Hybrid Context-awareness Modelling and Reasoning Approach for

Microgrid’s Intelligent Control

Soumoud Fkaier

1,2,3

, Mohamed Khalgui

2,3

and Georg Frey

Chair of Automation and Energy Systems, Saarland University, Saarbruecken, Germany

Tunisia Polytechnic School, Carthage University, Tunis, Tunisia

INSAT LISI Lab, Carthage University, Tunis, Tunisia

Keywords:

Microgrid, Context-awareness, Context Modelling, Context Reasoning.

Abstract:

Modern microgrids are promoting the integration of the information and communication technologies (ICT) in

order to enhance the emerging advanced power management functionalities such as the integration of renew-

able energy sources, distributed storage optimization, demand-response strategies, electric vehicles charging,

power generation rate forecasting and scheduling, etc. For this, sophisticated sensing and smart metering

infrastructures are incorporated in the used equipment as well as in the involved subsystems. Hence many

contextual data are become more and more available and its taking into consideration in the control tasks are

likely to provide promising results. However, making the microgrid control system understand the data and

take the proper decisions based on the identiﬁed context is not an easy task to perform. In fact, recognizing

the relations and meanings of the sensed data is difﬁcult and complex due to the heterogeneity and intricacy

of the involved parts. Hence, providing context-aware modelling and reasoning mechanisms for microgrids

becomes necessary. In this context, this paper contributes with two main solutions. First, a microgrid’s for-

malized design providing an easy and understandable view of the system is provided. This deﬁnition respects

the separation of concerns principle in order to tame the complexity of the complicated system. Second, an

ontology-based context modelling and a rule-based context reasoning in the framework of microgrids are pro-

vided. To show the suitability of the proposed processes, a formal case study is carried out. The proposed

processes are proved to be less resources consuming compared to some of the existing works.

1 INTRODUCTION

A microgrid is a smaller scale version of the tradi-

tional electricity grid that can operate independently

or in a connected mode with the main power grid.

In order to meet the requirements of its users, mi-

crogrids fulﬁl a set of advanced control and manage-

ment tasks such as the integration of renewable energy

sources, distributed storage optimization, demand-

response strategies, electric vehicles charging, power

generation forecasting and scheduling, etc. Such

functionalities are become possible thanks to the inte-

gration of ICTs that allow establishing a bidirectional

information ﬂow. In fact, the great development of

advanced sensing and electric metering devices has

paved the way for applying pervasive computing also

in the electricity management systems (Fkaier et al.,

2016b). Hence, big data ﬂows are available at any

time to the control agents of microgrids (Suslov et al.,

2016).

Although maximal information and details about

the system state are useful, their management and un-

derstanding are not easy to perform by microgrid’s

controller. In fact, controllers have to be aware about

the context of every collected data in order to use it

properly and exploit it in a fruitful way. Identifying

the proper context in a polyvalent system such as a

microgrid is really challenging because of the interre-

lated multidisciplinary “subsystems” constituting the

microgrid: microgrids are in general composed of

generation subsystem (such as photovoltaic panels or

diesel generators), storage subsystem (such as elec-

trical batteries), and consumption subsystem (such as

home area buildings). Deducing a conclusion about

the state of the microgrid at a given moment based on

large sets of sensed data from the mentioned subsys-

tems is difﬁcult since every subsystem can have many

states and each state can be inner to the subsystem

itself or inﬂuenced by other subsystems.

In the context-awareness computing scope, many

116

Fkaier, S., Khalgui, M. and Frey, G.

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control.

DOI: 10.5220/0009780901160127

In Proceedings of the 15th International Conference on Software Technologies (ICSOFT 2020), pages 116-127

ISBN: 978-989-758-443-5

methods and tools have been proposed in previous

works in order to provide solutions to extract accurate

contexts. It is acknowledged that the most efﬁcient

ones are those that build upon the ontology of well-

speciﬁc ﬁelds. Despite its robustness in the modelling

of a system facets and actors, ontology based meth-

ods are still less performant in terms of the consumed

computational power. This feature is important es-

pecially when it comes to control systems that may

require real-time functionalities such as microgrids.

In order to overcome the difﬁculties mentioned

above, the current paper proposes an efﬁcient context-

aware modelling and reasoning process for micro-

grids control. First, a formalization of the deﬁnition

of a microgrid is provided. The deﬁnition allows to

have a clear view on the subsystems composing mi-

crogrids and the relations between them. It is based

on the separation of concerns principle that allows to

reduce the complexity through separating the differ-

ent topics (generation, consumption, storage, and pro-

suming). This deﬁnition paves us the way for deﬁn-

ing a context modelling of the microgrid system. We

introduce a context model based on ontology. It is

performed using the Web Ontology Language (OWL)

(Zamazal, 2020) and it provides semantic microgrid

context representation. Then, a rule-based reasoning

mechanism is proposed to be the mechanism for con-

text extraction. We performed a formal case study to

validate the suitability of the proposed approach.

The outline of this paper is organized as follows.

Section 2 provides an overview of the related works.

Section 3 introduces the proposed microgrid deﬁni-

tion. Section 4 presents the deﬁned ontology-based

microgrids context modelling and rule-based context

reasoning. Section 5 provides the case study and the

performance evaluation. Finally, Section 6 concludes

the paper and presents future perspectives.

2 RELATED WORKS

In this section, an overview of the works in the ﬁeld

of context-awareness and microgrids is provided.

