Development of an outdoor lighting control system using expert system

Energy and Buildings 130 (2016) 773–786

Contents lists available at ScienceDirect

Energy and Buildings journal homepage: www.elsevier.com/locate/enbuild

Development of an outdoor lighting control system using expert system Selcuk Atis a,∗ , Nazmi Ekren b a b

Electrical and Energy Department, Vocational School of Technical Sciences, Marmara University, 34722 Istanbul, Turkey Department of Electrical and Electronic Engineering, Technology Faculty, Marmara University, 34722 Istanbul, Turkey

a r t i c l e

i n f o

Article history: Received 26 July 2016 Accepted 24 August 2016 Available online 4 September 2016 Keywords: Outdoor lighting Energy conservation Expert system

a b s t r a c t In this study, an intelligent energy-efficient outdoor lighting control system was developed that could be used in green buildings as well as intelligent building functions, and contribute to the reduction of carbon dioxide emissions by using more conservation of electric energy and daylight more efficiently. The intelligent energy-efficient lighting control system was based on expert system is one of the artificial intelligence techniques and has four functions running in real-time. The first function is controlling and monitoring of the lamp groups. The other functions are fault diagnosis in lambs and power lines connected to the lamp groups and the load estimation of lamp groups. The expert system was written in two separate computer and microcontroller based environments by using knowledge-based rules. The rule base of real-time control and monitoring function contains 213 rules. During the education semester operation mode which real-time control and monitoring function was implemented, an average of 33% conservation was achieved in energy consumption. The system is the first expert system application in which an expert system is used to control and monitor outdoor lighting as well as perform load estimate and fault diagnosis. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Nowadays, in which the importance of sustainable development increasingly understood, the value of the efforts directed to energy efficiency increases at the same rate. In this framework; enhancing energy efficiency, preventing unconscious usage and dissipation, decreasing energy intensity either in sectoral base or in macro level are preferencial and important components of national energy policies in all the stages from energy production and transmission to the final consumption [1]. As an indicator of energy efficiency, the “Energy Intensity”, in other words amount of primary energy consumed per GDP (Gross Domestic Product) is one of the most significant indicators. According to 2013 energy statistics of International Energy Agency (IEA), the energy intensity of Organisation for Economic Co-operation and Development (OECD) countries is recorded as 0.13 on average and 0.11 on average for European Union (28 countries) while this ratio was 0.18 in Turkey (based on USD in 2005) [2]. These ratios indicate that Turkey uses energy less efficiently than OECD and European Union countries. When these ratios are considered, different meth-

∗ Corresponding author. E-mail address: [email protected] (S. Atis). http://dx.doi.org/10.1016/j.enbuild.2016.08.066 0378-7788/© 2016 Elsevier B.V. All rights reserved.

ods need to be adopted for energy efficiency and conservation, and on a national scale extended to all fields using energy. One of the fields in which energy can be used more efficiently with broad individual and corporate participation is lighting. Worldwide, grid-based electric lighting consumes 19% of total global electricity production, slightly more electricity than used by the nations of OECD Europe for all purposes. Lighting requires as much electricity as is produced by all gas-fired generation and 15% more than produced by either hydro or nuclear power. The annual cost of this service including energy, lighting equipment and labour is USD 360 billion, which is roughly 1% of global GDP [3]. Global lighting-related CO2 emissions are estimated to be 1 528 million tonnes (Mt) from grid-based electric lighting, 190 Mt from fuel-based lighting and 181 Mt from vehicle lighting. This makes a total of 1 900 Mt of CO2 , which is 70% of the global emissions of light-duty passenger vehicles [3]. Globally an estimated 218 TWh of final electricity was consumed by outdoor stationary lighting in 2005, amounting to about 8% of total lighting electricity consumption [3]. In indoor and outdoor lighting, by: • Selecting elements fit for purpose, • Using lamps that consume less energy while performing the same task,

774

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

• Utilizing more daylight, • Using control and monitor systems, it is possible to achieve considerable conservation in energy consumption and contribute to the reduction of carbon dioxide emissions. Among these methods, one of the most important aspects is controlling and monitoring the lighting. In an uncontrolled lighting system, it is extremely difficult to achieve energy conservation. Lighting control systems are used in controlling and monitoring a lighting system by running either integrated into the building automation system or as a stand-alone unit. When studies conducted in recent years on energy efficiency and economic usage of indoor lighting are analyzed, the main areas of concern can be summarized as: the criteria for energy efficient lighting in buildings [4], new methods based on a sensor network for designing lighting control systems for industrial buildings [5], seeking an adaptive control strategy for lighting control in office spaces with the aim to reduce energy consumption and provide occupant comfort [6], the savings in various lighting control systems used in commercial buildings as well as factors affecting their performance [7], and the analysis of factors that influence a lighting control system’s energy performance [8]. When studies conducted on energy efficiency and economic usage of outdoor lighting are analyzed, the main areas of concern can be summarized as: innovative light design model frameworks to provide quality floodlight for outdoor lighting designs [9], integrated approaches for exterior lighting systems’ control and design based on formal graph-based models and methods [10], possible methods and recommendations regarding the factors influencing energy savings in street lighting [11], and major factors that needs to be taken into account for energy efficiency in outdoor lightings [12]. In studies where lighting and artificial intelligence are used together, the main focus is usually on lighting system designs [13] and energy management [14,15]. There are also studies in literature about using Power Line Communication technology [16–18] in lighting control systems. When all the relevant studies reviewed it was noted that there is a need for research focusing on an intelligent energy-efficient outdoor lighting control system that could contribute to the reduction of carbon dioxide emissions by using electric energy more conservation and that could also minimize the disadvantages presented by existing outdoor lighting control systems. There are three main methods for controlling the lamps in an outdoor lighting control system. These are lamp control based on level of illuminance (photocell), lamp control by using modular clocks and finally lamp control by motion sensors. However, none of these three methods are sufficient for energy savings. The disadvantages of these methods used in controlling the lamps can be listed as follows: • Typically high level of electric consumption, inability to set modular clocks and limited range of settings in photocells, • Potential risk in terms of safety, • Inability to interfere to the faults in the system, • Inability to generate reports for energy consumed, • Inflexibility of the construction, • Difficulty in controlling and monitoring from a specific centre. • Lack of a communications technology, • Inability to integrate into building management systems.

