A two-level model to define the energy procurement contract and daily operation schedule of microgrids

https://doi.org/10.1016/j.segan.2021.100459Get rights and content

Highlights

  • Integrating energy procurement contract, scheduling, demand response and optimal management of deferrable loads.

  • Framework to define optimal procurement contracts and operation schedule.

  • Flexible chance-constraints-based mathematical programming model.

  • New criterion to model chance-constraints based on historical min/max values.

  • Battery system and diesel generators reserves for overcoming forecast errors.

Abstract

A two-level energy management model that integrates energy procurement contract, operation scheduling under uncertainty, and demand response for connected microgrids is presented in this paper. It provides a framework for managing microgrid operations efficiently by prioritizing renewable over non-renewable generation. In this way, it is possible to size/measure both battery system and diesel generators reserves for overcoming forecast errors of both renewable generation and non-deferrable load demand; and this also enables deferrable load managing. A flexible chance-constraints-based mathematical programming model is also proposed, which can be used as an optimization model at both levels of the energy management model for defining an optimal energy procurement contract based on energy market trade policies and for generating an optimal daily operation schedule based on the energy contract. Furthermore, a new criterion for chance-constraints modeling based on historical minimum/maximum values is proposed. This modeling is intended to consider a high percentage of potential scenarios that may take place during execution. The percentage of presumed cases is estimated from simulation results.

Introduction

Microgrids are designed to operate at low or medium voltage, integrating renewable energy sources such as solar panels or wind turbines , controllable distributed generators such as fuel cells or micro turbines , and battery energy storage systems to meet local load demands [1]. They can operate in isolation or connected to the main grid through a macro station [2].

Renewable energy sources are highly dependent on weather conditions, which can also affect load demands. Generation and demand fluctuations and uncertainty could lead to service instability and energy deficit or excess at different times in a day. Therefore, even though microgrids can potentially reduce the need for an expansion of the traditional energy system, their operation management brings about new challenges.

Other types of renewable energy integration systems have emerged as ways to model new energy systems. Government policies related to renewable energy and climate variability are stimulating the transformation of energy users, who stop being consumers to become prosumers [3]. Prosumers are individuals who both consume and produce energy. They are emerging as a result of the development of household use of renewable energy sources. Since renewable energy sources of these prosumers as single entities consist of both solar systems and wind turbines at very small scale, they are often excluded from the wholesale energy market due to their perceived inefficiency and unreliability. Therefore, community energy systems such as virtual power plants, prosumer community groups, and integrated community energy systems are emerging in the literature to make their participation viable in the energy market [4].

Virtual power plants rely upon software systems to remotely and automatically dispatch and optimize generation or demand-side of storage resources in a single and secure Web-connected system. In short, virtual power plants represent an ‘Internet of energy’, which allows tapping into existing grid networks to tailor electricity supply and demand services for a customer, thus maximizing value for both end user and distribution utility through software innovations [5]. They use advanced communication technology and software architecture to aggregate distributed generations, energy storage devices, and adjustable loads. They can control and optimize renewable energy through a control coordination center, thus improving power quality and ensuring the effective utilization of renewable energy [6].

The term ‘integrated community energy’ refers to an integrated energy system based on the synergy and optimization of energy generation, transmission, conversion, distribution, storage, and consumption in a given region to meet various energy demands from different users within the region [7]. As a single entity, it has a suitable scale to participate in wholesale markets with the integration of distributed energy resources [8]. These systems aim to group prosumers sharing similar behaviors and interests to achieve the synergy for community self-provision and jointly compete in the energy market. Since network microgrid features are different from those of the prosumer community, the information and communication technologies required to manage both types of energy systems are not comparable, since their functional and non-functional requirements are significantly different.

The main concepts related to microgrids operation management this work focuses on are: unit commitment, economic dispatch, reserve sizing, demand response, and procurement contract. Unit commitment aims at scheduling the daily on/off-switch of controllable generators and the battery system charge/discharge mode to deal with energy deficit or excess. Economic dispatch is aimed at minimizing the operating cost of microgrid power dispatch by defining the hourly power that must be generated by each switched-on controllable generator, stored or delivered by the battery system, and supplied by the main grid [9], [10], [11].

The operation scheduling concept has been extensively used in the microgrid domain to refer to unit commitment and economic dispatch as a single decision problem. Moreover, when defining the operation schedule, properly sized reserves of controllable generators and battery system are defined with the aim of overcoming generation and demand uncertainties [12].

