In this paper, we propose the Development of a knowledge-based expert system applied to Electricity distribution Network maintenance.
The expert system utilizes the Matlab® platform, using its proprietary Fuzzy Logic Toolbox. The inputs to the expert system are the different situations related to cost variables to and the implementation of one of the maintenance optimizing solutions pertaining to the Live Work-Redundancy-Simultaneity-Security model as defined by the maintenance unit during a the global life cycle of the equipment. The output of the system indicates the economic feasibility and benefit of the solution to be adopted of for our Maintenance model. Application of the expert system methodology is shown as a simulation.
The expert system may be of valuable assistance to utility engineers or asset managers in making strategic maintenance decisions such as “Economic solution”.
Open Peer Review Details | |||
---|---|---|---|
Manuscript submitted on 07-11-2016 |
Original Manuscript | Innovative Maintenance Model: Methodology of Decision Making and Global Economic Cost – Based Choice Using Fuzzy Logic Concepts |
In a very tense electricity market, characterized by strong daily constraints to permanently meet the needs of increasingly demanding customers, it has become difficult to predict the conditions of operating and maintenance of electrical network facilities in a deterministic and flexible way [1J. Ashayeri, A. Teelen, and W. Selenj, "A production and maintenance planning model for the process industry", Int. J. Prod. Res., vol. 34, pp. 3311-3326, 1996.
[http://dx.doi.org/10.1080/00207549608905092] ]. Among the tools elaborated in order to control these constraints, the manager of the Moroccan National Electricity network has decided to develop a new maintenance methodology, which makes it possible to achieve maximum levels in terms of facility availability and service continuity.
In fact, and given that maintenance is a cost-generating function, managers often sought to reduce costs rather than improve functioning mechanisms. However, the great network evolution and intensified growth in demand have resulted in remarkable changes [2Y. Jacquemart, J.-N. Marquet, and P. Pruvot, "A comprehensive analysis of mid- term voltage stability", IEEE Trans. Power Syst., vol. 10, pp. 1173-1182, 1995.
[http://dx.doi.org/10.1109/59.466539] ] which, for their part, have affected, inter alia, the maintenance function and made its role a central one. It has, thus, become a strategic function which guarantees the safety of the functioning of the network and the performance of its facilities.
In fact, this research work is about maintenance optimization according to the maintenance approach called Live Work – Redundancy – Simultaneous work – Security (LRSS) [3Z. Bouzoubaa, A. Soulhi, and J. El Alami, "Optimization of maintenance methods to improve the availability of the national electrical network", J. Innovat. Technol. Educat., vol. 3, pp. 23-35, 2016.
[http://dx.doi.org/10.12988/jite.2016.612] ].
We, thus, underline the specific characteristics with regard to the cost of the implementation of aspects of LRSS maintenance, which makes it possible to elaborate of an overall frame of reference for a successful implementation of a new maintenance approach within our company.
The present study concerns the facilities of the national electricity distribution network in Morocco.
Fig. (1) Components of the national distribution network. |
The current policy of maintenance of these facilities takes into account the optimal configuration of the HV network and guarantees good functioning of the national electricity distribution network in order to ensure safe and economical electricity distribution and provide high quality services. It also makes it possible to standardize maintenance methods and keep its efficient practices [4L.A. Wehenkel, Automatic Learning Techniques in Power Systems, Kluwer Academic Publishers, 1997.]. By the end of 2015, the studied network is spread out over 230 transformer stations Fig. (1) and 23300 km of HV lines.
Given the increasing demand in electrical energy, it appears that improving the availability of facilities and reducing maintenance cut offs have become a strategic and inevitable requirement, which directly affects the quality of services and the brand image of the company.
Currently, the tendency is to exploit the electricity distribution network as fully as possible. This is particularly because it is economically and technically difficult to install new facilities and because of the increasingly growing demand compared to the relatively limited capacity of the existing facilities.
In this increasingly restrictive and constraining context, the operator of the network should ensure the continuity of services, hence the necessity to control the efficiency of maintenance methods. Continuous improvement of the availability of facilities is, therefore, a permanent activity of the operator of the network. The stakes identified are related essentially to scheduling and cost of maintenance operations of the network. Such operations should be carried out while minimizing the duration of cut-offs and possibly eliminate them.