energy-efficient outdoor lighting control system were run in two separate environments based on a computer and a microcontroller. The developed system is the first expert system in which an expert system is used to control and monitor an outdoor lighting system as well as perform load estimate and fault diagnosis. The Expert System was written with programming languages within the Human Machine Interface (HMI) and the microcontroller’s own software development platforms. While the system created in the computer environment is called the Central Real-Time Expert System (CRTES), the system developed in the microcontroller environment is referred as the Expert System Based Intelligent Control Nodes (ESICN). This system composed of CRTES and ESICN systems is named as Intelligent Energy Efficient Outdoor Lighting Control System (IEEOLCS). Since the outdoor lighting area in CRTES was large, a fieldbus technology was utilized. The I/O modules connected over the Programmable Controller (PLC) and the fieldbus were the interfaces where required input and output connections were made for running the CRTES. The fieldbus (Bit Serial Fieldbuses − Fieldbus) is the generic name for industrial communication networks. The purposes of four functions implemented in CRTES are as follows: • To contribute to the reduction of carbon dioxide emissions by using electrical energy more conservation and daylight more efficiently, • To control the lamps through with a phase angle technique (30% to 100% with triac) in the most appropriate time slot based on HMI, manual, daylight, real-time, the usage of university campus in specific time periods of the year (education semester, weekend and national holiday operation modes), the security level and motion, • To monitor whether the lamp groups are activated, • To diagnose the faults in the power lines the lamp groups are connected, • To diagnose the faults in the individual lamps of the lamp group, • To estimate the load connected to a specific phase. In the structure of IEEOLCS, in case of a fault on a fieldbus technology with a central control structure or deactivated modules on the fieldbus, CRTES functions cannot be performed in the block or blocks where the signal is cut off. In this case, to ensure continuity in control of the outdoor lighting, microcontroller based circuits were placed in each block of the buildings. By writing the ES rules to the microcontroller, expert system based intelligent control nodes were installed that control the outdoor lighting independent of the centre. The purposes of ESICN system are: • To contribute to the reduction of carbon dioxide emissions by using electrical energy more conservation and daylight more efficiently, • To control the lamps manually in the most appropriate time slot based on the real-time and motion. The developed IEEOLCS was implemented in a 5-blocks building in Marmara University’s Goztepe Campus. 2. Materials and methods

In this study, one of the artificial intelligence techniques, an Expert System (ES) was selected and used in various functions of the outdoor lighting to convert an intelligent system. This also proves the feasibility of artificial intelligence in different functions of outdoor lighting. The functions of the developed intelligent

2.1. IEEOLCS setup This research was implemented in the exterior lighting system of the buildings of Marmara University, Faculty of Technical Edu-

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

775

Fig. 1. Building Layout Plans for Marmara University, Faculty of Technical Education and Vocational School of Technical Sciences where IEEOLCS was implemented.

Fig. 2. Field positioning of all system elements in buildings of TEF and TBMYO.

cation (TEF) and Vocational School of Technical Sciences (TBMYO) in Istanbul. The layout of the buildings can be seen in Fig. 1. In the initial stage of system installation, the project of buildings’ existing outdoor lighting system was drawn. Then, the power consumption of outdoor lighting system was recorded for a year until the subsequent stages were completed. In the meantime, the input/output elements as well as the variables that will be used in IEEOLCS were identified. These include digital inputs, digital signals from motion and photo-electric sensors, analog voltage (0. . .10 V DC) generated from photovoltaic panel used to determine the illuminance of the ambient environment, real-time data, digital data for faults across the power lines the lamps are connected, analog current data for lamp faults (0. . .20 mA) and analog outputs of PLC for phase angled control of the lamps (4. . .20 mA) To be able to adjust the luminous flux of the lamps, a lamp dimmer circuit was developed using triac. The circuit controls the phase angle between 30% and 100% depending on the signal from 4 − 20 mA analog outputs of the PLC. The lamps were then classified into three groups taking daylight, real-time, security level and usage of the university campus during certain periods of the year into account. Group I: The lamp group that is activated during the time period called the most appropriate time slot in which lamps are driven by CRTES or ESICN by adjusting from 50% to 100% with phase control is called CTS (Continuous) lamp group. Group II: The lamp group that is activated by adjusting from 50% to 100% with phase control until 23:00, and from 30% to 100% with phase control after 23:00 is called 23-lamp group. Group III: Finally, the lamp group that is activated when a motion is detected by adjusting from 50% to 100% with phase control is called MOTION lamp group. A fieldbus system was selected through analyzing the number of inputs/outputs identified in IEEOLCS, the area of the field and the cost analysis. A fieldbus standard called Universal Remote Input/Output (URIO) was used. RIO has a central control structure.