2.1 Context Awareness, Context

Modelling and Reasoning

Context-aware computing is the type of comput-

ing that takes into account the user activity and/or

the changes in its surrounding environment (Khabou

et al., 2019). It was ﬁrst limited to the mobile ap-

plications that try to predict the location of the de-

vice (Fkaier et al., 2017). Then it was extended to

cover many ﬁelds ranging from health care systems

to smart cities (Fkaier et al., 2016a). In general, the

development of context-aware applications has many

steps and rely on different modules according to the

studied case; but the main steps are resumed in the so-

called “context life cycle” which is composed of four

phases: (i) context acquisition, (ii) context modelling,

(iii) context reasoning, and (iv) context dissemination.

The ﬁrst and last phases are considered conventional

phases since most of developers agree on the used

techniques and tools. However, the second and third

phases are still under research because there are open

questions about the improvements of the possible pro-

cesses, mechanisms, and methods. Performance and

suitability of one method over the other are still the

subject of discussions and are likely to be them selves

context-dependent. Every method, as it is the case of

every approach, has its advantages and drawbacks.

The modelling phase is considered as the con-

text representation phase which helps to understand

the context details, properties, and relationships.

Many solutions exist such as the key-value, markup,

graphical, object-oriented, multidisciplinary, domain-

focused, spatial, and ontology-based context mod-

elling. The selection of one technique is generally

based on the focus of the modelling, for example,

whether the relationships and dependencies are to be

emphasized or other feature such as the mobility, etc.

The reasoning phase is considered as the con-

clusion/knowledge extraction based on the modelled

contexts (Cheng et al., 2018). In this phase also

many techniques can be used such as the probabilis-

tic logic-based, Bayesian networks, ontology-based,

case-based, rule-based, supervised as well as unsuper-

vised learning, and fuzzy logic based (Al-Barazanchi

and Vural, 2015).

2.2 Microgrids, Smart Grids, and

Context Awareness

Context-awareness is acknowledged to be one impor-

tant mainstay paradigm for the development of smart

software applications (this is applicable also to the

case of smart grids and smart microgrids). Context-

awareness was involved in various microgrids func-

tionalities. In the following, a short overview of the

different use cases of context-awareness in the ﬁeld of

smart and micro grids is presented.

(Donohoe et al., 2015) present a survey of the

requirements and challenges of context-aware smart

grids. (Radzi et al., 2019) introduce context-

awareness trafﬁc scheduling algorithm for smart

grid’s network. (Obaidat et al., 2018) use the context-

awareness concepts to create a framework for the in-

telligent data management in smart grids. (Gomes

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control

117

et al., 2017) as well as (Stavropoulos et al., 2014)

use context-awareness for the development of energy

management systems of households. (Degha et al.,

2019) propose a solution for efﬁcient energy man-

agement based on ontological context to smart build-

ing. A smart energy equipment management solu-

tion for smart cities based on distributed context-

awareness is proposed by (Choi et al., 2018). (Dono-

hoe et al., 2013) analyze the adoption of context-

awareness paradigm for microgrids storage using

electric cars. (Zhuikov and Kyselova, 2013) introduce

a context-aware control process for microgirds. (Ky-

selova et al., 2016) propose a context-aware frame-

work for energy management systems. (Reddy and

Krishna, 2018) use context-awareness in the load bal-

ancing of power in smart grids. (Meng and Lu, 2015)

introduce a context-aware service customization strat-

egy for smart homes.

Despite their importance, none of the above-cited

researches have deﬁned a holistic context model that

could be used within the microgrids control strategies.

There is a need to an understandable model that cov-

ers all the factors and perspectives that might inﬂu-

ence the context. In addition, efﬁcient reasoning ap-

proach should be provided.

2.3 Summary

As it can be seen from the previous subsections,

context-awareness is a huge computation ﬁled that

has a large variety of techniques and tools and can

have multiple use cases in the one ﬁeld (i.e., smart

electricity management). A study of the manifold

context modelling techniques leads to conclude that

the ontology-based modelling is a very powerful

method. In fact, in comparison with other models

the ontology-based modelling technique is better in

terms of reusability, formality, and interoperabiliy (Li

et al., 2015). It also provides better expressive context

representations and can be independent from applica-

tions (Alegre et al., 2016). Thus, the current paper

adopts the ontology paradigm to represents the con-

text modelling thanks to the mentioned advantages.

However, concerning the reasoning some re-

searchers argue that ontology reasoning might be suit-

able for some kinds of applications and not for some

others. (Li et al., 2015) and (Sezer et al., 2017) ar-

gue that the ontology-based reasoning can be com-

plex and ambiguous. More importantly, ontology-

based reasoning cannot be suitable for time-critical

applications (Wang et al., 2004). However, rule-based

reasoning technique can be combined with ontology-

based models and provide promising efﬁciency. Ma-

chine/deep learning is an important technique that can

be used for reasoning and it is suitable for many use

cases. However, for microgirds it might have some

limitations. First, in order to accomplish its task as

expected, machine learning requires large amounts of

historic data to train. In this case, providing good and

enough training data for each particular microgrid is

required (given that not all microgrids have the same

composition of subsystems as described in Section

3, i.e., their behaviour is not the same). Therefore,

from one hand it is required to provide the appropri-

ate training data of the complicated parts, and from

the other hand this process should be re-done for each

different microgrid and for each time there is a change

in a microgrid structure. However, it is easier to de-

ﬁne rules, which makes the implementation quicker.

For control systems, such as microgirds, it is use-

ful even obligatory to provide understandable expla-

nation of the interpretations made by the reasoning

unit, especially when it comes to security and ﬁnan-

cial issues. Machine learning builds upon the “black

box” models which are not enough clear to be under-

standable by humans. Therefore, it is wiser to rely

on a clear controllable approach. In conclusion, given

that microgrids are a kind of control systems, rule-

based approach represents a better solution.

