Pasquale De Falco

A Probabilistic Competitive Ensemble Method for Short-Term Photovoltaic Power Forecasting

Photovoltaic systems are expected to play a key role in the planning and operation of future distribution systems due to the benefits associated with their use. Unfortunately, a great problem is involved in photovoltaic power utilization, i.e., the unpredictability of the solar source. Thus, many forecasting methods have been developed in order to provide tools with adequate consistency, quality, and value. The methods can provide either deterministic or probabilistic forecasts; the latter seem to be the most appropriate for taking into account the unavoidable uncertainties of the solar source. In this paper, a new probabilistic method based on a competitive ensemble of different base predictors is proposed for the short-term forecasting of photovoltaic power. Three probabilistic methods were selected and trained as base predictors in order to obtain an ensemble of the predictive distribution with optimal characteristics of sharpness and reliability. Numerical applications based on actual data were performed to test the effectiveness of the proposed method with respect to single predictors and to a benchmark method.

A Bayesian-based approach for the short-term forecasting of electrical loads in smart grids.: Part I: theoretical aspects

Accurate short-term load forecasting is extremely important for the optimal management of smart grids allowing the increase of energy efficiency and enabling profitable demand response strategies. However, most of industrial and domestic loads are intrinsically affected by uncertainties due to many factors such as devices operational characteristics, time of use, weather conditions and other random effects. Therefore, probabilistic load forecasting is a challenging task and in relevant literature a great interest has been recently developed towards this topic. In this paper, the uncertainties related to the load demand are modeled through different probability density functions and a probabilistic method based on the Bayesian inference and stochastic time series is proposed for the short-term forecasting of the probability density functions parameters. This paper shows the theoretical aspects of the proposed method and it is a companion paper to a paper of the same title, Part II, in which the numerical applications are reported.

New advanced method and cost-based indices applied to probabilistic forecasting of photovoltaic generation

The ability to forecast the production of power by photovoltaic(PV) systems accurately and reliably is of major importance for the appropriate management of future electrical distribution systems. Several forecasting methods have been proposed in the relevant literature, and many indices have been used to quantify the quality of the forecasts. The methods can provide either deterministic or probabilistic forecasts; the latter seem to be the most appropriate to take into account the unavoidable uncertainties of PV power production. Similarly, indices were used to quantify the quality of both deterministic and probabilistic forecasting methods, but they usually do not account for the economic consequences of forecasting errors. In this paper, two advanced probabilistic forecasting approaches based on the Bayesian inference method are applied to the short-term forecasting of PV power production. Moreover, new probabilistic indices were proposed with the aim of comparing the probabilistic forecasting methods in such way that the value of the forecast is not included only by the users in their decision-making process; instead, it is partially anticipated by the forecasters in their quality-assessment process. Numerical applications also are presented to provide evidence of the performances of the Bayesian-based approaches and the probabilistic indices that were considered.

A Bayesian-based approach for the short-term forecasting of electrical loads in smart grids.: Part II: numerical applications

This paper deals with the numerical applications performed on real data to validate the short term active load power forecasting method proposed in the companion paper of the same title, Part I. The method proposed in Part I is a hybrid method based on the application of time series model and Bayesian inference on a dataset of load power measurements; it provides probabilistic forecasts for different time horizons and is particularly suitable for risk assessment associated to the management of smart grids. The numerical applications presented in this paper are performed using domestic load measurements taken from an Italian distribution system. The method was tested for different number of loads and time horizons of forecasts; in particular, 15-minutes and 24- to 48- hours lead times were considered. The performances of the proposed method are also evaluated by testing its reliability, sharpness and calibration.

Bayesian estimation of inverse burr stress-strength model for power system components reliability assessment

A novel Inverse Burr stress-strength probabilistic model is proposed for reliability assessment of electrical components in the presence of overstresses, such as extreme wind-speed values for wind towers reliability modeling or extreme voltage surge amplitudes for insulation reliability modeling. These are kinds of stresses which can be modeled by means of adequate, well-known extreme value distributions, such as Gumbel and Inverse Weibull distributions. However, Inverse Burr models have also been recently proven to be efficient and flexible extreme values models. In this paper, also the “strength” of electrical components is characterized through an Inverse Burr distribution. The estimation of the overall reliability of the component is performed through classical Maximum Likelihood Estimation procedure and through a new proposed Bayesian methodology, based upon the assessment of prior distributions on given parameters of the stress and strength distributions. Indeed, usually only a limited amount of lifetime data are available for high-reliability electrical components, while data on strengths or stresses are easier to get. Therefore, the application of a Bayesian approach is particularly appropriate in situations characterized, on one hand, by lack of experimental data, and, on the other hand, by some degree of technical knowledge. The validation of the stress-strength model and the comparison between both reliability estimation methods is confirmed through a numerical application, considering typical values of reliability of electrical components. The results proved the usefulness of the Bayesian reliability estimation procedure and the feasibility of the Inverse Burr stress-strength model.

A New Ensemble Probabilistic Method for Short-Term Photovoltaic Power Forecasting

The high penetration of photovoltaic (PV) systems led to their growing impact on the planning and operation of actual distribution systems. However, the uncertainties due to the intermittent nature of solar energy complicate these tasks. Therefore, high-quality methods for forecasting the PV power are now essential, and many tools have been developed in order to provide useful and consistent forecasts. This chapter deals with probabilistic forecasting methods of PV system power, since they have recently drawn the attention of researchers as appropriate tools to cope with the unavoidable uncertainties of solar source. A new multi-model probabilistic ensemble is proposed; it properly combines a Bayesian-based and a quantile regression-based probabilistic method as individual predictors. Numerical applications based on actual irradiance data give evidence of the probabilistic performances of the proposed method in terms of both sharpness and calibration.