Hrvoje Keko

Wind Power Trading Under Uncertainty in LMP Markets

This paper presents a new model for optimal trading of wind power in day-ahead (DA) electricity markets under uncertainty in wind power and prices. The model considers settlement mechanisms in markets with locational marginal prices (LMPs), where wind power is not necessarily penalized from deviations between DA schedule and real-time (RT) dispatch. We use kernel density estimation to produce a probabilistic wind power forecast, whereas uncertainties in DA and RT prices are assumed to be Gaussian. Utility theory and conditional value at risk (CVAR) are used to represent the risk preferences of the wind power producers. The model is tested on real-world data from a large-scale wind farm in the United States. Optimal DA bids are derived under different assumptions for risk preferences and deviation penalty schemes. The results show that in the absence of a deviation penalty, the optimal bidding strategy is largely driven by price expectations. A deviation penalty brings the bid closer to the expected wind power forecast. Furthermore, the results illustrate that the proposed model can effectively control the trade-off between risk and return for wind power producers operating in volatile electricity markets.

Demand Dispatch and Probabilistic Wind Power Forecasting in Unit Commitment and Economic Dispatch: A Case Study of Illinois

In this paper, we analyze how demand dispatch combined with the use of probabilistic wind power forecasting can help accommodate large shares of wind power in electricity market operations. We model the operation of day-ahead and real-time electricity markets, which the system operator clears by centralized unit commitment and economic dispatch. We use probabilistic wind power forecasting to estimate dynamic operating reserve requirements, based on the level of uncertainty in the forecast. At the same time, we represent price responsive demand as a dispatchable resource, which adds flexibility in the system operation. In a case study of the power system in Illinois, we find that both demand dispatch and probabilistic wind power forecasting can contribute to efficient operation of electricity markets with large shares of wind power.

Risk management and optimal bidding for a wind power producer

This paper discusses risk management, contracting, and bidding for a wind power producer. A majority of the wind power in the United States is sold on long-term power purchase agreements, which hedge the wind power producer against future price risks. However, a significant amount is sold as merchant power and therefore is exposed to fluctuations in future electricity prices (day-ahead and real-time) and potential imbalance penalties. Wind power forecasting can serve as a tool to increase the profit and reduce the risk from participating in the wholesale electricity market. We propose a methodology to derive optimal day-ahead bids for a wind power producer under uncertainty in realized wind power and market prices. We also present an initial illustrative case study from a hypothetical wind site in the United States, where we compare the results of different day-ahead bidding strategies. The results show that the optimal day-ahead bid is highly dependent on the expected day-ahead and real-time prices, and also on the risk preferences of the wind power producer. A deviation penalty between day-ahead bid and real-time delivery tends to drive the bids closer to the expected generation for the next day.

EPSO: Evolutionary Particle Swarms

This chapter presents EPSO (Evolutionary Particle Swarm Optimization), as an evolutionary meta-heuristic that implements a scheme of self-adaptive recombination, borrowing the movement rule from PSO (Particle Swarm Optimization). Besides the basic model, it discusses a Stochastic Star topology for the communication among particles and presents a variant called differential EPSO or dEPSO. The chapter presents results in a didactic Unit Commitment/Generator Scheduling Power System problem and results of a competition among algorithms in an intelligent agent platform for Energy Re-tail Market simulation where EPSO comes out as the winner algorithm.

Reconstructing Missing Data in State Estimation With Autoencoders

This paper presents the proof of concept for a new solution to the problem of recomposing missing information at the SCADA of energy/distribution management systems (EMS/DMS), through the use of offline trained autoencoders. These are neural networks with a special architecture, which allows them to store knowledge about a system in a nonlinear manifold characterized by their weights. Suitable algorithms may then recompose missing inputs (measurements). The paper shows that, trained with adequate information, autoencoders perform well in recomposing missing voltage and power values, and focuses on the particularly important application of inferring the topology of the network when information about switch status is absent. Examples with the IEEE RTS 24-bus network are presented to illustrate the concept and technique.

Unit commitment and operating reserves with probabilistic wind power forecasts

In this paper we discuss how probabilistic wind power forecasts can serve as an important tool to efficiently address wind power uncertainty in power system operations. We compare different probabilistic forecasting and scenario reduction methods, and test the resulting forecasts on a stochastic unit commitment model. The results are compared to deterministic unit commitment, where dynamic operating reserve requirements can also be derived from the probabilistic forecasts. In both cases, the use of probabilistic forecasts contributes to improve the system performance in terms of cost and reliability.

