arXiv subjects

Salvador Pineda

Publications and source records attributed to Salvador Pineda.

At least 37 records · Page 2Linked to original sources

arXiv2022

Warm-starting constraint generation for mixed-integer optimization: A Machine Learning approach

Mixed Integer Linear Programs (MILP) are well known to be NP-hard (Non-deterministic Polynomial-time hard) problems in general. Even though pure optimization-based methods, such as constraint generation, are guaranteed to provide an optimal solution if enough time is given, their use in online applications remains a great challenge due to their usual excessive time requirements. To alleviate their computational burden, some machine learning techniques (ML) have been proposed in the literature, using the information provided by previously solved MILP instances. Unfortunately, these techniques report a non-negligible percentage of infeasible or suboptimal instances. By linking mathematical optimization and machine learning, this paper proposes a novel approach that speeds up the traditional constraint generation method, preserving feasibility and optimality guarantees. In particular, we first identify offline the so-called invariant constraint set of past MILP instances. We then train (also offline) a machine learning method to learn an invariant constraint set as a function of the problem parameters of each instance. Next, we predict online an invariant constraint set of the new unseen MILP application and use it to initialize the constraint generation method. This warm-started strategy significantly reduces the number of iterations to reach optimality, and therefore, the computational burden to solve online each MILP problem is significantly reduced. Very importantly, all the feasibility and optimality theoretical guarantees of the traditional constraint generation method are inherited by our proposed methodology. The computational performance of the proposed approach is quantified through synthetic and real-life MILP applications.

math.OC↗

arXiv2022

Cost-driven Screening of Network Constraints for the Unit Commitment Problem

In an attempt to speed up the solution of the unit commitment (UC) problem, both machine-learning and optimization-based methods have been proposed to lighten the full UC formulation by removing as many superfluous line-flow constraints as possible. While the elimination strategies based on machine learning are fast and typically delete more constraints, they may be over-optimistic and result in infeasible UC solutions. For their part, optimization-based methods seek to identify redundant constraints in the full UC formulation by exploring the feasibility region of an LP-relaxation. In doing so, these methods only get rid of line-flow constraints whose removal leaves the feasibility region of the original UC problem unchanged. In this paper, we propose a procedure to substantially increase the line-flow constraints that are filtered out by optimization-based methods without jeopardizing their appealing ability of preserving feasibility. Our approach is based on tightening the LP-relaxation that the optimization-based method uses with a valid inequality related to the objective function of the UC problem and hence, of an economic nature. The result is that the so strengthened optimization-based method identifies not only redundant line-flow constraints but also inactive ones, thus leading to more reduced UC formulations.

math.OC↗

arXiv2021

A bilevel framework for decision-making under uncertainty with contextual information

In this paper, we propose a novel approach for data-driven decision-making under uncertainty in the presence of contextual information. Given a finite collection of observations of the uncertain parameters and potential explanatory variables (i.e., the contextual information), our approach fits a parametric model to those data that is specifically tailored to maximizing the decision value, while accounting for possible feasibility constraints. From a mathematical point of view, our framework translates into a bilevel program, for which we provide both a fast regularization procedure and a big-M-based reformulation that can be solved using off-the-shelf optimization solvers. We showcase the benefits of moving from the traditional scheme for model estimation (based on statistical quality metrics) to decision-guided prediction using three different practical problems. We also compare our approach with existing ones in a realistic case study that considers a strategic power producer that participates in the Iberian electricity market. Finally, we use these numerical simulations to analyze the conditions (in terms of the firm's cost structure and production capacity) under which our approach proves to be more advantageous to the producer.

math.OC↗

arXiv2021

Learning the price response of active distribution networks for TSO-DSO coordination

The increase in distributed energy resources and flexible electricity consumers has turned TSO-DSO coordination strategies into a challenging problem. Existing decomposition/decentralized methods apply divide-and-conquer strategies to trim down the computational burden of this complex problem, but rely on access to proprietary information or fail-safe real-time communication infrastructures. To overcome these drawbacks, we propose in this paper a TSO-DSO coordination strategy that only needs a series of observations of the nodal price and the power intake at the substations connecting the transmission and distribution networks. Using this information, we learn the price response of active distribution networks (DN) using a decreasing step-wise function that can also adapt to some contextual information. The learning task can be carried out in a computationally efficient manner and the curve it produces can be interpreted as a market bid, thus averting the need to revise the current operational procedures for the transmission network. Inaccuracies derived from the learning task may lead to suboptimal decisions. However, results from a realistic case study show that the proposed methodology yields operating decisions very close to those obtained by a fully centralized coordination of transmission and distribution.