Demand response programs for the industrial sector refers to a set of measures aimed at encouraging a voluntary change in consumers’ electricity usage pattern in response to changes in electricity prices or grid reliability conditions. The increase in the demand-side capacity following the participation in such programs could effectively reduce electricity costs and enhance microgrid reliability [13], [14], [15]. Based on these programs, microgrid demands are classified into deferrable loads or non-deferrable loads. Deferrable load demand is deterministic and can be switched on/off when necessary; and non-deferrable load demand cannot be differed and present a random behavior. So, only deferrable loads can be managed to reduce purchasing costs.

Connected microgrids need to agree on energy procurement contracts with a distribution system manager (macro station) to meet local load demands through a quality service. Depending on trade policies, contracts must set energy and power prices for different time zones of a day, energy and power contracted for each time zone, and contract and penalty costs. So, trade policies and daily energy and power requirements must be taken into account when defining an economically efficient monthly energy contract; and the daily operation schedule must be generated considering the agreements settled by contract for defining an efficient daily economic dispatch.

Microgrid operation management presents meaningful differences as regards large power systems and their operation scheduling problem [14]. Traditional energy systems involve a large number of power units such as hydro, thermal, or nuclear sources to generate electric energy and distribute it in a given area, often corresponding to a country [16]. Their operation scheduling is classically addressed through either two-stage or multi-stage hierarchical decision models, which are referred to as unit commitment models. In these models, the unit commitment problem is solved in the upper stage and the economic dispatch problem is solved in lower stages. The potential for cost saving has led to extensive research on unit commitment and economic dispatch problems over several decades; and thousands of research articles concerning deterministic problem formulations and methodology for solving stochastic programs have been published [17].

The on/off dynamic of generation sources of traditional energy systems is limited to a daily programming because they are large-scale units with significant startup times, while the dynamic of load demands requires that economic dispatch is updated at least every hour. These characteristics together with the large number of involved units account for the separate resolution of unit commitment problems and economic dispatch problems.

As microgrids are consolidating all over the world, several approaches proposed in the literature for microgrid operation scheduling are based on the unit commitment model. For example, a convex non-linear optimization model to solve the economic dispatch problem is presented in [18]. The model considers the generation cost of a diesel generator and the degradation cost of the battery system capacity. In [19], a multi-objective optimization algorithm that takes into account economic, efficiency and security objectives for economic dispatch is proposed. The studied microgrid is composed of a photovoltaic generator, a battery system, electric vehicles, and transferable loads. The active power delivered by the photovoltaic generator is modeled as a mathematical function of both temperature and solar irradiation. In [20], a game theory approach is proposed to solve the economic dispatch problem. Electricity price is defined in order to maximize social benefit. In [21], a two-stage schedule model is presented. A dynamic programming algorithm is used first for solving the unit commitment problem and, secondly, for sizing micro-turbine reserves to face the uncertainty of photovoltaic generation. In [22], a mixed integer linear program to solve the economic dispatch problem by modeling reserve capacity is proposed. In [23], a two-stage scheduling framework of a microgrid is proposed. In the first stage, a day-ahead scheduling model solves the unit commitment problem and, in the second stage, a several-hour-ahead scheduling model solves the economic dispatch problem. In [1], also, a two-stage energy management strategy is proposed, whose objective is to minimize the microgrid operating cost and predefine the revenue risk at a certain level. In [24], a two-stage model is used for separately solving the unit commitment problem and the economic dispatch problem. In [9], a two-stage model is proposed. In the first stage, a historical data analysis is performed for predicting the demand and then, in the second stage, the unit commitment problem is solved.

Solving the unit commitment problem separately from the economic dispatch problem is suitable for large-scale electric systems; in [25], for example, this is proposed to solve the unit commitment problem for the island of Crete, in [26] for the British electricity system and, in [27] for the German electricity market. But, unlike traditional electric systems, microgrids involve a relatively small number of controllable medium-scale units, which can be hourly on/off-switched due to their reduced startup time. These characteristics do not account for separately solving unit commitment and economic dispatch problems as proposed by the unit commitment model. In other words, computational, physical, and economic viewpoints do not provide a well-founded justification for using the unit commitment model for microgrid operation scheduling.

Randomness of microgrid operating parameters is characterized by a wide range of uncertainties, which brought about the need of re-evaluating classic stochastic methods. Hundreds of research articles concerning uncertainty in microgrid operation scheduling have been published over the last decade, in which generation and load uncertainties are usually modeled by using scenario generation methods. A recent exhaustive review [12] focused on scenario generation methods identifies random search and analytical methods used for analyzing the uncertainty behavior and it remarks that methods based on random search are computationally expensive and that analytical methods can overcome this disadvantage but they need some mathematical assumptions to simplify the problem. Different optimization methods used for modeling the scheduling problem are also identified and classified into either mathematical or heuristic optimization methods. As regards mathematical methods, authors conclude that mixed integer linear programming is the most appropriate approach, and highlight the difficulty in solving nonlinear programming models and the feasibility of using mathematical techniques for linearizing them. Authors also remark drawbacks related to heuristic methods: they provide an estimated solution, convergence can be quite complicated, runtime can be high, and solution can converge to a local minimum instead of the global one.