The shortcoming of this system is that whole system pauses when the master device stops working. In the developed system, this disadvantage of the RIO standard was overcome by installing ESICN. Then, PLC and distributed I/Os were assembled on the fieldbus line. As PLC, SLC 500/03 coded device from Allen-Bradley was selected. For slave and distributed I/Os compatible with the master device in RIO standard, Flex modules from Allen-Bradley were chosen. Based on these choices, the distribution of the elements used in the IEEOLCS implementation field is shown in Fig. 2. The master device in the RIO standard with a central control structure was placed in block D whereas the modules used as slave and distributed I/Os were installed to the other blocks. Afterwards, the motion sensors to detect movements, the photo-electric sensors and the photovoltaic panel to measure the level of change of illuminance of the ambient light were assembled by choosing the most suitable locations. The mounting point of the photovoltaic panel can be seen in Fig. 3. If the fieldbus is disabled or there is a fault on it, CRTES functions cannot run at points where fieldbus signals are cut off. In this case, lighting control executes the ESICN functions with a microcontroller based circuit. Fig. 4 shows the locations of microcontroller board (1), power board (2) and the fault board (3) on the panel board. The programmable controller, distributed I/Os, the master and slave modules of RIO standard, the microcontroller board, fault board, power board and other hardware are placed on five separate panel boards in a suitable manner. Panel board connections of two blocks as examples can be seen in Figs. 5 and 6 [19]. CRTES and ESICN functions were written with software packages. Then the simulation phase of IEEOLCS was initiated. At this stage, all rules of ES designed in both computer and microcontroller environments were tested. After achieving ideal results during the simulations, control panels were assembled on appropriate locations in five buildings. Fig. 7 shows the setup used during the simulation. The fieldbus system in IEEOLCS, intelligent control

776

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

Fig. 3. Assembly point of photovoltaic panel.

Fig. 6. Slave Unit in RIO.

Fig. 4. Locations of microcontroller, power and fault boards.

Fig. 7. The simulation phase of IEEOLCS.

nodes, and the distribution of the devices in the blocks can be seen in Fig. 8 while Fig. 9 shows the general block diagram of IEEOLCS. 2.2. Research tools 2.2.1. Programmable controller PLC works both as the software and hardware interface of CRTES. As the processor module of programmable controller, SLC 500/03 module from Allen-Bradley was used.

Fig. 5. Master Unit in RIO.

2.2.2. Slave In fieldbus systems, there is a slave unit running the operations on modules connected to it based on the commands from

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

777

Fig. 8. Distribution of all system elements in IEEOLCS.

Fig. 9. The general block diagram of IEEOLCS.

the master unit that manages the communications in the central control structure. The RIO standard has a central control structure. At this point, a 1794-ASB RIO adapter from Allen-Bradley was used to enable communications with master 1747-SN RIO scanner in RIO standard.

2.2.3. Distributed I/O These are the input/output modules connected to the adapter module. The input and output elements in the field were connected to these modules. As input/output modules, Allen-Bradley’s products with codes 1793-IB4, 1793-OW4, 1793OE2 and 1793-IE4 were used.

778

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

2.2.4. Current transformer In order to diagnose lamp faults in CRTES with ES, the current of the lamp groups need to be measured. To reduce the phase current of block A’s CTS lamp group to a level that allows connection to the current transducer, a current transformer was used. The current transformer had a primary current of 10A, secondary current of 5A with 10VA and 0.5 class characteristics. 2.2.5. Current transducer A transducer transforms the electrical signals on its input into standard analog signals. The current transducer in this system converts the current transformer output of phase current for block A’s CTS lamp group in CRTES into standard 0–20 mA analog signal, while the distributed I/Os are connected to the appropriate analog input modules.

• Controlling the lamps according to the information from 7 different inputs. These are: - Manual, - Objects created on the screen of the HMI software, - Usage of university campus at specific time periods of the year (education semester, weekend and national holiday operation modes), - Level of security - Motions, - Real-time, - The level of luminance obtained by photovoltaic panel. The purpose of RTCOM is to control the lamps with phase angle control technique (30% to 100% with triac) at the most appropriate time slot based on 7 different information.

2.2.6. Photovoltaic panel In CRTES, this is used to measure the illuminance in the ambient environment. The analog voltage generated by the photovoltaic panel based on the daylight is then applied to the relevant channel input of the analog input module. Based on the level of change in illuminance, the lamp groups are either activated or deactivated. The maximum output voltage of the solar cell is 10 V and its maximum current is 160 mA.

• Interpreting the information received from the control system, and monitoring whether the lamps are active

2.3. Central real-Time expert system (CRTES)

• Activation and deactivation of the lamp group based on information from 7 separate inputs. • Following the activation of the lamps, a safe time slot was identified to prevent accidental deactivation of the lamps due to, for various reasons, an artificial light reflecting on the photovoltaic panel measuring the illuminance. In this time slot, the illuminance was disregarded. • Activation and deactivation of the lamp groups in the safe time slot. • Activation and deactivation of the lamp group based on the illuminance outside the safe time slot • Monitoring whether the lamp groups are activated. • Issuing an alarm for lamp groups when there is a mismatch between the real-time and illuminance. • Controlling the lamps with phase angle control technique (30% to 100% with triac). • Continuous activation of all the lamp groups in case of a security risk.

Within the scope of CRTES, ES were developed in four separate areas. These areas are control, monitoring, fault diagnosis and load estimate. The functions of these are given below: • • • •

Real-time control and monitoring, Real-time lamp fault diagnosis, Real-time fault diagnosis of power line, Real-time load estimate.

An expert system is composed of two main components. These are inference engine and knowledge base. The knowledge base for this study was created in three steps in CRTES. These are [20]: • Description of each field, • Obtaining the knowledge for each of these fields, • Selecting the knowledge representation model at the fields. The knowledge and experience of the expert person, mathematical models, users and the events experienced were used as the source of the knowledge. While creating the CRTES functions, one of the knowledge representation models, production rules model was used to form the objects and identify the relationship between these objects. This model was implemented with If· · ·Then rules. Forward chaining method was then used for the inference by reviewing the existing facts in the working memory of the ES structure for all functions contained in the CRTES and the If· · ·Then rules in the knowledge base. PLC supports both the interface and outdoor lighting control functions for CRTES. Reading and writing the data in the field was performed through PLC and distributed I/O’s input/output and analog modules. 2.3.1. Real-Time control and monitoring (RTCOM) ES’ block diagram designed for Real-Time Control and Monitoring of the outdoor lighting within the context of CRTES is shown in Fig. 10. RTCOM function has two main objectives. These are:

