Electricity Load Forecasting Using Machine Learning Techniques

Electricity load forecasting has become increasingly important due to the strong impact on the operational efficiency of the power system. However, the accurate load prediction remains a challenging task due to several issues such as the nonlinear character of the time series or the seasonal patterns it exhibits. A large variety of techniques have been proposed to this aim, such as statistical models, fuzzy systems or artificial neural networks. The Support Vector Machines (SVM) have been widely applied to the electricity load forecasting with remarkable results. In this chapter, the authors study the performance of the classical SVM in the problem of electricity load forecasting. Next, an algorithm is developed that takes advantage of the local character of the time series. The method proposed first splits the time series into homogeneous regions using the Self Organizing Maps (SOM) and next trains a Support Vector Machine (SVM) locally in each region. The methods presented have been applied to the prediction of the maximum daily electricity demand. The properties of the time series are analyzed in depth. All the models are compared rigorously through several objective functions. The experimental results show that the local model proposed outperforms several statistical and machine learning forecasting techniques.

Time Series Based House Sale Value Market Forecasting Using Genetically Evolved Neural Networks

Accurate forecasting of the house sale value market is important for individual investors, business investors, banks and mortgage companies. This chapter uses fundamentals of Genetic Algorithms (GAs) and Artificial Neural Networks (ANNs) to derive and implement a hybrid, genetically evolved feedforward ANN model that predicts next month house sale prices. Derived model results are compared with results obtained using a linear regression model and an Adaptive Neuro Fuzzy Inference System (ANFIS). The proposed model returned lower Root Mean Square Error (RMSE), Absolute Mean Error (MAE), Mean Square Error (MSE) and Mean Absolute Percent Error (MAPE) results compared with the linear regression and ANFIS models. For case studies real monthly data of USA housing prices from 1963 to 2007 were used.

Implementing Business Intelligence in Electricity Markets

The authors elaborate on three basic ideas that should guide the implementation of business intelligence tools. First, the authors advocate for closing the gap between structured information and contextual information. Second, they emphasize the need for adopting the point of view of the organization to assess the relevance of any proposal. In the third place, they remark that any new tool is expected to become a relevant instrument to enhance the learning of the organization and to generate explicit knowledge. To illustrate their point, they discuss how to set up a forecasting support system to predict electricity consumption that converts raw time series data into market intelligence, to meet the needs of a major organization operating at the Spanish electricity markets.

Analysis and Forecasting of Port Logistics Using TEI@I Methodology

In this chapter, the authors propose an integrated forecasting model based on the TEI@I methodology for port logistics forecasting. This model analyzes the port logistics time series data and other information several steps. In the first step, several econometric models are built to forecast the linear segment of port logistics time series. In the second step, a radial basis function neural network is developed to predict the non-linear segment of the time series. In the third step, the event-study method and expert system techniques are applied to evaluate the effects economic and other events that may impact port logistics. In the final step, synthetic forecasting results are obtained based on the integration of the predictions from first three steps. For illustration, Hong Kong port’ container throughout series is used for a case study. The empirical results show the effectiveness of the TEI@I integrated model for port logistics forecasting.

Electricity Demand Forecasting

Electricity demand forecasting has attracted the attention of many researchers and power company staff. It still does so because with better forecasting, power companies can approach exact plans with no over- or –under planning. This is reflected as being the right investment in terms of time, money, and performance. In essence a good demand forecast means the right investment plan and therefore, satisfied customers. In reality this is the objective of any business; to be able to estimate the demand as close to reality as possible. The number and extent of demand forecasting methodologies and models developed is tremendous, however, there exists no novel technique that can serve all situations. Basically forecasting models can be divided into statistically based and intelligence-based models. A description of forecasting models helps in identifying the characteristics, features, and strengths of each model. The selection of the most suitable forecasting algorithm is not an easy process. The time frame of the forecast, data availability, the accuracy and cost of the forecast, the application and purpose of the forecast are some of the important parameters in the selection process. A case study of two forecasting models used in Jordan is presented. The discussion of the case study shows that load forecasting in Jordan is based on an intelligence-based model for short term forecasting, and on a combination of traditional statistically-based models for long term forecasting.

Forecasting Crude Oil Demand Using a Hybrid SVR and Markov Approach

In this chapter, a hybrid support vector regression (SVR) and Markov forecasting approach is proposed for energy demand prediction. The original time series of energy demand is firstly decomposed into the general trend series and the random fluctuation series. Then the SVR method is used to model the general trend series and the Markov forecasting method is used to model the random fluctuation series so that the tendencies of two series can be accurately predicted. The prediction results of two series are integrated to formulate an ensemble output for future energy demand. The proposed forecasting approach makes full use of the historical time series information so as to improve the forecasting precision of time series with large random fluctuation. To illustrate the applicability and capability of the proposed approach, it is used to analyze and forecast world crude oil demand. For verification, the proposed approach is compared with SVR method, Markov forecasting method and ARIMA. The results show that the hybrid SVR and Markov forecasting approach proposed in this chapter can be applied successfully and provide high accuracy and reliability for forecasting world crude oil demand.