Another recent review analyzes eleven different techniques used for modeling generation and load uncertainty in a microgrid environment [28]. Authors stress that for any optimization technique, whether classical or advanced, obtaining a conclusive result for problems depends on how uncertainties of both renewable generation and demand are calculated. They highlight the inadequacy of uncertainty modeling methods applicable to renewable sources, in terms of both number and accuracy, and also envisage the scope and need for more flexible models for specific applications.

The operation scheduling modeling usually requires a lot of scenarios, which gives rise to a high computational burden. So, to address this issue, the number of scenarios needs to be drastically reduced because it affects the method performance, which is the main drawback of scenario generation approaches [28]. Since a few scenarios cannot capture the whole spectrum of uncertainty, the inclusion of reserve requirements in stochastic programming is proposed [24]. The combination of stochastic and reserve methods outperforms classical deterministic and stochastic formulations in terms of both cost and reliability. A similar approach that uses both scenario and reserve requirements to obtain a risk-averse uncertainty management of isolated microgrids was later proposed [29]. However, establishing optimal reserve requirements is not an obvious task [17].

The family of stochastic programming approaches explicitly takes uncertainty into account within the model by using a set of discrete scenarios with associated probabilities. The three optimization methods primarily used are stochastic optimization, robust optimization, and chance-constrained optimization [17].

Stochastic optimization requires considering a large number of scenarios to obtain a robust and effective solution, which increases the computational complexity and calls for advanced decomposition techniques. Furthermore, in order to properly generate the scenarios and characterize their probability, it is necessary to know the statistical distribution function of all uncertainty parameters, which is not always available [10].

Robust optimization, as opposed to stochastic optimization, defines uncertainty as a mathematical space within which every infinite potential realization is included regardless of its probability. The identified solution is feasible for all possible realizations encompassed by the mathematical space [10]. But, the system operation could be only ensured if all possible realizations of parameters are involved into their mathematical space, which is difficult to guarantee.

Chance-constrained optimization is a stochastic programming approach that uses probabilistic measures over constraints with uncertain parameters [30]. It is an effective way for modeling fluctuations of uncertain parameters, which are intended as random variables modeled using probability density functions [31]. So, uncertain parameters are represented by their probability distribution functions, rather than scenarios, which can be seen as samples of these distributions. Therefore, random variables representing parameters are not limited to discrete values, and the parameter correlation in consecutive hours is explicitly accounted for in the chance-constrained formulation [17].

A chance-constrained optimization model for daily operation scheduling of hybrid microgrids has been recently presented in the literature [9]. Based on the historical data of demanded power, an ARMA model is used to obtain demand predictions. Then, the scheduling problem is addressed by considering reserves of dispatch units. The three-sigma criteria are used to impose a probability value of 0.9973 on demand constraint. Three main weaknesses of this approach are: uncertainty management is limited to load demands by using the well-established normal distribution function N (0; σ) to represent the random behavior of the forecast error; deferrable loads are not considered; and the stochastic unit commitment model is used, which as discussed above is not a proper strategy for solving the microgrid operation scheduling problem. In [32], for the daily programming of multi-microgrid systems, wind and solar generation units and load demands are considered as sources of uncertainty and modeled using the chance-constrained approach.

In scenario-based models, all possible realizations of uncertainty are considered. However, the number of scenarios could be infinite and a scenario reduction method would be necessary [5]. In [31], a chance-constrained programming restricted scheduling method is used along with the scenario reduction approach for reducing the number of generated scenarios to overcome challenges regarding complexity, high computational burden, and time-consumption caused by a large number of scenarios.

A recently proposed method for microgrid uncertainty modeling is the worst case scenario. Instead of using an explicit probability distribution function, variations of the random parameter are confined to an uncertainty set defined by both upper and lower boundaries. Unlike robust optimization, the worst case scenario estimates prediction intervals for evaluating an uncertainty measure of prediction. Each of them has a range defined by upper and lower limits within which the actual value of the parameter lies with a known confidence level [28].

This work proposes a novel energy management model that integrates energy procurement contract, operation scheduling under uncertainty, and demand response for connected microgrids. If compared to previous works addressing the operation scheduling of connected microgrids, the main outstanding contributions of this paper can be the following:

  • Proposing a novel two-level energy management model that provides a framework for managing microgrid operations efficiently by prioritizing renewable over non-renewable generation, dimensioning both battery system and diesel generators reserves for covering forecast errors of both renewable generation and non-deferrable load demands and also managing deferrable loads.