2.3.1.1. Description of the field in RTCOM knowledge base. In describing the knowledge base of ES for RTCOM, the following stages were used: 2.3.1.1.1. Identifying the characteristics of RTCOM problem. The main topics can be summarized as follows:

2.3.1.1.2. Identifying the objects in RTCOM problem. The objects identified according to the characteristics of the problem: • The real-time (month, day, hour, minute), • The analog voltage generated by the photovoltaic panel that measures the change in illuminance of the ambient environment (0–10 V DC). • Analog outputs of PLC for phase control of the lamps (4–20 mA). • PLC lamp group’s outputs (Block A: CTS, MOTION, 23 − Block B: CTS, MOTION, 23 − Block C: CTS, MOTION, 23 − Block D: CTS − Block E: CTS, 23/MOTION). • Analog current data of phase current for block A’s CTS lamp group (0–20 mA). • Power line fault inputs at PLC (Block A: phases L1, L2, L3 − Block B: phases L1, L2, L3 − Block C: phases L1, L2, L3 − Block D: phases L1, L2, L3 − Block E: phases L1, L2). • Usage of university campus at specific time periods of the year (education semester, weekend and national holiday operation modes). • Level of security.

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

779

Fig. 10. Block diagram of real-time control and monitoring expert system.

2.3.1.1.3. Identifying the relations between RTCOM objects. Generating the necessary control signal to activate and deactivate the lamp groups at appropriate value range of real-time and illuminance: The analog voltage generated from the photovoltaic panel that measures the change in illuminance in the ambient environment was compared to the predetermined reference values. The reference value at the appropriate time slot is 1.5 V. This value corresponds to an illuminance of 4.5 lx. If the ambient illuminance is below the reference value during the delay time and the appropriate real-time slot for the lamp group has been reached, the lamp groups (23-lamp group and CTS lamp group) are activated. The lamps are driven with phase angle control technique (30% to 100% with triac). When the illuminance reaches the appropriate value (after safe time slot) or after the appropriate real-time slot, all lamp groups are deactivated. Generating the necessary control signal to activate and deactivate the lamp groups at the reference value range of the ambient environment’s illuminance outside the appropriate real-time slot: The lamp groups can be activated outside the appropriate real time slot if the ambient light fades suddenly. In order to ensure this, the measured illuminance was compared to the identified reference values. The reference value range outside the appropriate time slot is 0.1 V (1 lx) = < analogvalue < = 1V(2.74Lux). If the illuminance is within the reference values during the delay time, lamp groups (23-lamp group and CTS lamp group) are activated. The lamps are driven with phase angle control technique (30% to 100% with triac). All lamp groups are deactivated outside the appropriate real-time slot or when the illuminance reaches the appropriate value (after the safe time slot). Following the activation of the lamps, disregarding the illuminance at certain time slots (safe time slot) to prevent deactivation of the lamps due to, for various reasons, an artificial light reflecting on the photovoltaic panel: At this time slot, the lamp groups are triggered according to the real-time. Whether the lamp groups are activated was monitored by recording the outputs of the lamp groups and the fault inputs of the power lines. Specifically the phase current of block A’s CTS lamp group was monitored. If the ambient illuminance is lower than the reference value of 1 lx before reaching the appropriate real-time slot in which the lamp groups will be activated or if the illuminance is higher than the reference value of 7 lx after reaching the appropriate real-time slot in which the lamp groups will be activated, then warnings are issued.

In the study, three different operation situations were specified as education semester, weekend and national holiday operation modes based on the user preference for 23-lamp group according to the rate of usage of the university campus at certain time periods. If the education semester operation mode is selected, 23-lamp group is activated by adjusting from 50% to 100% with phase control until 23:00, and from 30% to 100% with phase control after 23:00. If the weekend operation mode is selected, the 23-lamp group, in line with the campus’ usage rate, is not activated on Sundays. If national holiday or summer term operation method is selected, 23-lamp group is activated from 30% to 100% with continuous phase control since the campus has a low rate of usage.

2.3.1.2. ES variables and inference engine variables for RTCOM. For the ES rule-base which is with the objects identified in the RTCOM problem, a total of 85 variables was used in the If part, and in the Then part, 43 variables were used for the inference engine where numerical values and descriptive sentences were assigned. The ES’ rule base designed with If· · ·Then rule for RTCOM, has 213 rules The following paragraphs give the variables contained in the “if” and “then” part of 104 rules where control signal is generated depending on real-time and the daylight and the monitor function of the lamps only. 41 variables in the “if” part and 21 variables for the inference engine in the “then” part of the 104 rules are listed below. In the variables list of the If part, the italics refer to the variables; the statements in the parenthesis represent the variables in Figs. 11 and 12 where only the 104 rules of RTCOM’s total 213 rule bases are shown in the form of a network while other statements are brief descriptions of the variables. • Month: (G1), DayOfMonth: (G2), Hour: (G3), Minute: (G4), • TIME1 is the appropriate real time slot during the first fourmonths period; TIME2 is the appropriate real-time slot for the second four-months period; TIME3 is the appropriate real-time slot for the third four-months period; TIME is the appropriate real-time slot in a twelve-months time period; analog\N7 8 daylight is the analog voltage generated with the photovoltaic panel (G5), Safe TIME is the time slot during which the analog voltage generated with the photovoltaic panel is disregarded. • Block A’s lamp outputs are CTS, 23 and MOTION lamp outputs (respectively represented by output\O 1 2 0:(G6),

780

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

Fig. 11. Network structure of the monitoring part in the ES of RTCOM.