math.OC↗

arXiv2021

An exact dynamic programming approach to segmented isotonic regression

This paper proposes a polynomial-time algorithm to construct the monotone stepwise curve that minimizes the sum of squared errors with respect to a given cloud of data points. The fitted curve is also constrained on the maximum number of steps it can be composed of and on the minimum step length. Our algorithm relies on dynamic programming and is built on the basis that said curve-fitting task can be tackled as a shortest-path type of problem. Numerical results on synthetic and realistic data sets reveal that our algorithm is able to provide the globally optimal monotone stepwise curve fit for samples with thousands of data points in less than a few hours. Furthermore, the algorithm gives a certificate on the optimality gap of any incumbent solution it generates. From a practical standpoint, this piece of research is motivated by the roll-out of smart grids and the increasing role played by the small flexible consumption of electricity in the large-scale integration of renewable energy sources into current power systems. Within this context, our algorithm constitutes an useful tool to generate bidding curves for a pool of small flexible consumers to partake in wholesale electricity markets.

math.OC↗

arXiv2021

Forecasting the Price-Response of a Pool of Buildings via Homothetic Inverse Optimization

This paper focuses on the day-ahead forecasting of the aggregate power of a pool of smart buildings equipped with thermostatically-controlled loads. We first propose the modeling of the aggregate behavior of its power trajectory by using a geometric approach. Specifically, we assume that the aggregate power is a homothet of a prototype building, whose physical and technical parameters are chosen to be the mean of those in the pool. This allows us to preserve the building thermal dynamics of the pool. We then apply inverse optimization to estimate the homothetic parameters with bilevel programming. The lower level characterizes the price-response of the ensemble by a set of marginal utility curves and a homothet of the prototype building, which, in turn, are inferred in the upper-level problem. The upper level minimizes the mean absolute error over a training sample. This bilevel program is transformed into a regularized nonlinear problem that is initialized with the solution given by an efficient heuristic procedure. This heuristic consists in solving two linear programs and its solution is deemed a suitable proxy for the original bilevel problem. The results have been compared to state-of-the-art methodologies.

eess.SY↗

arXiv2021

Inverse Optimization with Kernel Regression: Application to the Power Forecasting and Bidding of a Fleet of Electric Vehicles

This paper considers an aggregator of Electric Vehicles (EVs) who aims to learn the aggregate power of his/her fleet while also participating in the electricity market. The proposed approach is based on a data-driven inverse optimization (IO) method, which is highly nonlinear. To overcome such a caveat, we use a two-step estimation procedure which requires solving two convex programs. Both programs depend on penalty parameters that can be adjusted by using grid search. In addition, we propose the use of kernel regression to account for the nonlinear relationship between the behaviour of the pool of EVs and the explanatory variables, i.e., the past electricity prices and EV fleet's driving patterns. Unlike any other forecasting method, the proposed IO framework also allows the aggregator to derive a bid/offer curve, i.e. the tuple of price-quantity to be submitted to the electricity market, according to the market rules. We show the benefits of the proposed method against the machine-learning techniques that are reported to exhibit the best forecasting performance for this application in the technical literature.

eess.SY↗

arXiv2021

A novel embedded min-max approach for feature selection in nonlinear support vector machine classification

In recent years, feature selection has become a challenging problem in several machine learning fields, such as classification problems. Support Vector Machine (SVM) is a well-known technique applied in classification tasks. Various methodologies have been proposed in the literature to select the most relevant features in SVM. Unfortunately, all of them either deal with the feature selection problem in the linear classification setting or propose ad-hoc approaches that are difficult to implement in practice. In contrast, we propose an embedded feature selection method based on a min-max optimization problem, where a trade-off between model complexity and classification accuracy is sought. By leveraging duality theory, we equivalently reformulate the min-max problem and solve it without further ado using off-the-shelf software for nonlinear optimization. The efficiency and usefulness of our approach are tested on several benchmark data sets in terms of accuracy, number of selected features and interpretability.

cs.LG↗

arXiv2020

Day-ahead Operation of an Aggregator of Electric Vehicles via Optimization under Uncertainty