On the Efficient Dynamical Financial Analysis Computation Supported by UML(VR)

Dynamic Financial Analysis (DFA) is disciplines of actuarial science developed to asses and manage risk. The structure of the problem, simulations-based approach as well as great user demands issue a challenge in terms of defining and controlling a computer software system. In order to properly support an efficient software and data allocation onto a distributed hardware environment we have to specify their basic characteristics. In this chapter we suggest using UML notation extended with vertical relations joining the information represented by different kinds of UML diagrams. Next this information is transformed into a graph notation that supports the optimal allocation of a new DFA process with the polynomial computational complexity. This way we are able to optimize DFA calculations on a given hardware environment without any changes to the software system.

Increasing Translation Invariant Morphological Forecasting Models for Stock Market Prediction

Statistical models have been widely used for the purpose of forecasting. However, it has some limitations regarding its performance, which prevents an automatic forecasting system development. In order to overcome such limitations, Artificial Neural Networks (ANNs), Evolutionary Algorithms (EAs) and Fuzzy Systems (FSs) based approaches have been proposed for nonlinear time series modeling. However, a dilemma arises from all these models regarding financial time series, which follow a Random Walk (RW) model, where the forecast of such time series exhibits a characteristic one step shift regarding original data. In this way, this work presents a new approach, referred to as Increasing Translation Invariant Morphological Forecasting (ITIMF) model, to overcome the RW dilemma for financial time series forecasting, which performs an evolutionary search for the minimum dimension to determining the characteristic phase space that generates the financial time series phenomenon. It is inspired on Takens Theorem and consists of an intelligent hybrid model composed of a Modular Morphological Neural Network (MMNN) combined with a Modified Genetic Algorithm (MGA), which searches for the particular time lags capable of a fine tuned characterization of the time series and estimates the initial (sub-optimal) parameters (weights, architecture and number of modules) of the MMNN. Each individual of the MGA population is trained by the Back Propagation (BP) algorithm to further improve the MMNN parameters supplied by the MGA. After adjusting the model, it performs a behavioral statistical test and a phase fix procedure to adjust time phase distortions observed in financial time series. Furthermore, an experimental analysis is conducted with the proposed model using ten real world financial time series. Five well-known performance metrics and an evaluation function are used to assess the performance of the proposed model and the obtained results are compared to classical models presented in literature.

A Mixture Price Trend Model for Long-Term Risk Management

In financial forecasting, a long-standing challenging issue is to develop an appropriate model for forecasting long-term risk management of enterprises. In this chapter, using financial markets as an example, we introduce a mixture price trend model for long-term forecasts of financial asset prices with a view to applying it for long-term financial risk management. The key idea of the mixture price trend model is to provide a general and flexible way to incorporate various price trend behaviors and to extract information from price trends for long-term forecasting. Indeed, the mixture price trend model can incorporate model uncertainty in the price trend model, which is a key element for risk management and is overlooked in some of the current literatures. The mixture price trend model also allows the incorporation of users’ subjective views on long-term price trends. An efficient estimation method is introduced. Statistical analysis of the proposed model based on real data will be conducted to illustrate the performance of the model.

Macroeconomic Forecasting Using Genetic Programming Based Vector Error Correction Model

Vector autoregressions are widely used in macroeconomic forecasting since they became known in the 1970s. Extensions including vector error correction models, co-integration and dynamic factor models are all rooted in the framework of vector autoregression. The three important extensions are demonstrated to have formal equivalence between each other. Above all, they all emphasize the importance of “common trends” or “common factors”. Many researches, including a series of work of Stock and Watson, find that “common factor” models significantly improve accuracy in forecasting macroeconomic time series. This study follows the work of Stock and Watson. The authors propose a hybrid framework called genetic programming based vector error correction model (GPVECM), introducing genetic programming to traditional econometric models. This new method could construct common factors directly from nonstationary data set, avoiding differencing the original data and thus preserving more information. The authors’ model guarantees that the constructed common factors satisfy the requirements of econometric models such as co-integration, in contrast to the traditional approach. Finally but not trivially, their model could save lots of time and energy from repeated work of unit root tests and differencing, which they believe is convenient for practitioners. An empirical study of forecasting US import from China is reported. The results of the new method dominates those of the plain vector error correction model and the ARIMA model.