•

•

• •

•

•

•

• •

•

output\O 1 2 1:(G7), output\O 1 2 2:(G8)). Block A’s lamp drive outputs are 4–20 mA (analog\A A1:(G61), analog\A A2,:(G71)), Block B’s lamp outputs are CTS, 23 and MOTION lamp outputs (respectively represented by output\O 1 10 0:(G9), output\O 1 10 1:(G10), output\O 1 10 2:(G11)). Block B’s lamp drive outputs are 4–20 mA (analog\B A1:(G91), analog\B A2,:(G101)), Block C’s lamp outputs are CTS, 23 and MOTION lamp outputs (respectively represented by output\O 1 14 0:(G12), output\O 1 14 1:(G13), output\O 1 14 2:(G14)). Block C’s lamp drive outputs are 4–20 mA (analog\C A1:(G121), analog\C A2,:(G131)), Block D’s lamp output is CTS lamp output (output\O 3 0 0: (G15)). Block E’s lamp outputs are CTS, 23/MOTION lamp outputs (respectively represented by output\O 1 6 0:(G16), output\O 1 6 1:(G17)). Block E’s lamp drive outputs are 4–20 mA (analog\E A1:(G161), analog\E A2,:(G171)). Fault information on the phase where block A’s CTS, 23 and MOTION lamp groups are connected (respectively represented by ariza\B3 6 12:(G18), ariza\B3 6 13:(G19), ariza\B3 6 14:(G20)). Fault information on the phase where block B’s CTS, 23 and MOTION lamp groups are connected (respectively represented by ariza\B3 6 15:(G21), ariza\B3 7 0:(G22), ariza\B3 7 1:(G23)). Fault information on the phase where block C’s CTS, 23 and MOTION lamp groups are connected (respectively represented by ariza\B3 7 2:(G24), ariza\B3 7 3:(G25), ariza\B3 7 4:(G26)). Fault information on the phase where block D’s CTS lamp group is connected (ariza\B3 6 9:(G27)). Fault information on the phase where block E’s CTS and 23/MOTION lamp groups are connected (respectively represented by ariza\B3 7 5:(G28), ariza\B3 7 6: (G29)). Current value of block A’s CTS lamp group (current\F8 6:(G30)), second: (G31), the initial delay time for triggering the lamp groups (delay1:(G32)), second delay time for triggering the lamp groups (delay2: (G33)).

Below are the inference engine’s variables where numerical values and descriptive sentences are assigned from the then part of 104 rules out of RTCOM’s total 213 rule-bases. The italics refer to the variables of the inference engine; the statements in the parenthesis represent the variables of inference engine in Fig. 11 and Fig. 12 where only the 104 rules are shown in the form of a net-

work while other statements are brief descriptions of the inference engine’s variables. • Control signal based on the real-time and daylight (EXPERT L CONTROL:(C6)), Information about the warning given in case of a mismatch between the real-time and illuminance (TimeAndDaylight:(C7) • The appropriate real time in the first four-months period (TIME1:(C1)), the appropriate real time in the second fourmonths period (TIME2:(C2)), the appropriate real time in the third four-months period (TIME3:(C3)), the appropriate real time in the twelve-months period (TIME:(C4)), the real time in which illuminance is disregarded (Safe TIME:(C5)). • Monitoring whether Block A’s CTS, 23 and MOTION lamp groups are activated through the expert system’s screen (A CTS Lamp:(C8), A 23 Lamp:(C9), A MOTION Lamp:(C10)). • Monitoring whether Block B’s CTS, 23 and MOTION lamp groups are activated through the expert system’s screen (B CTS Lamp:(C11), B 23 Lamp:(C12), B MOTION Lamp:(C13)). • Monitoring whether Block C’s CTS, 23 and MOTION lamp groups are activated through the expert system’s screen (C CTS Lamp:(C14), C 23 Lamp:(C15), C MOTION Lamp:(C16)). • Monitoring whether Block D’s CTS lamp group is activated through the expert system’s screen (D CTS Lamp:(C17)). • Monitoring whether Block E’s CTS and 23/MOTION lamp groups are activated through the expert system’s screen (E CTS Lamp:(C18), E 23/MOTION Lamp:(C19)). • The initial delay time for triggering the lamp groups (delay1:(C20)), second delay time for triggering the lamp groups (delay2:(C21)). Fig. 11 shows the 14 rules that allow monitoring the lamp groups in the ES rule base of RTCOM in the form of a network. Fig. 12 on the other hand shows 90 rules that allow generating the control signal based on the real-time and daylight only for the 213 rules written for RTCOM in the form of a network. Table 1 gives the certainty factors (CF) of the information shown on RTCOM ES screen. 2.3.2. Real-Time fault diagnosis of lamp (RTFD-L) In an outdoor lighting system, the purpose of RTFD-L function is to determine whether the lamp groups are activated by interpreting the phase current value of Block A’s lamp group in real time, and to

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

Fig. 12. Network structure of the control part based on the real-time and daylight only in the ES.

781

782

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

Table 1 Certainty Factors in RTCOM ES screen. Inference engine

CF value of Rule’s If part

CF value of Rule’s result part

Rule’s CF value

Inference7

Time = 0.8 N7 8 daylight = 0.6

Inference8

Lamp group’s output = 0.8 Current Transformer measurement = 0.6 Fault phase input = 0.51 Lamp group’s output = 0.8 Fault phase input = 0.51 Lamp group’s output = 0.8 Fault phase input = 0.51 Lamp group’s output = 0.8 Fault phase input = 0.51 Lamp group’s output = 0.8 Fault phase input = 0.51

Time not appropriate, very low daylight = 0.9 Time appropriate, very high daylight = 0.7 Photovoltaic panel working normal = 0.9 Analog output = 0.8

0.6 × 0.9 0.6 × 0.7 0.6 × 0.9 0.51 × 0.8

Analog output = 0.8

0.51 × 0.8

Analog output = 0.8

0.51 × 0.8

Analog output = 0.8

0.51 × 0.8

Analog output = 0.8

0.51 × 0.8

Inference11 Inference14 Inference17 Inference18

identify lamp faults as well as generate the descriptions of these. In order to achieve this purpose, PLC performs the interface function for the ES.

D, E blocks are connected and to generate necessary descriptions. In order to achieve this purpose, PLC performs the interface function for the ES.