Besides, and despite the network being fitted out with a set of regulation controllers and devices, which make it possible to optimize its operating [5S. Henry, C. Lebrevelec, and Y. Schlumberger, "Defining operating rules against voltage collapse using a statistical approach: The EDF experience", In: Electric Power Engineering, 1999. PowerTech Budapest 99. International Conference, 1999, pp. 114-126.
[http://dx.doi.org/10.1109/PTC.1999.826461] ], the implementation of different maintenance schedules makes its behavior more complicated. It results in a successive disruptions due to frequent and extended stops of facilities. Thus, in order to address this issue, we have developed an appropriate methodology, called LRSS.
The present work is essentially a comparison of the costs of the implementation of the three solutions of this methodology (cf. logical flow chart below), based on the results of expert system and the principles of fuzzy logic (discussed in details below). In fact, the parameters impacting the costs of the said solutions can be summarized as follows:
Thus, consider the following value: C= C_Sol/C_End
If we take:
The economic feasibility and benefit of the solution to be adopted as one of the LRSS model solutions depends on the following conditions:
The methodology we have adopted is schematized in Fig. (2), taking into consideration the following definitions:
- A(i,k): Availability value equivalent to stop during operational Top(i,k)
The method includes the 4 phases described hereafter, of which the sequence is summarized in (Fig. 3).
The purpose of this phase is to design a set of plausible situations by enhancing cost variables related to the implementation of one of the maintenance optimizing solutions pertaining to the LRSS model. The choice of these variables is, therefore, determined by the purpose of the study and the scope of expected results and calls for engineering expertise and know-how [7S.D. Kaminaris, B.C. Papadias, and A.V. Machias, "Substation maintenance using fuzzy sets theory and expert system methodology", In: Athens Power Tech, APT 93, Joint International Power Conference, 1993, pp. 611-613.
[http://dx.doi.org/10.1109/APT.1993.673871] ].
Thus, any technical or economic variable pertaining to the modeling of our system can be potentially sampled, such as:
Fig. (2) Logical flow chart of LRSS model. |
Fig. (3) Main phases of the method. |
Each selected solution should be able to represent a preservation of an operating state that is sufficiently realistic and achievable, close to a typical state of continuous upkeep of facility operating. This requires finding out the facilities and operation modes concerned.
Moreover, for every selected solution:
It should be stressed that the availability of such a robust and efficient tool that has permitted the operational implementation of the approach should be elaborated.
If all scenarios have a realistic implementation point, operators of the electrical system will have at their disposal a primary database likely to result in a distinction between acceptable and unacceptable situations (See phase 4). For this purpose, they have numerous results, such as, among others, adjusted maintenance programs, flexibility of electricity distribution means, estimation of costs of optimization solutions (redundancy…).
The aim of this phase is to have a graphic profile of the examined cost parameters. The characteristics of the resulting curves and graphs vary depending on the input elements (input) of the expert system.
Concerning the IT modeling tool, we have opted for the use of the MATLAB software program [8 MATLAB User’s Guide, The MathWorks, Inc., Natick, MA, 1994-2016.], which proves to be perfectly compatible with the concepts of fuzzy logic as well as with the fuzzification methodology detailed below [9H.T. Nguyen, and M. Sugeno, Fuzzy Systems: Modeling and Control, Kluwer Academic Publisher: Massachusetts, 1998.].
Depending on the objectives of this study, decision-makers/ network [10A. Soulhi, S. Guedira, and N.-E. El Alami, "Decision-making automation fuzzy decision-making in industry", In: Proceedings of the 8th WSEAS International Conference on Artificial Intelligence, Knowledge Engineering & Data Bases, 2009, pp. 181-185.] operators should define precisely the optimal solution that they have selected to maximize the availability of facilities, through optimizing the overall cost of their maintenance. In fact, only by relying on a combination of these factors can make decision-makers and network operators decide an efficient asset policy [11A. Soulhi, S. Hayat, S. Hammadi, and P. Borne, "New strategy for the aid decision-making based on the fuzzy inferences in the traffic regulation of an urban bus network", In: Systems, Man, and Cybernetics, 1999. IEEE SMC'99 Conference Proceedings. 1999 IEEE International Conference, 1999, pp. 12-16.].
Representing and using knowledge are central to any relatively new scientific discipline - artificial intelligence [12A. Soulhi, "Artificial Intelligence Contribution to decision-making aid in Management of the urban collective transport systems", M.S. thesis, University of Lille 1, Villeneuve-d'Ascq, France, 2000.]. Until fairly recently, this discipline has had limited impact on industrial applications, for it has laid emphasis exclusively on the symbolic processing of knowledge as opposed to digital modeling traditionally used in engineering sciences.