In summary, given the above presented analysis

of the state of the art, the ontology-based modelling

technique along with the rule-based reasoning tech-

nique are adopted in the contribution of this paper.

3 MICROGRID DEFINITION

In order to clearly introduce the microgrid context on-

tology, it is necessary ﬁrst to deﬁne the conceptual

model of a microgrid. A microgrid is deﬁned as a

small scale electricity grid system that is composed

of a set of subsystems. These subsystems inﬂuence

each-other and their operations are performed under

the control and supervision of a control unit called the

microgrid controller as depicted in Fig. 1. An infor-

mation and communication infrastructure (composed

mainly of sensors, measurement units, and smart me-

tering devices) enables the exchange of bidirectional

information ﬂows between the involved actors.

Four main subsystems can be deﬁned to describe

the involved disciplines in the microgrid system: (i)

generation, (ii) consumption, (iii) storage, and (iv)

prosuming. Each subsystem is characterized by spe-

ciﬁc properties, has well-deﬁned equipment as well as

tasks, and has optimization challenges.

ICSOFT 2020 - 15th International Conference on Software Technologies

118

Figure 1: Subsystems composing a micorgrid.

3.1 Electricity Generation Subsystem

Traditionally, the electricity is generated in big facili-

ties in high voltage power then transmitted as medium

voltage power, and distributed as low voltage power.

However, with the advent of smart grids the genera-

tion process has been changed where especially clean

energy sources are integrated. To reach intelligent

control of the electricity usage, detailed information

about the generation must be provided to the con-

troller so that it can make right decisions like the

switching between the on/off-grid modes. The aware-

ness about the available generators and their mini-

mal/maximal production abilities are important for

the awareness consolidation.

To make the task of perception of electricity gen-

eration information easier, two main generation types

are deﬁned: renewable and non-renewable. Renew-

able energy generation relies on the natural renew-

able energy sources such as the wind, the water, and

the sun. Properties of the available equipment like

the solar photovoltaic panels or wind turbines are

tremendous in the context information creation. Ca-

pacity, size, number, and utilization rate can all im-

pact the context. More importantly, it is true that re-

newable energies have many advantages like reducing

the emission of greenhouse gas but they have non-

trustful behaviour. Despite the weather forecasting

and prediction mechanisms, the generation amounts

remain unknown or at least non-precise. Therefore,

clear view about the considered renewable energies

may help in the control task. Non-renewable energy

generation relies on some materials such as the oil or

coal to produce electricity. Despite the bad impact on

the globe safety, non-renewable energies guarantee a

trustful generation behaviour. Therefore they are in

general used as the backbone of the power grid. The

percentage of reliance on these energies may help in

many tasks such as the tariff deﬁnition.

3.2 Electricity Consumption Subsystem

Electricity is nowadays required in nearly all the ﬁelds

of the people’s daily life. Information about the con-

sumption proﬁles of the consumers supplied within

one microgrid is required to determine the control

contexts. Also analysing the human activities like the

working hours, the holidays, the social events, etc.,

helps in the context recognition. Similarly, the exis-

tence of industrial establishments, big shopping cen-

ters, or even product stores among the consumers is

necessary to take into account in the control strategy.

It is required that a microgrid provides supply for

its consumers, however in some cases (peak hours

consumption) shortages can occur, consequently not

all consumers can be satisﬁed. In that case, a mi-

crogrid controller need to understand the context and

must serve consumers according to their emergency

level and priorities. To make the context perception

easier for the controller, consumers are classiﬁed into

two groups: critical and non-critical consumers. Crit-

ical consumers are consumers whom the shortages or

faults in the electricity supply to their activities imply

catastrophic losses in terms of people lives, money,

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control

119

or security issues (for example hospitals, banks, and

police ofﬁces). Electricity should be continuously

supplied to such kind of consumers. Thus, consid-

erable awareness should be attached to this informa-

tion. Non-critical consumers are consumers whom

the supply can be deffered to later time, and short

blackouts do not induce losses rather they induce a

deterioration of the quality-of-service.

3.3 Prosumers Subsystem

Equipments and tools of managing the renewable en-

ergy are witnessing continuous evolution which facil-

itate their adoption and integration in many emerging

use cases such as the electricity generation from the

roof-top photovoltaic panels. Hence, the house lord

can play the role of electricity producer and consumer

at the same time (from where the term “prosumer” is

born). Another use case, is that the produced electric-

ity can be supplied to electric cars as a type of storage.

Hence, enabling such advanced functionalities to the

users can be the source of sudden big energy demand

(in case of simultaneous bad weather and high use) or

also the release (during good weather and low use).

The behaviour of prosumers is generally unpre-

dictable, at least if we assume that the metering and

sensing infrastructure ensures security and privacy of

the user’s data which makes difﬁcult forecasting its

activity. That is why being aware of such kind of

users allows to take into account any potential sharp

demand-response changes.

3.4 Storage Subsystem

Storage systems play a pivotal role in almost all func-

tionalities of the microgrid control. It allows to store

electricity and provide it back during the peak shav-

ing process (Dongol et al., 2018), dynamic demand-

response mechanisms, load shifting, etc., in order to

guarantee better supply continuity and better distur-

bances avoidance. It is necessary to a microgrid con-

troller to be aware about the size, the type, the capac-

ity, the distribution and other parameters of storage

equipment in order to be ensure reliable functionali-

ties.

4 CONTEXT MODELLING AND

REASONING APPROACH FOR

MICROGRIDS

In this section, we provide the deﬁnition of the pro-

posed microgrid’s context model as well as the pro-

posed reasoning mechanism. But before starting, it is

necessary to show how the context information is to

be used in the control process.