2.3.2.1. Description of the field in RTFD-L knowledge base. In describing the knowledge base of ES for RTFD-L, the following stages were used: 2.3.2.1.1. Identifying the characteristics of RTFD-L problem. The main topics can be summarized as follows:

2.3.3.1. Description of the field in RTFD-PL knowledge base. In describing the knowledge base of ES for RTFD-PL, the following stages were used: 2.3.3.1.1. Identifying the characteristics of RTFD-PL problem. The main topics can be summarized as follows:

• • • •

Determining whether the Block A’s CTS lamp group is activated. Using the data on lamp power obtained from the ES. Identifying the number of lamps in Block A’s CTS lamp group. Measuring the real-time phase current value of Block A’s CTS lamp group in order to identify the faults in the lamp. • Identifying the limit values of the phase current. Taking into account the heating factor and the differences in the phase current values according to the changes in the phase voltage for activation of the lamps during the initial start-up phase, • Determining whether a fault exists in Block A’s CTS lamp group and generating the description information.

• Identifying the single phase faults in L1, L2, L3 phases where the lamps in A, B, C, D, E blocks are connected. • Identifying the dual phase faults in L1, L2, L3 phases where the lamps in the blocks are connected. • Identifying the three-phase faults in L1, L2, L3 phases where the lamps in the blocks are connected. • Taking the maintenance condition on the power line into consideration.

2.3.2.1.2. Identifying the objects in RTFD-L problem. The objects identified according to the characteristics of the problem:

• A phase signal for each of the L1, L2, L3 (L1, L2 in block E) phases where the lamps are connected.

• There are three luminaires in the Block A’s CTS lamp group. The power of lamps installed to these luminaires. • The number of lamps in Block A’s CTS lamp group. • The output of Block A’s CTS lamp group. • Real-time analog current data of phase current for block A’s CTS lamp group (0–20 mA). • Based on the power of lamps, the current values of Lamp1 + Lamp2 + Lamp3, Lamp1, Lamp2, Lamp3, Lamp1 + Lamp2, Lamp1 + Lamp3, Lamp2 + Lamp3.

2.3.3.1.3. Identifying the relations between RTFD-PL objects. The location of the fault was identified and necessary descriptions were generated by examining the relations between the phase signals on L1, L2, L3 (L1, L2 in Block E) phases where lamps were connected in each block.

2.3.2.2. ES variables and inference engine variables for RTFD-L. For the ES rule base which was written with the objects identified in the RTFD-L problem, a total of 16 variables was identified in the If part, and in the Then part, 2 variables were identified for the inference engine where numerical values and descriptive sentences were assigned. For RTFD-L, the ES’ rule base designed with If· · ·Then rule has 19 rules. 2.3.3. Real-Time fault diagnosis of power line (RTFD-PL) The main purpose of the RTFD-PL function in an outdoor lighting system is to determine whether there is a real-time power line fault in the L1, L2, L3 (L1, L2 in Block E) phases where the lamps in A, B, C,

2.3.3.1.2. Identifying the objects in RTFD-PL problem. The object identified according to the characteristics of the problem:

2.3.3.2. ES variables and inference engine variables for RTFD-PL. For the ES rule base, written with the objects identified in the RTFD-PL problem, a total of 14 variables were identified in the If part, and in the Then part, 5 variables were identified for the inference engine where numerical values and descriptive sentences were assigned. For RTFD-PL, the ES’ rule base, designed with If· · ·Then rule, was created with 36 rules. 2.3.4. Real-Time load estimate (RTLE) In the outdoor lighting system, the main purpose of RTLE function was to estimate the power value of lamps connected to the Block A’s CTS lamp group. Two methods were used for estimation. These are: • Presenting the estimate values to the user based on the values of real-time phase current, • Selection from a list by the user.

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

In order to achieve this purpose, the screen designed for PLC and RTLE performs the interface function for the ES. 2.3.4.1. Description of the field in RTLE knowledge base. In describing the knowledge base of ES for RTLE, the following stages were used: 2.3.4.1.1. Identifying the characteristics of RTLE problem. The main topics can be summarized as follows: • Active participation of the user to RTLE ES. • Assessing the powers of 300W, 500W and 750W halogen lamps used in outdoor lighting as estimated lamp powers (halogen lamps were selected due to their color rendering index close to 100%). • Leaving the selection of estimation method for power levels of the lamps connected to the block A’s CTS lamp group to the user. • Ability of ES to offer the user the forecasted powers between 300W and 2300W for load estimates according to the real-time phase current values. • In order to present the user options to use in load estimate, using the real-time phase current of Block A’s CTS lamp group. • Taking into account the differences in the phase current values according to the changes in the phase voltage for activation of the lamps during the initial start-up phase, • In user selected load estimate, choosing the lamp powers from a list without considering the phase current. • Identifying the largest possible load that could be connected to the power line by evaluating the current transformer’s primary current connected to the block A’s CTS lamp group and generating the necessary warnings. 2.3.4.1.2. Identifying the objects in RTLE problem. The objects identified according to the characteristics of the problem: • For the block A’s CTS lamp group, the real-time phase current’s analog current data. • Estimate values; lamp powers up to four separate options according to the current the lamps draw for real-time load estimate. • In real-time load estimate, lamp powers selected either from estimated values or from a list by the user. • Confirmation information that allows selecting the first choice directly in real-time load estimate.

