presented Bayesian network could be used later for scenario and decision analysis of using solar energy in Egypt, as a stable renewable source for generating any type of energy needed.

Time Series Simulation by Conditional Generative Adversarial Net

Generative Adversarial Net (GAN) has proved to be a powerful machine learning tool in image data analysis and generation. In this paper, we propose to use Conditional Generative Adversarial Net (CGAN) to learn and simulate time series data. The conditions include both categorical and continuous variables with different auxiliary information. Our simulation studies show that CGAN has the capability to learn different types of normal and heavy-tailed distributions, as well as dependent structures of different time series. It also has the capability to generate conditional predictive distributions consistent with training data distributions. We also provide an in-depth discussion on the rationale behind GAN and the neural networks as hierarchical splines to establish a clear connection with existing statistical methods of distribution generation. In practice, CGAN has a wide range of applications in market risk and counterparty risk analysis: it can be applied to learn historical data and generate scenarios for the calculation of Value-at-Risk (VaR) and Expected Shortfall (ES), and it can also predict the movement of the market risk factors. We present a real data analysis including a backtesting to demonstrate that CGAN can outperform Historical Simulation (HS), a popular method in market risk analysis to calculate VaR. CGAN can also be applied in economic time series modeling and forecasting. In this regard, we have included an example of hypothetical shock analysis for economic models and the generation of potential CCAR scenarios by CGAN at the end of the paper.

Forecasting Stock Indexes Using Bayesian Additive Regression Tree

Forecasting the stock market is a very challenging task. Various economic indicators such as GDP, exchange rates, interest rates, and unemployment have a substantial impact on the stock market. Time series models are the traditional methods used to predict stock market changes. In this paper, a machine learning method, Bayesian Additive Regression Tree (BART) is used in predicting stock market indexes based on multiple economic indicators. BART can be used to model heterogeneous treatment effects, and thereby works well when models are misspecified. It also has the capability to handle non-linear main effects and multi-way interactions without much input from financial analysts. In this research, BART is proposed to provide a reliable prediction on day-to-day stock market activities. By comparing the analysis results from BART and with time series method, BART can perform well and has better prediction capability than the traditional methods.

The Non-Stationary BINARMA(1,1) Process with Poisson Innovations: An Application on Accident Data

This paper considers the modelling of a non-stationary bivariate integer-valued autoregressive moving average of order one (BINARMA(1,1)) with correlated Poisson innovations. The BINARMA(1,1) model is specified using the binomial thinning operator and by assuming that the cross-correlation between the two series is induced by the innovation terms only. Based on these assumptions, the non-stationary marginal and joint moments of the BINARMA(1,1) are derived iteratively by using some initial stationary moments. As regards to the estimation of parameters of the proposed model, the conditional maximum likelihood (CML) estimation method is derived based on thinning and convolution properties. The forecasting equations of the BINARMA(1,1) model are also derived. A simulation study is also proposed where BINARMA(1,1) count data are generated using a multivariate Poisson R code for the innovation terms. The performance of the BINARMA(1,1) model is then assessed through a simulation experiment and the mean estimates of the model parameters obtained are all efficient, based on their standard errors. The proposed model is then used to analyse a real-life accident data on the motorway in Mauritius, based on some covariates: policemen, daily patrol, speed cameras, traffic lights and roundabouts. The BINARMA(1,1) model is applied on the accident data and the CML estimates clearly indicate a significant impact of the covariates on the number of accidents on the motorway in Mauritius. The forecasting equations also provide reliable one-step ahead forecasts.

Forecast of Polyethylene Properties in the Gas Phase Polymerization Aided by Neural Network

A major problem that affects the quality control of polymer in the industrial polymerization is the lack of suitable on-line measurement tools to evaluate the properties of the polymer such as melt and density indices. Controlling the polymerization in ordinary method is performed manually by taking samples, measuring the quality of polymer in the lab and registry of results. This method is highly time consuming and leads to producing large number of incompatible products. An online application for estimating melt index and density proposed in this study is a neural network based on the input-output data of the polyethylene production plant. Temperature, the level of reactors' bed, the intensity of ethylene mass flow, hydrogen and butene-1, the molar concentration of ethylene, hydrogen and butene-1 are used for the process to establish the neural model. The neural network is taught based on the actual operational data and back-propagation and Levenberg-Marquart techniques. The simulated results indicate that the neural network process model established with three layers (one hidden layer) for forecasting the density and the four layers for the melt index is able to successfully predict those quality properties.

River Stage-Discharge Forecasting Based on Multiple-Gauge Strategy Using EEMD-DWT-LSSVM Approach

This study presented hybrid pre-processing approach along with a conceptual model to enhance the accuracy of river discharge prediction. In order to achieve this goal, Ensemble Empirical Mode Decomposition algorithm (EEMD), Discrete Wavelet Transform (DWT) and Mutual Information (MI) were employed as a hybrid pre-processing approach conjugated to Least Square Support Vector Machine (LSSVM). A conceptual strategy namely multi-station model was developed to forecast the Souris River discharge more accurately. The strategy used herein was capable of covering uncertainties and complexities of river discharge modeling. DWT and EEMD was coupled, and the feature selection was performed for decomposed sub-series using MI to be employed in multi-station model. In the proposed feature selection method, some useless sub-series were omitted to achieve better performance. Results approved efficiency of the proposed DWT-EEMD-MI approach to improve accuracy of multi-station modeling strategies.