We pose the aggregator's problem as a bilevel model, where the upper level minimizes the total operation costs of the fleet of EVs, while each lower level minimizes the energy available to each vehicle for transportation given a certain charging plan. Thanks to the totally unimodular character of the constraint matrix in the lower-level problems, the model can be mathematically recast as a computationally efficient mixed-integer program that delivers charging schedules that are robust against the uncertain availability of the EVs. Finally, we use synthetic data from the National Household Travel Survey 2017 to analyze the behavior of the EV aggregator from both economic and technical viewpoints and compare it with the results from a deterministic approach.

eess.SY↗

arXiv2020

An Efficient Robust Approach to the Day-ahead Operation of an Aggregator of Electric Vehicles

The growing use of electric vehicles (EVs) may hinder their integration into the electricity system as well as their efficient operation due to the intrinsic stochasticity associated with their driving patterns. In this work, we assume a profit-maximizer EV-aggregator who participates in the day-ahead electricity market. The aggregator accounts for the technical aspects of each individual EV and the uncertainty in its driving patterns. We propose a hierarchical optimization approach to represent the decision-making of this aggregator. The upper level models the profit-maximizer aggregator's decisions on the EV-fleet operation, while a series of lower-level problems computes the worst-case EV availability profiles in terms of battery draining and energy exchange with the market. Then, this problem can be equivalently transformed into a mixed-integer linear single-level equivalent given the totally unimodular character of the constraint matrices of the lower-level problems and their convexity. Finally, we thoroughly analyze the benefits of the hierarchical model compared to the results from stochastic and deterministic models.

math.OC↗

arXiv2020

Data-Driven Screening of Network Constraints for Unit Commitment

The transmission-constrained unit commitment (TC-UC) problem is one of the most relevant problems solved by independent system operators for the daily operation of power systems. Given its computational complexity, this problem is usually not solved to global optimality for real-size power systems. In this paper, we propose a data-driven method that leverages historical information to screen out network constraints in the TC-UC problem. First, past data on demand and renewable generation throughout the network are used to learn the congestion status of transmission lines. Then, we infer the lines that will not become congested for upcoming operating conditions based on such learning and disregard their capacity constraints. This way, we formulate a reduced TC-UC problem that is easier to solve. Numerical results on a medium- and a large-size power system show that the proposed approach outperforms existing ones by significantly reducing the computational time while obtaining solutions that are equal or close to the one obtained with the original TC-UC problem. Furthermore, the purely data-driven method we propose can be seamlessly complemented with a constraint generation procedure to guarantee that the optimal solution to the original TC-UC problem is eventually recovered.

math.OC↗

arXiv2020

Feature-driven Improvement of Renewable Energy Forecasting and Trading

Inspired from recent insights into the common ground of machine learning, optimization and decision-making, this paper proposes an easy-to-implement, but effective procedure to enhance both the quality of renewable energy forecasts and the competitive edge of renewable energy producers in electricity markets with a dual-price settlement of imbalances. The quality and economic gains brought by the proposed procedure essentially stem from the utilization of valuable predictors (also known as features) in a data-driven newsvendor model that renders a computationally inexpensive linear program. We illustrate the proposed procedure and numerically assess its benefits on a realistic case study that considers the aggregate wind power production in the Danish DK1 bidding zone as the variable to be predicted and traded. Within this context, our procedure leverages, among others, spatial information in the form of wind power forecasts issued by transmission system operators (TSO) in surrounding bidding zones and publicly available in online platforms. We show that our method is able to improve the quality of the wind power forecast issued by the Danish TSO by several percentage points (when measured in terms of the mean absolute or the root mean square error) and to significantly reduce the balancing costs incurred by the wind power producer.

math.OC↗

arXiv2019

How Can Smart Buildings Be Price-Responsive?

The prospective participation of smart buildings in the electricity system is strongly related to the increasing active role of demand-side resources in the electrical grid. In addition, the growing penetration of smart meters and recent advances on home automation technologies will spur the development of new mathematical tools to help optimize the local resources of these buildings. Within this context, this paper first provides a comprehensive model to determine the electrical consumption of a single-zone household based on economic model predictive control. The goal of this problem is to minimize the electricity consumption cost while accounting for the heating dynamics of the building, smart home appliances, and comfort constraints. This paper then identifies and analyzes the key parameters responsible for the price-responsive behaviour of smart households.