In fact, Fuzzy logic [13S.D. Kaehler, Fuzzy logic – an introduction, part 2, Available at http://www.seattlerobotics.org/encoder/mar98/fuz/fl_part2.html, 2006.] is a problem-solving control system methodology that lends itself to the implementation in systems ranging from simple, small, embedded micro-controllers to large, networked, multi-channel PC or workstation-based data acquisition and control systems. It can be implemented in hardware, software, or a combination of both. Fuzzy logic provides a simple way to arrive at a definite conclusion based upon vague, ambiguous, imprecise, noisy, or missing input information. Fuzzy logic requires some numerical parameters in order to operate such as what is considered a significant error and significant rate-of-change-of-error, but exact values of these numbers are usually not critical unless very responsive performance is required in which case empirical tuning would determine them.
Fuzzy systems are part of knowledge-based systems or expert systems and their main objective is to implement a human know-how or linguistic rules through an IT program. Fuzzy logic [14A.B. Tee, M.D. Bowman, and K.C. Sinha, "Application of fuzzy logic to condition assessment of concrete slab bridges", Transport. Res. Record., pp. 22-30, 1988.] provides mathematical formalism for uncertain linguistic concepts.
The concept of fuzzy logic comes from the observation that the Boolean variable [15P. Borne, J. Rozinoer, J.Y. Dieulot, and L. Dubois, Introduction à la commande floue., Edition Technip: France, 1998.], which considers only two values (true or false), is inadequate and ill-adapted to the representation of most current phenomena.
Whereas, classical logic considers that a proposition is either true or false, fuzzy logic distinguishes infinity of truth values (between 0 and 1). It is then a generalization of binary logic to a multi-valued logic.
Fig. (4) Principle of fuzzy logic-based reasoning. |
Fuzzy logic –based reasoning can be schematically represented as follows (Fig. 4):
This method is based on the principle of degrees of membership from 0 to 1.
A function: 0 means not a member to a function, 1 means a member of the function and anything in between denotes partial memberships.
Membership functions can be defined using linguistic properties related to an human reasoning fast, high, and low…
The methods presented above have their own merits and each would have its applications suitable for specific reasons. Fuzzy logic is used in the expert system application presented in this paper due to:
The indicator modelling is mad by fuzzification. The fuzzification is the Numerical/Linguistic conversion of different variables characterizing the different indicators. The different indicators are presented by the same membership functions µ(x).
In this step, we proceed as follows:
This block indicates the relation between input variables (expressed as linguistic variables) and output variables (also expressed as linguistic variables) by means of intermediary rules.
The fuzzy inference allows us to develop a decision by using the decision rules. The decision rules are described by linguistic terms. For example:
In general, the description of a given situation, phenomenon or a process includes fuzzy terms, such as:
Instead of belonging to the set « True » or the set « false » of classical binary logic, fuzzy logic admits degrees of membership to a given set. The degree of membership to a fuzzy set is realized by a number between 0 and 1. A precise value of the membership function related to a variable value is referred to as µ and called « membership factor ».
In theory, membership functions can take any form. However, they are often defined by segments and are referred to as « piecewise linear », (widely used, for they are simple and include areas where the notion is true and areas where it is true, a fact which facilitates the collection of expertise).
Several values of linguistic variables are interrelated by rules that make it possible to draw conclusions.
Rules can be expressed in a general form:
Conditions can depend on various variables interrelated by operators OR or AND.
A simplified description of inferences can be obtained by means of a tabular representation, called inference matrix.
The fuzzification is the Linguistic/Numerical conversion of different variables characterising the global efficiency.
The method which is used here is the method of the centre of gravity. This method takes into account all available information.
In fact, Inference methods provide a membership function resulting from the output variable.
It is then fuzzy information. This fuzzy information should be transformed into a given value to be applied to the process control interface. Such transformation is called defuzzification.
The most frequently and widely used method of defuzzification is that of determining the gravity centre. Results of preceding calculation should be adapted to the control interface that is implemented.
For the purposes of our study, we have identified three parameters in order to summarize the status of performance of the studied system [17A. Soulhi, S. Hayat, S. Hammadi, and P. Borne, "Scheduling and decision-making in the traffic exploitation of the urban bus networks", In: Conference ISAS’99, 1999, pp. 8-12.].