Finding representative wind power scenarios and their probabilities for stochastic models

This paper analyzes the application of clustering techniques for wind power scenario reduction. The results have shown the unimodal structure of the scenario generated under a Monte Carlo process. The unimodal structure has been confirmed by the modes found by the information theoretic learning mean shift algorithm. The paper also presents a new technique able to represent the wind power forecasting uncertainty by a set of representative scenarios capable of characterizing the probability density function of the wind power forecast. From an initial large set of sampled scenarios, a reduced discrete set of representative scenarios associated with a probability of occurrence can be created finding the areas of high probability density. This will allow the reduction of the computational burden in stochastic models that require scenario representation.

A Multiple Scenario Security Constrained Reactive Power Planning Tool Using EPSO

Evolutionary particle swarm optimization (EPSO) is a robust optimization algorithm belonging to evolutionary methods. EPSO borrows the movement rules from particle swarm optimization (PSO) and uses it as a recombination operator that evolves under selection. This paper presents a reactive power planning approach taking advantage of EPSO robustness, in a model that considers simultaneously multiple contingencies and multiple load levels. Results for selected problems are summarized including a trade-off analysis of results.

Wind power forecasting, unit commitment, and electricity market operations

In this paper we discuss the use of wind power forecasting in electricity market operations. In particular, we demonstrate how probabilistic forecasts can contribute to address the uncertainty and variability in wind power. We focus on efficient use of forecasts in the unit commitment problem and discuss potential implications for electricity market operations.

Development and testing of improved statistical wind power forecasting methods.

Wind power forecasting (WPF) provides important inputs to power system operators and electricity market participants. It is therefore not surprising that WPF has attracted increasing interest within the electric power industry. In this report, we document our research on improving statistical WPF algorithms for point, uncertainty, and ramp forecasting. Below, we provide a brief introduction to the research presented in the following chapters. For a detailed overview of the state-of-the-art in wind power forecasting, we refer to [1]. Our related work on the application of WPF in operational decisions is documented in [2]. Point forecasts of wind power are highly dependent on the training criteria used in the statistical algorithms that are used to convert weather forecasts and observational data to a power forecast. In Chapter 2, we explore the application of information theoretic learning (ITL) as opposed to the classical minimum square error (MSE) criterion for point forecasting. In contrast to the MSE criterion, ITL criteria do not assume a Gaussian distribution of the forecasting errors. We investigate to what extent ITL criteria yield better results. In addition, we analyze time-adaptive training algorithms and how they enable WPF algorithms to cope with non-stationary data and, thus, to adapt to new situations without requiring additional offline training of the model. We test the new point forecasting algorithms on two wind farms located in the U.S. Midwest. Although there have been advancements in deterministic WPF, a single-valued forecast cannot provide information on the dispersion of observations around the predicted value. We argue that it is essential to generate, together with (or as an alternative to) point forecasts, a representation of the wind power uncertainty. Wind power uncertainty representation can take the form of probabilistic forecasts (e.g., probability density function, quantiles), risk indices (e.g., prediction risk index) or scenarios (with spatial and/or temporal dependence). Statistical approaches to uncertainty forecasting basically consist of estimating the uncertainty based on observed forecasting errors. Quantile regression (QR) is currently a commonly used approach in uncertainty forecasting. In Chapter 3, we propose new statistical approaches to the uncertainty estimation problem by employing kernel density forecast (KDF) methods. We use two estimators in both offline and time-adaptive modes, namely, the Nadaraya-Watson (NW) and Quantilecopula (QC) estimators. We conduct detailed tests of the new approaches using QR as a benchmark. One of the major issues in wind power generation are sudden and large changes of wind power output over a short period of time, namely ramping events. In Chapter 4, we perform a comparative study of existing definitions and methodologies for ramp forecasting. We also introduce a new probabilistic method for ramp event detection. The method starts with a stochastic algorithm that generates wind power scenarios, which are passed through a high-pass filter for ramp detection and estimation of the likelihood of ramp events to happen. The report is organized as follows: Chapter 2 presents the results of the application of ITL training criteria to deterministic WPF; Chapter 3 reports the study on probabilistic WPF, including new contributions to wind power uncertainty forecasting; Chapter 4 presents a new method to predict and visualize ramp events, comparing it with state-of-the-art methodologies; Chapter 5 briefly summarizes the main findings and contributions of this report.