Fig. 2 depicts the abstract architecture view of the

microgrid control system which is composed of three

main layers: (i) Physical Layer: which represents the

process and ﬁeld components of the subsystems de-

ﬁned in Section 3. (ii) Context Acquisition Layer:

which is the layer subject of the current paper con-

tributions. Finally, (iii) Control Layer: which holds

the controller of the microgrid. In the following, the

functionalities of the second layer, i.e., Context Ac-

quisition Layer, are introduced.

Functionalities of the physical layer such as gath-

ering data from the heterogeneous sources, as well as

functionalities of the control layer such as deﬁning

reconﬁguration and adaptation policies are out of the

scope of this paper.

Figure 2: Microgrid control system abstract architecture.

4.1 Ontology-based Microgrid Context

The purpose behind deﬁning the microgrid context

model using the ontology is to provide an easy un-

derstandable model. Also, ontology based models are

easily extensible, so whenever new concepts need to

be added the basic model could be reused. More-

over, although the contexts in a microgrid are not easy

to deﬁne and even complex, the ontology allows us

to derive the formal semantic knowledge about the

microgrid subsystems, their components, and con-

straints.

At a given moment, the microgrid controller may

not need every context detailed information, but it is

of great usability to provide a holistic context model.

Most of the previous works provide a context mod-

elling for some speciﬁc use cases like the context-

aware load-balancing, the context-aware electric ve-

hicle charging, the context-aware power generation

ICSOFT 2020 - 15th International Conference on Software Technologies

120

Figure 3: Microgrid Context Model Using OWL.

forecasting. However, none of them provide a context

deﬁnition for the whole microgrid. It is difﬁcult to

model every single case and every single scenario, es-

pecially with the expanded subsystems as presented in

Section 3. Nevertheless, we provide a holistic context

model for microgrids context-awareness which incor-

porates the most used (conventional) facets. When-

ever necessary, the model could be extended, i.e., new

ontology entities and deeper details could be added.

Fig. 3 depicts the proposed microgrid context on-

tology. The model is presented based on the OWL

(Web Ontology Language). The considered entities

are the entities related to the control task and that can

affect the intelligent functionalities such as the load

balancing, the peak shaving, the generation optimiza-

tion, the storage optimization and others. The micro-

grid context is formed mainly with six main parts:

(i) generators, (ii) storage, (iii) consumers, (iv) pro-

sumers, (v) environment, and (vi) historical state.

Generators Ontology: This ontology entity is used

to represent the electricity generators within a micro-

grid system. It is the generalization of the renewable

and traditional energy generators where each kind is a

generalization of some speciﬁc classes of generation

such as the Solar Panels or the Coal Energy Plant.

For space constraints, we detail one example of con-

text information of solar panels. It is important for the

controller to get some parameters such the voltage at

maximum power point (VMP), the open circuit volt-

age (VOC), the ﬁll factor (FF), etc. Also the number

of panels, their size, etc.

Storage Ontology: This ontology entity is used to

represent the electricity storage system in a micro-

grid. The Battery speciﬁc class is a conventional type

of electricity storage. Information required are for

example the power requirement range (expressed in

kW), the standing time interval (in hours), the nomi-

nal AC coupling voltage (VAC), the dimensions, the

percentage of over and under voltage allowance, etc.

Consumers Ontology: This ontology entity is used

to represent the consumers of electricity within a mi-

crogrid. In general, consumers can be characterized

by their culture (for example consumers in a devel-

oped country do not consume the same amount of

consumers in developing ones). Even the ﬁnancial

level and psychology (how is the susceptibility to ac-

cept and use new restrictions and technologies, etc.)

can inﬂuence the consumption rates. Consumers are

classiﬁed into two classes according to their priority:

CriticalConsumers and NonCriticalConsumers.

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control

121

Figure 4: Context Reasoning Process.

Prosumers Ontology: This ontology entity is used

to represent the prosumers such as the ElectricVehicle

and HomeWithSolarPanelRoofTop speciﬁc classes.

Information about how many home has the roof-top

solar panels and what are their characteristics is nec-

essary. Also the number of potential usage of electric

vehicles and whether it disposes of solar panels are

useful to know.

Environment Ontology: This ontology entity is used

to describe the environmental inﬂuencers on the con-

text of control of the microgrid. Three main con-

ditions are taken into account which are the timing,

weather forecast, and physical equipment. The time is

represented through the speciﬁc class Time which is a

generalization of the Season, DayOfWeek, PeriodOf-

Day, TimeOfDay which impact directly the schedul-

ing, the generation, and the load balancing tasks.

The Weather speciﬁc class also is important to

the renewable energy generation and that is generally

feeded by the weather forecast service providers. The

PhysicalEquipment speciﬁc class represents the phys-

ical sources of the context or let say the low level con-

text. It is classiﬁed into two types of microgrid equip-

ment: Devices such as the sensors, actuators, and

smart meters, and the power lines especially the emer-

gency and trading power lines. Being aware about the

emergency power lines for example, helps to set the

emergency reconﬁgurations in case of internal faults

or also in case of faults in neighbor microgrids.

Historical Microgrid State Ontology: This ontol-

ogy entity is used to describe the previous activities

that a microgrid has performed. Historical informa-

tion has an inﬂuence on the decision making for cur-

rent and future conﬁgurations/control tasks.

There are dependencies between the different en-

tities, we present among them the dependency of the

renewable energy generators and the prosumers to the

environment ontology.

4.2 Rule-based Context Reasoning

In this section, the proposed lightweight context rea-

soning process is introduced. The process ﬂow is de-

scribed with three steps as depicted in Fig. 4: (1) for-

matting the collected data, (2) comparison with the

current context, and (3) search for matching rules.

