Sergio Valero

Short-Term Predictability of Load Series: Characterization of Load Data Bases

This paper proposes the use of two indicators of the predictability of the load series along with an accuracy value such as mean average percentage error as standard measures of load forecasting performance. Over the last 10 years, there has been a significant increase in load forecasting models proposed in engineering journals. Most of these models provide a description of the inner design of the model, the results from applying this model to a specific data base and the conclusions drawn from this application. However, a single accuracy value may not be sufficient to describe the performance of the model when applied to other data bases. The aim of this paper is to provide researchers with a tool that is able to assess the predictability of a load series and, therefore, contextualize the forecasting accuracy reported. Nine different data bases from the U.S. have been used; all of them include hourly load and temperature data.

Development of a methodology for improving the effectiveness of customer response policies through electricity-price patterns

The main objective of electricity regulators when establishing electricity markets is to decrease the cost of electricity through competition. However, this scenario can not be achieved without a full participation of the electricity demand by reacting against electricity prices. The aim of this research is to develop tools for helping customers and aggregators to join price and demand response programs, while helping them to hedge against the risk of short-term price volatility. In this way, the capacity of and hybrid methodology (self-organizing maps and statistical wards linkage) to classify high electricity market prices is analyzed. Besides, with the help of non-parametric estimation, some price-patterns were found in the above mentioned clusters. The knowledge contained within these patterns supplies customer market-based information on which to base its energy use decisions. The interest for this participation of customers in markets is growing in developed countries to obtain a higher elasticity in demand. Results show the capability of this approach to improve data management and select coherent policies to accomplish cleared demand offers amongst different price scenarios in a more flexible way.

Development of a model for short-term load forecasting with neural networks and its application to the electrical Spanish market

The study presented in this paper used Kohonens Self-Organized Maps, which is one of the more uncommon techniques based on neural networks in load forecasting. The aim of this study is not only to show that this technique is capable of producing accurate short-term load forecasting results which should not be neglected, but also to provide a deep and thorough analysis of these results in order to extract solid conclusions about the inner design of the network, the selection of variables and also about the training periods. In addition, an application for the Spanish electricity market is developed.

Standardization of short-term load forecasting models

There has been a significant production of load forecasting models over the last 5 years. These models present a wide variety of techniques, most of them using novel artificial intelligence approaches. Load forecasting is a complex matter and it is the result of several processes that, depending on the database, may be of more or less importance. However, most models focus their attention only on one process like the “forecasting engine”, neglecting other processes like variable selection or pre-processing. This paper proposes a standard scheme for load forecasting models that includes all sub-processes within load forecasting. The analysis of load forecasting models through this scheme allows identifying the effect of each process on the overall performance of the model. Also, proposing load forecasting models following this scheme will enhance benchmarking possibilities and hybridization of models. Finally, this paper presents such analysis of an actual load forecasting model.

Using a Self Organizing Map Neural Network for Short-Term Load Forecasting, Analysis of Different Input Data Patterns

This research uses a Self-Organizing Map neural network model (SOM) as a short-term forecasting method. The objective is to obtain the demand curve of certain hours of the next day. In order to validate the model, an error index is assigned through the comparison of the results with the real known curves. This index is the Mean Absolute Percentage Error (MAPE), which measures the accuracy of fitted time series and forecasts. The pattern of input data and training parameters are being chosen in order to get the best results. The investigation is still in course and the authors are proving different patterns of input data to analyze the different results that they will be obtained with each one. Summing up, this research tries to establish a tool that helps the decision making process, forecasting the short-term global electric load demand curve.

“Self-organizing maps” for identification of tire model longitudinal braking parameters of a vehicle on a roller brake tester and on flat ground

This paper discusses how to identify tire model coefficients that are used to compare longitudinal forces when braking. Those in the automotive world have worked extremely hard to obtain these parameters and different methods have been used to match the values of these parameters. This paper proposes the use of self-organizing maps to tackle this problem whereby interactively searches are carried out to find the optimum tire model parameters. The objective of this research is to prove the capability of self-organizing maps (SOMs) to classify a vehicle’s braking formula on a roller brake tester from the MOT (Ministry of Transport) and on flat ground. The neural network produced a good brake-slip ratio when presented with data that are not used in network training. This means that the methodology is feasible. This tool easily obtains the brake-slip equation of each experiment and the braking on two different experimental tests will be compared.

Short-term load forecasting of multiregion systems using mixed effects models

This paper presents an application of linear mixed models to short-term load forecasting. The starting point of the design is a currently working model at the Spanish Transport System Operator, which is based on linear autoregressive techniques and neural networks. The forecasting system currently forecasts each of the regions within the Spanish grid separately, even though the behavior of the load in each region is affected by the same factors in a similar way. The integration of several regions into a linear mixed model allows using the information from other regions to learn general behaviors present in all regions while also identifying individual deviation in each regions. This technique is especially useful when modeling the effect of special days for which information from the past is scarce. The model described has been applied to the three most relevant regions of the system, focusing on special day and improving the performance of both currently working models used as benchmark.

Short-term load forecasting: Revising how good we actually are

This paper proposes the use of an indicator of the predictability of the load series along with an accuracy value such as Mean Average Percentage Error as standard measures of load forecasting performance. Over the last 10 years, there has been a significant increase in load forecasting models proposed in engineering journals. Most of these models provide a description of the inner design of the model, the results from applying this model to a specific data base and the conclusions drawn from this application. However, a single accuracy value may not be sufficient to describe the performance of the model when applied to other data bases. The aim of this paper is to provide researchers with a tool that is able to assess the predictability of a load series and, therefore, contextualize the forecasting accuracy reported. Thirteen different data bases were used to determine its validity.

SOM for Getting the Brake Formula of a Vehicle on a Brake Tester and on Flat Ground

The objective of the research is to prove the capability of Self-Organizing Map (SOM) to classify brake formula of a vehicle on a bank of roller tester from the MOT (Ministry of transport) and on flat ground. The neural network demonstrated good generation of the brake-slide relationship when presented with data not used in network training. This tool will easily find brake-slide equation of each experience and we will compare the brake on two different experimental tests. This article demonstrates that the MOT brake testing do not check the car brake in its usual way of driving. We will provide data and graphs to prove that tyre pressure is a determining factor when assessing the condition of brakes.

Development of a methodology for forecasting electricity-price series to improve Demand Response Initiatives

The main objective of electricity regulators when establishing electricity markets is to decrease the cost of electricity through competition. However, this scenario can not be achieved without the full participation of the electricity demand. The aim of this paper is to propose a procedure, through the detection of electricity-price patterns based on what happened in the energy markets the previous day, which could help customers and aggregators to take decisions in Electricity Markets. In this way, the capacity of a methodology (Statistical Wards Linkage) to classify and forecast high electricity market prices is analyzed. Besides, some price-patterns were found in the abovementioned clusters. The knowledge contained within these patterns supplies customers with market-based information on which to focus its energy use decision to improve the usefulness of Demand Response Initiatives.