783

2.3.4.2. ES variables and inference engine variables for RTLE. For the ES rule base, written with the objects identified in the RTLE problem, a total of 9 variables were identified in the If part, and in the Then part, 15 variables were identified for the inference engine where numerical values and descriptive sentences were assigned. For RTLE, the ES’ rule base, designed with If· · ·Then rule, has 138 rules. 2.4. Expert system based intelligent control node (ESICN) In a CRTES, a fault in PLC causes deactivation of the master unit in RIO communication; interruption of the communication between the master and adapter units, or an adapter fault causes the system to enter into a fault condition in area or areas where communication is cut off. With CRTES, fault area or areas do not control the lighting. To use with RIO system as well as to ensure continuity in IEEOLCS regionally during the fault condition, a microcontroller board was installed both to the central node and at each node where adapters are located. Then an ES based microcontroller was developed by writing an ES into the microcontroller program memory that would allow activation and deactivation of the lamp groups at the most appropriate time based on the real-time and motion, i.e. that could keep the energy consumption at the optimum level. In other words, an Expert System Based Intelligent Control Node was created. ES based microcontroller contributes to the reduction or carbon dioxide emissions by using electrical energy more conservation and daylight more efficiently. In ESICN, control inputs, motion and photo-electric sensors were connected to the input ports of the microcontroller to activate and deactivate the lamps regionally. Furthermore, an output signal from the output module at the PLC and distributed I/Os was connected to the microcontroller’s input as a communications signal. This digital signal is the output signal for the module and input signal for the microcontroller. When the output signal of the module is cut off, microcontroller assumes all the tasks involved to ensure continuity in IEEOLCS at the corresponding block and performs intelligent control with ES in the program memory. The outputs of the microcontroller are used to activate and deactivate the lamp groups. A block diagram of the simple hardware setup for ES based microcontroller is shown in Fig. 13. For forming the objects and identifying the relations between these in the ES function at the Intelligent Control Node, one of the

Fig. 13. ES based microcontroller’s simple hardware setup.

784

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

Fig. 14. Intelligent control node expert system’s block diagram.

knowledge representation models − the production rules model was used. This model is implemented with If· · ·Then rules. Intelligent Control Node had three different operating methods at five installation locations. By taking these methods into consideration, three separate ES were written. • ES designed for blocks D and B, • ES designed for blocks A and C, • ES designed for block E. ES’ block diagram designed for real-time control of the outdoor lighting at the intelligent control node is shown in Fig. 14. Fig. 8 shows the location of the Intelligent Control Node in the field device layout of IEEOLCS while Fig. 9 shows the location of ESICN in the general block diagram of IEEOLCS. 2.5. Illuminance measurement The photovoltaic panel was connected to the channel 0 input of the analog module in block A’s control panel board. From the moment the illuminance reads 0 lx, the voltage generated by the photovoltaic panel and the illuminance is measured and recorded simultaneously with a luxmeter in units of lux. These measurements were taken 10 times on different days and similar results were obtained. The rate of change between the measured analog voltage and the lux values is shown with a curve, and by applying curve fitting method to the obtained values, a mathematical model was extracted. The mathematical model is shown in Equation (1). Fig. 15 shows the real value curve and the curve obtained through curve fitting method. E = 0, 0994.v5gp − 0, 3998.v4gp + 0, 6346.v3gp + 0, 4473.v2gp +1, 1758.vgp + 0, 7899

(1)

vgp : Voltage generated by the photovoltaic panel (Volt) E : Illuminance (Lux) 3. Results The power consumption of the traditional outdoor lighting system was recorded with a smart electric meter for a year until the IEEOLCS installation phase was completed. The power consumption was again recorded for the second year after IEEOLCS was installed and RTCOM was activated. Fig. 16 shows the energy consumption of the outdoor lighting system.

Fig. 15. The change between illuminance and analog voltage.

Monthly measurements taken for a two-year period is given in Table 2. As can be seen in Fig. 16, RTCOM was deactivated between July 16 and 31, and ESICN system was taken into operation. The power consumption of the traditional outdoor lighting system and the ESICN system between July 16 and 31 is presented in Table 3. With the aid of the expert system, the energy consumption was kept at an appropriate level, and the activation status of the lamp groups was monitored through a window prepared for RTCOM on the CRTES screen. The active CRTES screen can be seen in Fig. 17. It is possible to monitor the lamp group’s status on RTCOM window. Fig. 17 shows block D’s active CTS lamp group. CFs calculated for CTS lamp groups at each block can be seen under the heading “status of lamp groups”. Since the information on current of block A’s CTS lamp groups was also taken into account, the CF value (0.46) is higher than the others. The relation between the real-time and daylight was also monitored. The statement: “photovoltaic panel works normal” in the RTCOM window in Fig. 17 shows that the real-time and the daylight are congruent with each other. The CF of this status can be

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

785

Fig. 16. Comparison of the energy consumption of the outdoor lighting system.

Fig. 17. The active CRTES screen.

seen in the upper right corner of the RTCOM window. In case of a mismatch between the time and the daylight, the following statements can be observed in the window: Although daylight is low, time is not appropriate and Although time is appropriate, daylight is high. The RTLE window to estimate the powers of lamps connected to the block A’s CTS lamp group in real-time, and the RTFD-L win-

dow used to fault diagnosis of the same lamp group in the screen prepared for CRTES can be seen in Fig. 17. The RTFD-PL window used to determine whether there is a power line fault in the L1, L2, L3 (L1, L2 in Block E) phases supplying to the lamps in A, B, C, D, E blocks in the screen prepared for CRTES can also be seen in Fig. 17.

786

S. Atis, N. Ekren / Energy and Buildings 130 (2016) 773–786

Table 2 Monthly energy consumption values of traditional lighting system and RTCOM system. Month

Energy consumption of Classical System (kWh)

Energy consumption of RTCOM (kWh)

January February March April May June July August September October November December TOTAL

5653,81 4854,78 4846,15 4162,31 3850,35 3504,84 1804,96 4141,09 4485,66 5189,30 5451,42 5871,01 53815,68

3826,03 3285,68 3199,22 2801,95 2598,11 2375,05 1196,73 2779,67 3017,57 3423,57 3652,04 3902,99 36058,61

Table 3 Energy consumption values of traditional lighting system and RTCOM system between July 16 and 31. July(July 16–31)

16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 TOTAL

Energy consumption of Classical System (kWh) 120,03 120,24 120,65 121,01 121,58 123,00 119,98 120,98 122,74 123,13 123,60 123,99 124,30 124,91 125,24 124,82 1960,20