eess.SY↗

arXiv2019

Time-Adaptive Unit Commitment

The short-term operation of a power system is usually planned by solving a day-ahead unit commitment problem. Due to historical reasons, the commitment of the power generating units is decided over a time horizon typically consisting of the 24 hourly periods of a day. In this paper, we show that, as a result of the increasing penetration of intermittent renewable generation, this somewhat arbitrary and artificial division of time may prove to be significantly suboptimal and counterproductive. Instead, we propose a time-adaptive day-ahead unit commitment formulation that better captures the net-demand variability throughout the day. The proposed formulation provides the commitment and dispatch of thermal generating units over a set of 24 time periods too, but with different duration. To do that, we use a clustering procedure to select the duration of those adaptive time periods taking into account the renewable generation and demand forecasts. Numerical results show that, without increasing the computational burden, the proposed time-adaptive unit commitment allows for a more efficient use of the system flexibility, which translates into a lower operating cost and a higher penetration of renewable production than those achieved by a conventional hourly unit commitment problem.

math.OC↗

arXiv2018

Solving Linear Bilevel Problems Using Big-Ms: Not All That Glitters Is Gold

The most common procedure to solve a linear bilevel problem in the PES community is, by far, to transform it into an equivalent single-level problem by replacing the lower level with its KKT optimality conditions. Then, the complementarity conditions are reformulated using additional binary variables and large enough constants (big-Ms) to cast the single-level problem as a mixed-integer linear program that can be solved using optimization software. In most cases, such large constants are tuned by trial and error. We show, through a counterexample, that this widely used trial-and-error approach may lead to highly suboptimal solutions. Then, further research is required to properly select big-M values to solve linear bilevel problems.

math.OC↗

arXiv2016

On the Inefficiency of the Merit Order in Forward Electricity Markets with Uncertain Supply

This paper provides insight on the economic inefficiency of the classical merit-order dispatch in electricity markets with uncertain supply. For this, we consider a power system whose operation is driven by a two-stage electricity market, with a forward and a real-time market. We analyze two different clearing mechanisms: a conventional one, whereby the forward and the balancing markets are independently cleared following a merit order, and a stochastic one, whereby both market stages are co-optimized with a view to minimizing the expected aggregate system operating cost. We first derive analytical formulae to determine the dispatch rule prompted by the co-optimized two-stage market for a stylized power system with flexible, inflexible and stochastic power generation and infinite transmission capacity. This exercise sheds light on the conditions for the stochastic market-clearing mechanism to break the merit order. We then introduce and characterize two enhanced variants of the conventional two-stage market that result in either price-consistent or cost-efficient merit-order dispatch solutions, respectively. The first of these variants corresponds to a conventional two-stage market that allows for virtual bidding, while the second requires that the stochastic power production be centrally dispatched. Finally, we discuss the practical implications of our analytical results and illustrate our conclusions through examples.

math.OC↗

arXiv2014

Impact of Equipment Failures and Wind Correlation on Generation Expansion Planning

Generation expansion planning has become a complex problem within a deregulated electricity market environment due to all the uncertainties affecting the profitability of a given investment. Current expansion models usually overlook some of these uncertainties in order to reduce the computational burden. In this paper, we raise a flag on the importance of both equipment failures (units and lines) and wind power correlation on generation expansion decisions. For this purpose, we use a bilevel stochastic optimization problem, which models the sequential and noncooperative game between the generating company (GENCO) and the system operator. The upper-level problem maximizes the GENCO's expected profit, while the lower-level problem simulates an hourly market-clearing procedure, through which LMPs are determined. The uncertainty pertaining to failures and wind power correlation are characterized by a scenario set, and their impact on generation expansion decisions are quantified and discussed for a 24-bus power system.

math.OC↗

arXiv2014

Impact of Forecast Errors on Expansion Planning of Power Systems with a Renewables Target

This paper analyzes the impact of production forecast errors on the expansion planning of a power system and investigates the influence of market design to facilitate the integration of renewable generation. For this purpose, we propose a stochastic programming modeling framework to determine the expansion plan that minimizes system-wide investment and operating costs, while ensuring a given share of renewable generation in the electricity supply. Unlike existing ones, this framework includes both a day-ahead and a balancing market so as to capture the impact of both production forecasts and the associated prediction errors. Within this framework, we consider two paradigmatic market designs that essentially differ in whether the day-ahead generation schedule and the subsequent balancing re-dispatch are co-optimized or not. The main features and results of the model set-ups are discussed using an illustrative four-node example and a more realistic 24-node case study.

math.OC↗