Before providing the pseudo-code of each step,

it is noted that we assume that at any given instant,

the microgrid has a context (i.e., a state) that can be

changed according to the results of the provided rea-

soning process. It is also assumed that the new found

conclusions are forwarded to the controller of the mi-

crogrid (control layer in Fig. 2), and the latter has

to make decisions based on its control protocols (re-

lated to tariff calculation, sustainability policies, se-

curity policies, optimization policies, etc.). The role

of the proposed processes in this paper is to get and

recognize context changes.

4.2.1 Formatting Data

Sensors and metering devices provide real numerical

values that are not meaningful or at least not helpful

in deducing the semantic relations between the facts

related to the context entities (as shown in the ontol-

ogy deﬁnition in Section 3). That is why, we propose

the following process to represent the sensed data.

We note that, a Context Attributes Model Store

(CAMS) is provided as input to the process. The store

contains context attributes models where each model

ICSOFT 2020 - 15th International Conference on Software Technologies

122

represents one context attribute and its related labels

as depicted in Fig. 5. A label is a description of the

range of values of the attribute. For example, temper-

ature is a context attribute of the weather ontology.

The interval of values [0, 10] is described as low tem-

perature, [10, 25] is medium temperature, and [25, 35]

is high temperature. Hence, low, medium, and high

are deﬁned as labels. In the same way, the consump-

tion rate (in kW) of one residential unit is an attribute

of the non-critical consumers entity that belongs to

the consumers ontology. The values between [0, 70]

is labelled as very low, between [70, 250] is labelled

as low, between [250, 600] is labelled as medium, and

between [600, 1200] is labelled as high.

Figure 5: Context attributes model.

Some of the context attributes are given a criti-

cal importance and receiving an information related

to one of them implies the immediate sent of an alert

to the controller. For example, if the sensor of ﬁres

detects a ﬁre the process has to end immediately and

send an alert message to the controller. Same case for

situations related to the storms or blackouts.

Data formatting pseudo-code is as follows:

Step 1: Parse received values and check if one of

the critical attributes contains a value.

Step 2: For each received attribute, read the re-

lated context attribute model from the context at-

tribute model store CAMS and decide which label is

to be assigned to the attribute by comparing the value

to the minimum and maximum values of each range,

i.e., we say that the label of a range of one model is

assignable to the input x

if:

∈ [RangeMinValue,RangeMaxValue] (1)

Step 3: Generate a context row using the re-

trieved labels. A context row, denoted by CR is

deﬁned as a set of context items denoted by ci,

hence CR = {ci

,ci

,...,ci

} where n is the number

of context items. A context item is deﬁned as a

tuple ci =< type,label,value > where value is the

collected/sensed/measured value, label is the label

of the context attribute determined in the previous

step, and type is the semantic type of the input and

it is presented by its root through the ontologies

, entities e

, and attributes a

. For example, the

context item of the temperature will be as follows:

ci =< o

environment

weather

temperature

,medium,15 >.

The context item of the residential unit

consumption will be as follows: ci =<

consumers

noncritical

residence

consumption

,low,100 >.

4.2.2 Comparison with Current Context

Given that context awareness should be provided in

real time in order to be efﬁcient, many calculations

have to be processed. However, the input data could

be the same or slightly different from the current

one in many situations. So calculations will produce

nearly same results which have not an impact on the

context. For example, for the demand response han-

dling task, knowing the period of the day (morning

/afternoon or night) may help to reduce the calcula-

tions during some hours of each period. For instance,

supposing that we are in the night, the periodic super-

vision of the consumption rates indicates that it is a

low consumption. Thus, there is a need to a method

allowing to avoid triggering the reasoning process that

might produce similar results.

To avoid leading non-useful computations, we

propose to create an intermediate step that compares

the new context values with the current ones. The

comparison is based on a predeﬁned similarity thresh-

old of the values of the context row. The threshold is

deﬁned according to the assumptions and preferences

of the system owners, for example 10%.

Let X = {x

,...,x

} be the sensed values of

the input context row and V = {v

,...,v

} be

the values of the current context row CR. To see how

much the new sensed values are different from the

current ones the following formula is applied:

P =

∑

j=1

(|x

− v

|/[(x

+ v

)/2] × 100)/n (2)

For example, let the current context row values

be V = {35,1, 1100} and let the sensed values be

X = {30,1,1000}. The percentage P is 8.3% which

is less than the threshold 10%. In this case, it is rec-

ommended to not re-trigger the search for rules and

rather “no changes recommendation” can be sent to

the controller because the difference is negligible and

does not imply a context change.

If the difference is greater than the threshold then

a search for the matching context rule is necessary.

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control

123

4.2.3 Search for Matching Rules

In order to ﬁnd the proper context, a context rule

store (CRS) is deﬁned. A rule has the form of

condition −→ conclusion, where the condition is a

conjunction of a set of premises. The context items

of the input context row are used as premises of the

rules, and the conclusions of rules will be sent to the

controller as recommendations of reconﬁgurations. In

this phase of the process, premises are given prior-

ities allowing to facilitate the rule matching process

as detailed in the following. The pseudo-code of the

matching rule search is given as follows:

Step 1: Select the set of triggerable rules. This is

done based on which attribute has changed its values

(i.e., we select rules containing as premises the related

attributes).

Step 2: If only one rule is selected in Step 1 then

go to Step 3. Else if more than one rule is selected

in Step 1 then compare the number of premises of the

rules. The longest rule is the winner rule. Else if there

are more than one winner rule then;

Check the priorities of the attributes. Select the

rule containing the attribute having the highest prior-

ity. If all attributes are of equal priorities then calcu-

late the Euclidean distance between the input context

row values X and the conditions of the selected rules

as follows

d(X,R

) =

∑

i=1

− r

)

(3)

Where r

is the mid-range of the relevant attribute

range. The winner rule is the one having the minimal

distance to X.