  • Proposing a novel chance-constrained-based mathematical programming model that can be used for defining an optimal energy procurement contract based on energy market trade policies, and for generating an optimal daily operation schedule based on the energy contract.

  • Presenting a novel criterion for chance-constrained modeling based on historical minimum/maximum values. The modeling aims to consider a high percentage of potential scenarios that might take place during execution. The percentage of covered cases is estimated from simulation results.

The remainder of this paper is structured as follows: Section 2 discusses features and trade policies of the energy market and the main component parts of electric microgrids; Section 3 describes the energy management system, its architecture and the management model, the main characteristics of involved decision problems, and a chance-constrained-based mathematical programming model; Section 4 presents a case study; and finally, conclusions are drawn in Section 5.

Section snippets

Energy market

Microgrids, connected to the main grid through a macro station, need to agree on energy procurement contracts with the macro station to meet local load demands with a resilient and stable service. Depending on trade policies of distribution systems, contract duration can usually vary from one to six months; and both energy and power prices are set for different time zones of a day. So, common aspects defined in energy procurement contracts (Fig. 1) are: contract duration, energy and power

Energy management system

Microgrids management is carried out by the energy management system (EMS) by implementing operation and control functions. Due to the uncertainty of renewable generation and the random behavior of non-deferrable loads, the EMS must both agree on energy procurement contracts with reserves capacity to meet local load demands with a quality service and schedule and control daily operations to meet contract obligations at the lowest cost. To this end, the EMS must define an operation schedule of

Study case

The study case is part of the economic feasibility studies carried out within the framework of a microgrid design project for Gato Colorado, a district in the northeastern region of Santa Fe, Argentina, conducted by the Provincial Energy Company (EPE, for its Spanish initials) of Santa Fe. Homer Pro software [35] was used for design tasks.

Gato Colorado involves an area of 3018 km2 with a population of 1412 people, 70% of which inhabit the urban area and 30% rural clusters. There are 362

Conclusion

The two-level energy management model proposed in this work allows the EMS of microgrids to implement, in the management layer of its architecture, functions required to assist procurement contract definition and operation schedule generation. The first function allows the EMS to define an optimal energy procurement contract based on energy market trade policies, which specifies the daily contracted energy, energy and power contracted for each time zone, energy and power prices to pay in each

CRediT authorship contribution statement

Pedro Luis Querini: Conceptualization, Funding acquisition, Formal analysis, Writing - original draft, Writing - review & editing. Ulises Manassero: Conceptualization, Funding acquisition, Formal analysis, Writing - original draft, Writing - review & editing. Erica Fernádez: Conceptualization, Funding acquisition, Formal analysis, Writing - original draft, Writing - review & editing. Omar Chiotti: Conceptualization, Funding acquisition, Formal analysis, Writing - original draft, Writing -

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

All authors approved the version of the manuscript to be published. This research was funded by National Technological University (Grant numbers: PID-ENUTIFE0007716TC, PID-ENUTIFE0005151TC) and CONICET (Grant numbers: PIP-593).

References (39)

Cited by (8)

  • Optimal management of a hybrid and isolated microgrid in a random setting

    2022, Energy Reports
    Citation Excerpt :

    They adopt a formulation called Affinely Adjustable Robust Optimization (AARO) and, among the possible approaches to handle uncertainty, select the decision rule-based approach, which allows different response strategies, depending on the sign of the forecast error. In Querini et al. (2021) they present a two-level SUC that integrates the power procurement contract, the uncertainty-driven operating schedule and the response to the demand for connected MGs. They use LINGO optimization software that includes solvers that perform linearization automatically.

  • Modeling the risk-based decisions of the microgrid in day-ahead energy and reserve markets considering stochastic dispatching of electrical and thermal energy storages

    2022, Energy Conversion and Management: X
    Citation Excerpt :

    Reviewing the previous studies showed that there are two main types of models proposed for the EM problem of the MGs besides the MGO’s participation in the markets. In the first type, the MGO only supplies the electrical energy demand and it participates in the energy market [1–4,6,8–13,15] or the energy and ancillary service markets [5,7,16–21]. In the second type, the MGO supplies its electrical and thermal energy demands and it only participates in the energy markets [22–24].

  • Clustering-Based Serverless Edge Computing Assisted Federated Learning for Energy Procurement

    2022, APNOMS 2022 - 23rd Asia-Pacific Network Operations and Management Symposium: Data-Driven Intelligent Management in the Era of beyond 5G
View all citing articles on Scopus
View full text