Minas C. Alexiadis

A fuzzy model for wind speed prediction and power generation in wind parks using spatial correlation

In this paper, a fuzzy model is suggested for the prediction of wind speed and the produced electrical power at a wind park. The model is trained using a genetic algorithm-based learning scheme. The training set includes wind speed and direction data, measured at neighboring sites up to 30 km away from the wind turbine clusters. Extensive simulation results are shown for two application cases, providing wind speed forecasts from 30 min to 2 h ahead. It is demonstrated that the suggested model achieves an adequate understanding of the problem while it exhibits significant improvement compared to the persistent method.

Long-term wind speed and power forecasting using local recurrent neural network models

This paper deals with the problem of long-term wind speed and power forecasting based on meteorological information. Hourly forecasts up to 72-h ahead are produced for a wind park on the Greek island of Crete. As inputs our models use the numerical forecasts of wind speed and direction provided by atmospheric modeling system SKIRON for four nearby positions up to 30 km away from the wind turbine cluster. Three types of local recurrent neural networks are employed as forecasting models, namely, the infinite impulse response multilayer perceptron (IIR-MLP), the local activation feedback multilayer network (LAF-MLN), and the diagonal recurrent neural network (RNN). These networks contain internal feedback paths, with the neuron connections implemented by means of IIR synaptic filters. Two novel and optimal on-line learning schemes are suggested for the update of the recurrent networks weights based on the recursive prediction error algorithm. The methods assure continuous stability of the network during the learning phase and exhibit improved performance compared to the conventional dynamic back propagation. Extensive experimentation is carried out where the three recurrent networks are additionally compared to two static models, a finite-impulse response NN (FIR-NN) and a conventional static-MLP network. Simulation results demonstrate that the recurrent models, trained by the suggested methods, outperform the static ones while they exhibit significant improvement over the persistent method.

A neural network short term load forecasting model for the Greek power system

This paper presents the development of an artificial neural network (ANN) based short-term load forecasting model for the Energy Control Center of the Greek Public Power Corporation (PPC). The model can forecast daily load profiles with a lead time of one to seven days. Attention was paid for the accurate modeling of holidays. Experiences gained during the development of the model regarding the selection of the input variables, the ANN structure, and the training data set are described in the paper. The results indicate that the load forecasting model developed provides accurate forecasts.

Wind speed and power forecasting based on spatial correlation models

Wind energy conversion systems (WECS) cannot be dispatched like conventional generators. This can pose problems for power system schedulers and dispatchers, especially if the schedule of wind power availability is not known in advance. However, if the wind speed can be reliably forecasted up to several hours ahead, the generating schedule can efficiently accommodate the wind generation. This paper illustrates a technique for forecasting wind speed and power output up to several hours ahead, based on cross correlation at neighboring sites. The authors develop an artificial neural network (ANN) that significantly improves forecasting accuracy comparing to the persistence forecasting model. The method is tested at different sites over a year.

Modeling and Field Measurements of Photovoltaic Units Connected to LV Grid. Study of Penetration Scenarios

The operation of photovoltaic (PV) units connected to the grid is characterized by several uncertainties due to the number of currently operating units, the points where these units are sited, the exported power, and the injection of harmonic currents. The objective of this paper is to investigate the impact of the penetration of PV units in the low-voltage (LV) network. Thus, a model has been developed for the computation of load flows, harmonics, and voltages in the feeders. A PV panel is modeled as an irradiance-driven current generator, with an embodied maximum power point (MPP) tracking algorithm; it is connected to the network via electronic switches that represent the PV inverter. Simulations are performed with PSIM and Harmoniques simulator packages. Results from the simulated model are validated with data acquired from field measurements and bibliography. Finally, the acceptable penetration level of PV systems is investigated for several scenarios, depending on the topology of the LV network, the number, size, and locations of loads and PV units.

Harmonic impact of small photovoltaic systems connected to the LV distribution network

Since the penetration of photovoltaic (PV) systems in the low voltage (LV) distribution network is increasing, the need to register and model the contribution of these systems to the harmonic distortion of current and voltage waveforms is becoming an up-to-date issue. As PV systems incorporate power conditioning units, which are harmonic generating devices, they will have an influence on quality of supply, reliable operation of system equipment as well as component life expectancy. This paper investigates the harmonic impact of a 20 kWp PV system connected to the LV distribution network in Greece. The harmonic behavior of the PV plant as a function of the solar radiation under several weather conditions is analyzed. Measurements results are compared to those obtained from the power simulator package PSIMcopy. The level of penetration of PV systems in the LV distribution network without harmonic limits been exceeded is investigated.

Load profiling in the deregulated electricity markets: A review of the applications

This paper provides a state of the art survey on the load profiling applications in the deregulated market. The survey is focused on topics like tariff design, load forecasting and various power distribution issues. The procedure of the formulation of the load profiles is analyzed, as well as the algorithms used for the aforementioned procedure. Furthermore, the paper presents the two general models that are used in load profiling, namely the area (or regional) and the category (or the consumer-group-related) model. The strength and weaknesses of each model are discussed. Finally, the authors contribute to the existing load profiling literature by introducing a clustering algorithm that is used in other clustering applications. The algorithms performance is compared with a set of algorithms that have been proposed and the results are discussed.

Application of time series and artificial neural network models in short-term forecasting of PV power generation

This paper addresses two practical methods for electricity generation forecasting of grid-connected PV plants. The first model is based on seasonal ARIMA time-series analysis and is further improved by incorporating short-term solar radiation forecasts derived from NWP models. The second model adopts artificial neural networks with multiple inputs. Day-ahead and rolling intra-day forecast updates are implemented to evaluate the forecasting errors. All models are compared in terms of the Normalized (with respect to the PV installed capacity) Root Mean Square Error (NRMSE). Simulation results from the application of the forecasting models in different PV plants of the Greek power system are presented.

Evaluation of the performance of clustering algorithms for a high voltage industrial consumer

Abstract Load profiling refers to a procedure which leads to the formulation of daily load curve clusters based on the similarity of the curves shapes. This paper focuses on the investigation of the consumption patterns of an existing high voltage industrial consumer. The profiling process involves stages like the normalization of the recorded load data, the utilization of pattern recognition algorithms, the selection of the appropriate validation scheme and the exploitation of the profiling findings. Certain improvements are proposed for each of these stages. More specifically, the most common algorithms of the related literature are implemented and a detailed investigation of their performance is presented. A new algorithm is proposed, presenting, in the majority of the cases, the best performance. Additionally, all the clustering validity indicators of the literature are considered to evaluate the clustering results. After the formulation of the load curve clusters, the load profiles are extracted and based on specific indices conclusions are drawn regarding the implementation of suitable demand side management schemes.

Electricity customer characterization based on different representative load curves

Load profiling provides the necessary information about daily demand patterns for the short and medium-term actions οf retailers and utilities. Consumer characterization is a two stage approach: In the first stage, the daily load curves of each consumer are classified in a certain number of clusters. Each cluster constitutes a load profile. In the second stage, one of these profiles is chosen as representative for the consumer and a new classification takes places between the load profiles of each customer, leading to the formulation of customer classes. This paper examines various approaches for the formulation of the consumer classes during the first stage. A specific profile is chosen and the second stage procedure takes place. A criterion based on cost of purchased electricity is introduced in order to evaluate the results of the clustering of the second stage.