Step 3: Trigger winner rule. If all rules selected

in Step 1 are triggered, then ﬁnish. Else, go to Step 2.

The conclusions of all triggered rules will be sent

to the controller as recommendations/suggestions that

help on establishing the proper control strategies. For

example, a recommendation to sell energy when there

is surplus of renewable energy generation and low

consumption rates, etc.

5 FORMAL CASE STUDY

Let us consider a microgrid called mg that has these

features: mg relies on the utility main power grid and

a ten solar panels plant. mg should supply one resi-

dential area (composed only of homes) and one hos-

pital. mg has also one battery as a storage system.

Using the proposed microgrid context ontology, in

the current case the controller of mg should handle:

(i) the renewable energy generation subsystem, (ii)

the consumers subsystem, and (iii) the storage sub-

system. The traditional electricity generators and the

prosumers are not considered. Table 1 depicts some

of the context attributes stored in CAMS.

Table 1: Context Attribute Labels.

Attribute name Values

Connection to main grid on, off

Solar panel production high, medium, low

Battery reserve empty, full, half full

Battery activity charging, discharging, idle

Consumer priority critical, non-critical

Weather windy, cloudy, sunny, rainy, stormy

Season summer, winter, autumn, spring

Day of the week working day, week-end day, holiday

Period of the day morning, afternoon, night

Time of the day peak, normal, zero-consumption hours

Consumption very low, low, medium, high, very high

Let the current context row of the microgrid, de-

noted by CR be composed with the set of context

items deﬁned in Table 2.

Table 2: Current Context Items.

Id tuples

< o

HistState

GridConn

mode

,OnMode,1 >

< o

Generators

Renew

ProdRate

,medium, 50% >

< o

Storage

Battery

Reserve

,hal f F ull, 50% >

< o

Environment

Weather

Sunlight

,sunny,80% >

< o

Environment

Time

Season

,summer,3 >

< o

Environment

Time

DayO f Week

,workingday, 1 >

< o

Environment

Time

PeriodO f Day

,morning,1 >

< o

Consumers

NonCritical

Residence

,low, 110 >

Let the input vector of collected data be

as follows I= { (Connection to the grid =

true), (SolarPanelProduction= 1000 Wh), (Bat-

teryReserve= 20%), (Weather= 60% sunny), (Date-

Time= 04.07.2010:11:30:00), (ConsumptionRate=

350 kW)}. Following the proposed approach, the

ﬁrst step, which is the data formatting step, pro-

cesses the data vector I using the CAMS and re-

sults on the following context row, denoted by R =

{ci

,ci

Table 3: Generated Context Items.

Id tuples

< o

HistState

GridConn

mode

,OnMode,1 >

< o

Generators

Renew

ProdRate

,medium, 50% >

< o

Storage

Battery

Reserve

,empty, 20% >

< o

Environment

Weather

Sunlight

,sunny,80% >

< o

Environment

Time

Season

,summer,3 >

< o

Environment

Time

DayO f Week

,workingday, 1 >

< o

Environment

Time

PeriodO f Day

,morning,1 >

< o

Consumers

NonCritical

Residence

,medium, 350 >

ICSOFT 2020 - 15th International Conference on Software Technologies

124

Since any of the attributes is critical, no alerts will

be sent to the controller and the process goes to the

next step which is the comparison with the current

context step. A quick parsing of the attributes returns

those that have changed their values which are the

battery reserve and the consumption level. Therefore,

calculating the percentage of the change is to be done.

The corresponding numerical values of the con-

texts are given as follows: V = {1, 50%, 50%,

80%, 3, 1, 1, 110}, and X= {1, 50%, 20%, 80%,

3, 1, 1, 350}. The difference calculation gives

P =

∑

j=1

(|x

−v

|/[(x

)/2]×100)/8 = 23.75%.

Given that the threshold of the difference between the

current and new context is of 10%, it is necessary

to move to the next step which is the search for the

matching context rules. For this, the CRS is deﬁned

as depicted in Table 4.

Applying the ﬁrst step of the third phase results in

the recognition of R1, R7 and R8 as triggerable rules.

Then, according to the second step of this search a

comparison between the selected rules in terms of

number of premises give the following results: (R1,

2), (R7, 3), and (R8, 3). There is no longest rule be-

cause R7 and R8 have the same size of premises.

In this case, a comparison of the premises priority

is performed. In this case study, more importance is

given to the integration of renewable energy. Thus,

whenever a rule is dealing with renewable energies it

has more priority over other rules. As conclusion, R8

is the ﬁrst rule to trigger. Triggering R8 gives as a

result Switch on to main grid. Then, R7 and R1 are

triggered successively. Therefore these recommenda-

tions are sent to the controller of mg: switch on to

main grid and start charging batteries.

6 PERFORMANCE EVALUATION

6.1 Numerical Analysis

A major challenge of today’s smart applications is

its level of awareness and its efﬁciency in terms of

both: (i) reliable behaviours, and (ii) computational

resources consumption. For this, efﬁciently manage

and exploit the collected (and/or sensed/measured)

data is a key factor that allows to assess the quality

of an approach. In this context, the proposed context

reasoning in this paper introduces a comparison phase

as an intermediate phase between the data representa-

tion phase and the rules matching phase at the aim

of reducing possible non-needed computations, i.e.,

those that end with the same conclusion. Therefore,

to make our process smart, a threshold of difference

is deﬁned to decide about the similarity of the current

context values and the new ones, i.e., how much cur-

rent and new contexts are different.