Improving Flash Flood Forecasting with a Bayesian Probabilistic Approach: A Case Study on the Posina Basin in Italy

The Flash Flood Guidance (FFG) provides the rainfall amount of a given duration necessary to cause flooding. The approach is based on the development of rainfall-runoff curves, which helps us to find out the rainfall amount that would cause flooding. An alternative approach, mostly experimented with Italian Alpine catchments, is based on determining threshold discharges from past events and on finding whether or not an oncoming flood has its magnitude more than some critical discharge thresholds found beforehand. Both approaches suffer from large uncertainties in forecasting flash floods as, due to the simplistic approach followed, the same rainfall amount may or may not cause flooding. This uncertainty leads to the question whether a probabilistic model is preferable over a deterministic one in forecasting flash floods. We propose the use of a Bayesian probabilistic approach in flash flood forecasting. A prior probability of flooding is derived based on historical data. Additional information, such as antecedent moisture condition (AMC) and rainfall amount over any rainfall thresholds are used in computing the likelihood of observing these conditions given a flash flood has occurred. Finally, the posterior probability of flooding is computed using the prior probability and the likelihood. The variation of the computed posterior probability with rainfall amount and AMC presents the suitability of the approach in decision making in an uncertain environment. The methodology has been applied to the Posina basin in Italy. From the promising results obtained, we can conclude that the Bayesian approach in flash flood forecasting provides more realistic forecasting over the FFG.

Comparison of Machine Learning Models for the Prediction of System Marginal Price of Greek Energy Market

The Greek Energy Market is structured as a mandatory pool where the producers make their bid offers in day-ahead basis. The System Operator solves an optimization routine aiming at the minimization of the cost of produced electricity. The solution of the optimization problem leads to the calculation of the System Marginal Price (SMP). Accurate forecasts of the SMP can lead to increased profits and more efficient portfolio management from the producer`s perspective. Aim of this study is to provide a comparative analysis of various machine learning models such as artificial neural networks and neuro-fuzzy models for the prediction of the SMP of the Greek market. Machine learning algorithms are favored in predictions problems since they can capture and simulate the volatilities of complex time series.

An IM-COH Algorithm Neural Network Optimization with Cuckoo Search Algorithm for Time Series Samples

Back propagation algorithm (BP) is a widely used technique in artificial neural network and has been used as a tool for solving the time series problems, such as decreasing training time, maximizing the ability to fall into local minima, and optimizing sensitivity of the initial weights and bias. This paper proposes an improvement of a BP technique which is called IM-COH algorithm (IM-COH). By combining IM-COH algorithm with cuckoo search algorithm (CS), the result is cuckoo search improved control output hidden layer algorithm (CS-IM-COH). This new algorithm has a better ability in optimizing sensitivity of the initial weights and bias than the original BP algorithm. In this research, the algorithm of CS-IM-COH is compared with the original BP, the IM-COH, and the original BP with CS (CS-BP). Furthermore, the selected benchmarks, four time series samples, are shown in this research for illustration. The research shows that the CS-IM-COH algorithm give the best forecasting results compared with the selected samples.

Forecasting Issues in Energy Markets within a Reg-ARIMA Framework

Electricity markets throughout the world have undergone substantial changes. Accurate, reliable, clear and comprehensible modeling and forecasting of different variables (loads and prices in the first instance) have achieved increasing importance. In this paper, we describe the actual state of the art focusing on reg-SARMA methods, which have proven to be flexible enough to accommodate the electricity price/load behavior satisfactory. More specifically, we will discuss: 1) The dichotomy between point and interval forecasts; 2) The difficult choice between stochastic (e.g. climatic variation) and non-deterministic predictors (e.g. calendar variables); 3) The confrontation between modelling a single aggregate time series or creating separated and potentially different models of sub-series. The noteworthy point that we would like to make it emerge is that prices and loads require different approaches that appear irreconcilable even though must be made reconcilable for the interests and activities of energy companies.

Load Forecasting Using Neural Network Integrated with Economic Dispatch Problem

High cost of fossil fuels and intensifying installations of alternate energy generation sources are intimidating main challenges in power systems. Making accurate load forecasting an important and challenging task for optimal energy planning and management at both distribution and generation side. There are many techniques to forecast load but each technique comes with its own limitation and requires data to accurately predict the forecast load. Artificial Neural Network (ANN) is one such technique to efficiently forecast the load. Comparison between two different ranges of input datasets has been applied to dynamic ANN technique using MATLAB Neural Network Toolbox. It has been observed that selection of input data on training of a network has significant effects on forecasted results. Day-wise input data forecasted the load accurately as compared to year-wise input data. The forecasted load is then distributed among the six generators by using the linear programming to get the optimal point of generation. The algorithm is then verified by comparing the results of each generator with their respective generation limits.