Energy consumption of ESICN System (kWh) 82,32 81,98 83,17 84,25 83,47 85,86 85,96 84,00 84,34 86,50 87,10 86,90 87,55 87,97 86,01 86,15 1363,53

4. Conclusions In this study, ES which is run in real time in two separate environments were developed in order to eliminate the disadvantages of existing lighting control systems. These ES are called CRTES and ESICN. The developed system is the first expert system in which an expert system is used to control and monitor an outdoor lighting system as well as perform load estimate and fault diagnosis. In RTCOM, one of the four separate functions of CRTES, the lamp groups were activated and deactivated at the most appropriate time slot depending on seven different conditions. This practice ensured savings around 33% in energy consumption. This rate proves that in comparison to a traditional outdoor lighting system, RTCOM contributes to the reduction of carbon dioxide emissions by using electrical energy more conservation and daylight more efficiently. On the other hand, in the ESICN system, the lamp groups were activated and deactivated at the most appropriate time slots according to the ES. This achieved energy savings around 30%. Such a saving indicates that ESICN system can also contribute to the reduction of carbon dioxide emissions. In CRTES, the user plays an active role in the system. Since CRTES runs on a computer platform, it offers wide possibilities in recording and processing the data in addition to changing and adding new rules to the rule base of ES. It is also possible to interfere to the system through internet when necessary. The ESICN developed can

meet these demands to a certain extend. In terms of installation costs, ESICN system provides a much more cost effective alternative to CRTES. At a time when energy is extremely valuable, more studies that focus on using outdoor lighting efficiently are needed. In this context, this study plays a significant role. The developed intelligent energy-efficient outdoor lighting control system is a system that could be integrated into green buildings and intelligent building management systems since it uses daylight more efficiently and electrical energy more conservation, contributes to the reduction of carbon dioxide emissions by using energy more conservation, allows timely response to the faults, offers a flexible architecture, and offers the possibility to control and monitor the system through Ethernet. Acknowledgement This research has been supported by Marmara University, Directorate of Scientific Research Projects. Project Number: FEN035/030203. References [1] Ministry of Energy and Natural Resources, “Energy Efficiency Strategy Paper.” Available: http://www.eie.gov.tr/verimlilik/document/Energy Efficiency Strategy Paper.pdf. [Accessed: 05.06.16]. [2] International Energy Agency, Available from: http://www.iea.org/statistics/statisticssearch/. [Accessed: 14.05.6]. [3] International Energy Agency, Light’s Labour’s Lost. Available from: https:// www.iea.org/publications/freepublications/publication/light2006.pdf. [Accessed: 12-May-2016]. [4] W.R. Ryckaert, C. Lootens, J. Geldof, P. Hanselaer, Criteria for energy efficient lighting in buildings, Energy Build. 42 (2010) 341–347. [5] L. Wang, H. Li, X. Zou, X. Shen, Lighting system design based on a sensor network for energy savings in large industrial buildings, Energy Build. 105 (2015) 226–235. [6] Z. Nagy, F.Y. Yong, M. Frei, A. Schlueter, Occupant centered lighting control for comfort and energy efficient building operation, Energy Build. 94 (2015) 100–108. [7] M.A. ul Haq, M.Y. Hassan, H. Abdullah, H.A. Rahman, M.P. Abdullah, F. Hussin, D.M. Said, A review on lighting control technologies in commercial buildings, their performance and affecting factors, Renew. Sustain. Energy Rev. 33 (May 2014) 268–279. [8] L. Bellia, F. Fragliasso, A. Pedace, Lighting control systems: factors affecting energy savings ’ evaluation, Energy Procedia 78 (2015) 2645–2650. [9] N. Das, N. Pal, S.K. Pradip, Economic cost analysis of LED over HPS flood lights for an efficient exterior lighting design using solar PV, Build. Environ 89 (2015) 380–392. [10] I. Wojnicki, S. Ernst, L. Kotulski, A. Se, Advanced street lighting control, Expert Syst. Appl. 41 (no. 4) (2014) 999–1005. [11] M. Kostic, L. Djokic, Recommendations for energy efficient and visually acceptable street lighting, Energy 34 (10) (Oct. 2009) 1565–1572. [12] S. Axel, Energy efficiency measures for outdoor lighting, Light Eng. 19 (no. 1) (2011) 15–19. [13] H. Rocha, I.S. Peretta, G. Flávio, M. Lima, L.G. Marques, Exterior lighting computer-automated design based on multi-criteria parallel evolutionary algorithm: optimized designs for illumination quality and energy efficiency, Expert Syst. Appl. 45 (2016) 208–222. [14] D.J. Fonseca, K.B. Bisen, K.C. Midkiff, G.P. Moynihan, An expert system for lighting energy management in public school facilities, Expert Syst. 23 (no. 4) (2006) 194–211. [15] H. Doukos, K. D.Patlitzianas, I. Konstantinos, J. Psarras, Intelligent building energy management system using rule sets, Build. Environ. 42 (2007) 3562–3569. [16] J.J. Lukkien, Exploring User-Centered Intelligent Road Lighting Design: A Road Map and Future Research Directions, 57, (no. (2)) (2011) 788–793. [17] C. Li, J. Wu, X. He, Realization of a general LED lighting system based on a novel power line communication technology, Appl. Power Electron. 230 (2010) 0–2304. [18] Q. Li-jun, S. Zi-zheng, J. Feng, Intelligent streetlight energy-saving system based on LonWorks power line communication technology, Electric Utility Deregulation and Restructuring and Power Technologies (DRPT), 2011 4th International Conference (2011) 663–667. [19] S. Atis, Control of Outdoor Lighting System by Expert System Supported Supervisory Control and Data Acquisition, Marmara University, 2007. [20] N. Allahverdi, Expert Systems. An Artificial Intelligence Application, Atlas Publishing House, Istanbul, 2002.