To show more the beneﬁts of the contribution, let

us consider the current context row A = {ci

,ci

}

composed of three context items where the values are

given by V = {35,1,1100}. The ﬁrst attribute repre-

sents the percentage of the renewable energy genera-

tion and it has this context attribute labels: low for the

interval [0%, 33%], medium for the interval [34%,

66%], and high for the interval [67%, 100%]. The

second attribute represents the connection to the util-

ity grid (1 for true, and 0 for false). Lastly, the third

attribute represents the consumption (in Wh) and it

has this context attribute labels: low for the values in

[0, 1000], medium for the values [1000, 2000], and

high for the values [2000, 3000].

Let us consider the following ﬁve vectors rep-

resenting ﬁve periodic context measurements during

one morning: V

= {32,1,1020}, V

= {40,1,1200},

= {35, 1,900}, V

= {35, 1,1400}, and V

{35,1,1488}. Calculating the percentage difference

P, between A and the collected vectors results in val-

ues less than the threshold as depicted in Fig. 6.

Figure 6: Comparison between current and new rows.

Given that the reconﬁgurations are costly, the

slight ﬂuctuations of the generation and consumption

inside one interval of values or around its bounds and

below the threshold value should not generate context

changes. Deﬁning a threshold values allows to deter-

mine the supportable possible ﬂuctuations within one

deﬁned context. This phase allows to reduce consider-

ably non-necessary rule matching processes, thus al-

lows to improve the computational resources usage.

6.2 Discussion

The proposed approach shows the ability to express

the relation of the control task to the related context.

It introduces a deﬁnition of the context ontology’s that

may inﬂuence the behaviour of the microgrid control-

ling tasks and contributes to the enrichment of the de-

velopment of the system intelligence. Compared to

the works presented in Section 2.2, and to the best of

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control

125

Table 4: Rules Samples.

Id Rules

R1 (BatteryReserve is empty) and (Weather is sunny) −→ Start charging batteries

R2 (Weather is sunny) and (PeriodOfDay is morning) and (SolarPanelProduction is high) −→ Switch off from main grid

R3 (Weather is sunny) and (PeriodOfDay is afternoon) −→ Stop charging batteries

R4 (Season is winter) and (ConsumerPriority is critical) −→ Switch on to main grid

R5 (PeriodOfDay is night) and (BatteryActivity is discharging) −→ Turn batteries on idle mode

R6 (Consumption is low) and (Weather is sunny) and (Season is summer) −→ Switch off from main grid

R7 (Consumption is medium) and (Weather is sunny) and (Season is summer) −→ Switch on to main grid

R8 (Consumption is medium) and (Weather is sunny) and (SolarPanelProduction is medium) −→ Switch on to main grid

our knowledge, the current work is the ﬁrst to deﬁne

a microgrid ontology that takes into consideration the

four system perspectives: generation, consumption,

prosuming, and storage.

The three-steps reasoning process is lightweight

and provides computational resources reduction

thanks to the second step, i.e., the “comparison with

current context step”. In fact, it allows to improve the

semantics of the context information and avoid non-

useful calculations. Compared to the works reported

in (Skillen et al., 2014), (Quinn et al., 2017), our pro-

posed reasoning method provides better resources us-

age thanks to the ﬁltering of the “comparison with

current context step”. Also, recognizing urgent at-

tributes in step 1 allows to lighten the computational

overhead for critical situations. Search for matching

rules process execution is not triggered if critical at-

tributes are obtained also when there are small negli-

gible changes.

The efﬁciency of the proposed contributions

would be more veriﬁable by the development of the

physical and control layers which allow to provide a

complete process of a context-aware application work

ﬂow.

7 CONCLUSIONS

The current paper introduces an approach of context

modelling and reasoning for microgrids. A general-

ized and holistic design of the microgrid is presented,

the design deﬁnes four subsystems: generators, con-

sumers, prosumers, and storage. The current paper in-

troduces a microgrid ontology based on the proposed

design to present the different parts that can impact

the context of the system control. The ontology is de-

ﬁned using the OWL language and it is general, holis-

tic, and extensible. The proposed context model also

shows the dependencies between the different subsys-

tems of a microgird and helps to ﬁnd common un-

derstanding of the microgrid context. Then, a three-

steps process of rule-based context reasoning is intro-

duced. The reasoning process has as originality the

lightweight process thanks to a ﬁltering mechanism

that is processed before the rules search. This pro-

cess allows to reduce computational resources usage.

A formal case study is conducted and the suitability

of the context deduction is demonstrated. In future

work, many interesting research directions could be

tackled. The deﬁnition of reconﬁguration and adapta-

tion mechanisms to be performed by the controller of

the microgrid needs to be studied. Also, application

of the proposed concepts to a real-world case study

would be of great importance.

REFERENCES

Al-Barazanchi, S. A. M. and Vural, A. M. (2015). Model-

ing and intelligent control of a stand-alone pv-wind-

diesel-battery hybrid system. In 2015 International

Conference on Control, Instrumentation, Communi-

cation and Computational Technologies (ICCICCT),

pages 423–430. IEEE.

Alegre, U., Augusto, J. C., and Clark, T. (2016). Engineer-

ing context-aware systems and applications: A survey.

Journal of Systems and Software, 117:55–83.

Cheng, X., Yang, J., and Xia, L. (2018). A service-oriented

context-awareness reasoning framework and its im-

plementation. The Electronic Library.

Choi, C., Esposito, C., Wang, H., Liu, Z., and Choi, J.

(2018). Intelligent power equipment management

based on distributed context-aware inference in smart

cities. IEEE Communications Magazine, 56(7):212–

217.

Degha, H. E., Laallam, F. Z., and Said, B. (2019). Intelli-

gent context-awareness system for energy efﬁciency

in smart building based on ontology. Sustainable

Computing: Informatics and Systems, 21:212–233.

Dongol, D., Feldmann, T., and Bollin, E. (2018). A model

predictive control based peak shaving application for

a grid connected household with photovoltaic and bat-

tery storage. In SMARTGREENS, pages 54–63.

Donohoe, M., Jennings, B., and Balasubramaniam, S.

(2013). Context-aware microgrid storage using elec-

tric cars. In IEEE PES ISGT Europe 2013, pages 1–5.

IEEE.

Donohoe, M., Jennings, B., and Balasubramaniam, S.

(2015). Context-awareness and the smart grid: Re-

ICSOFT 2020 - 15th International Conference on Software Technologies

126

quirements and challenges. Computer Networks,

79:263–282.

Fkaier, S., Romdhani, M., Khalgui, M., and Frey, G.

(2016a). Enabling reconﬁguration of adaptive control

systems using real-time context-aware framework. In

2016 IEEE/ACS 13th International Conference of

Computer Systems and Applications (AICCSA), pages

1–8. IEEE.

Fkaier, S., Romdhani, M., Khalgui, M., and Frey,

G. (2016b). R2tca: New tool for develop-

ing reconﬁgurable real-time context-aware frame-

work—application to baggage handling systems. In

Proc. Int. Conf. Mobile Ubiquitous Comput., Syst.,

Services Technol.(UBICOMM), pages 113–119.

Fkaier, S., Romdhani, M., Khalgui, M., and Frey, G.

(2017). Context-awareness meta-model for reconﬁg-

urable control systems. In ENASE, pages 226–234.

Gomes, L., Sousa, F., and Vale, Z. (2017). Enaplug–an envi-

ronmental awareness plug to test energy management

solutions for households. In EPIA Conference on Ar-

tiﬁcial Intelligence, pages 253–259. Springer.

Khabou, N., Rodriguez, I. B., and Jmaiel, M. (2019). An

overview of a novel analysis approach for enhancing

context awareness in smart environments. Information

and Software Technology, 111:131–143.

Kyselova, A., Verbitskyi, I., and Kyselov, G. (2016).

Context-aware framework for energy management

system. In 2016 2nd International Conference on In-

telligent Energy and Power Systems (IEPS), pages 1–

4. IEEE.

Li, X., Eckert, M., Martinez, J.-F., and Rubio, G. (2015).

Context aware middleware architectures: survey and

challenges. Sensors, 15(8):20570–20607.

Meng, Z. and Lu, J. (2015). A rule-based service cus-

tomization strategy for smart home context-aware au-

tomation. IEEE Transactions on Mobile Computing,

15(3):558–571.

Obaidat, M. S., Reddy, N. R., Venkata, K. P., and Saritha,

V. (2018). Context-aware middleware architectural

framework for intelligent smart grid data manage-

ment. In 2018 International Conference on Computer,

Information and Telecommunication Systems (CITS),

pages 1–5. IEEE.

Quinn, S., Bond, R., and Nugent, C. (2017). Ontological

modelling and rule-based reasoning for the provision

of personalized patient education. Expert Systems,

34(2):e12134.

Radzi, N., Suhaimy, N., Ahmad, W., Ismail, A., Abdullah,

F., Jamaludin, M., and Zakaria, M. N. (2019). Con-

text aware trafﬁc scheduling algorithm for power dis-

tribution smart grid network. IEEE Access, 7:104072–

104084.

Reddy, N. R. S. and Krishna, P. V. (2018). Context aware

power management in smart grids using load balanc-

ing approach. International Journal of Advanced In-

telligence Paradigms, 11(3-4):335–347.

Sezer, O. B., Dogdu, E., and Ozbayoglu, A. M. (2017).

Context-aware computing, learning, and big data in

internet of things: a survey. IEEE Internet of Things

Journal, 5(1):1–27.

Skillen, K.-L., Chen, L., Nugent, C. D., Donnelly, M. P.,

Burns, W., and Solheim, I. (2014). Ontological user

modelling and semantic rule-based reasoning for per-

sonalisation of help-on-demand services in pervasive

environments. Future Generation Computer Systems,

34:97–109.

Stavropoulos, T. G., Rigas, E. S., Kontopoulos, E., Bassil-

iades, N., and Vlahavas, I. (2014). A multi-agent co-

ordination framework for smart building energy man-

agement. In 2014 25th International Workshop on

Database and Expert Systems Applications, pages

126–130. IEEE.

Suslov, K., Solodusha, S., and Gerasimov, D. (2016). Deter-

mination of parameters of adaptive law for the control

of an off-grid power system. In 2016 5th International

Conference on Smart Cities and Green ICT Systems

(SMARTGREENS), pages 1–7. IEEE.

Wang, X. H., Zhang, D. Q., Gu, T., and Pung, H. K. (2004).

Ontology based context modeling and reasoning using

owl. In IEEE annual conference on pervasive comput-

ing and communications workshops, 2004. Proceed-

ings of the second, pages 18–22. Ieee.

Zamazal, O. (2020). A survey of ontology benchmarks for

semantic web ontology tools. International Journal

on Semantic Web and Information Systems (IJSWIS),

16(1):47–68.

Zhuikov, V. and Kyselova, A. (2013). Integration of

context-aware control system in microgrid. In

2013 IEEE XXXIII International Scientiﬁc Conference

Electronics and Nanotechnology (ELNANO), pages

386–390. IEEE.

Hybrid Context-awareness Modelling and Reasoning Approach for Microgrid’s Intelligent Control

127