We, thus, suggest that, for each parameter, the following modeling Fig. (5) should be adopted:
Fig. (5) Modeling functions. |
Cost of acquisition cost, maintenance and replacement of tooling and facilities makes it possible to carry out works in accordance with the concept of live work maintenance, in an interrupted way, and particularly live line work [6G. Gela, A.E. Lux, H. Kientz, D.A. Gillies, J.D. Mitchell, and P.F. Lyons, "Application of portable protective gaps for live work on compact 550 kV transmission lines", IEEE Trans. Power Deliv., vol. 11, pp. 1419-1429, 1996.
[http://dx.doi.org/10.1109/61.517500] ].
Cost of training (basic, updating, retraining courses) during the useful life of the facility.
In order to increase the overall availability of the facility, one of the LRSS model solutions involves implementing a redundant facility, which ensures the total takeover of the function of a unit scheduled to be stopped for maintenance.
This solution is possible provided the following conditions be obtained:
This cost includes the main items below:
Fig. (6) Live Work function. |
Fig. (7) Simultaneous work function. |
Given that works comprise various facilities, each with its own scheduled stops for maintenance, this situation results in frequent and repetitive stops, which leads to important disruptions in terms of provided services as well as unavailability during periods of peak electrical energy consumption.
Fig. (8) Redundancy function. |
In fact, in order to minimize the impact of the implementation and operation of two facilities on the overall availability of the system, LRSS model provides for a fusion of operation modes of the two facilities following specific rules.
The resulting costs are mainly those pertaining to accelerated depreciation of the performance and efficiency of facilities due to the delay in the execution of the operation modes recommended by manufacturers.
This function represents the analysis and decision making parameter, which will enable decision makers to decide on the technical and economic feasibility of the planned solution.
Fig. (9) LRSS Global cost function. |
Further on in the present article, we will examine the mechanisms of assessing the risk of overall cost, which constitutes a strategic lever in our approach.
Below is the correlation table Fig. (10) of the rules assigned to the fuzzy subsets:
Fig. (10) Inference rules assigned to retained functions. |
Following the modeling of the different functions and parameters in accordance with the principles of fuzzy logic and with the provisions of the software application employed, we have come up with the graphs of the studied variables correlation as well as their impact on the resulting parameter, namely LRSS’s GLOBAL COST.
The graphs, which highlight this strong dependence and interaction between the three levers chosen at the beginning can be presented as follows (Fig. 12).
Fig. (12) Surface viewers. |
It appears that this modeling constitutes a real tool box, including decision making support levers. In the diagram above, we notice that controlling the impact of the overall cost of the solutions implemented via LRSS model with regard to improving the system availability, at a value inferior to 50% (Fig. (13) hereafter), starts by observing maximum rates of input functions as follows:
Fig. (13) Rules viewers (MATLAB). |
Henceforth, these variables constitute input elements, of which the limit values should be in tune with the impact levels set by the management.
The main results of the simulations carried out, using this modelling via Matlab, can be summarized as follows:
It appears that the officer in charge of defining the new maintenance policy, based on the principle of our model, should consider the following results:
Face to the more and more constraining, uncertain and sometimes « surprising » operating conditions, which arise due to the higher demand for electricity and the increase in users’ requirements, operators of the electricity distribution network seek to regularly use more advanced and up-to-date techniques, methods and procedures able to face these hazards and continue to run the facilities of the network safely and provide high quality service. In fact, application of part of the expert system concerning the evaluation of maintenance cost is shown in this paper as an example. The output of the expert system can be utilized as a basis for utility asset managers to make maintenance decisions, thus improving maintenance quality and overall availability. A fuzzy logic based technique has been presented for the identification and classification of cost impact of innovative maintenance method. The proposed technique requires considering the condition to choose or not to apply one of methods defined in LRSS Maintenance Model. The concept is applied in a knowledge-based expert system that converts maintenance cost into a more objective and useful representation of overall Cost. Resorting to concepts of fuzzy logic has made it possible to establish decision making criteria with regard to the choice of investment based on an exhaustive assessment of potential solutions.
The authors confirm that this article content has no conflict of interest.
We would like to thank Prof. M. S. Sadik Maliki, Department of English Studies, Faculty of Letters and Humanities, Ain Chock, Hassan II University of Casablanca.
[1] | J. Ashayeri, A. Teelen, and W. Selenj, "A production and maintenance planning model for the process industry", Int. J. Prod. Res., vol. 34, pp. 3311-3326, 1996. [http://dx.doi.org/10.1080/00207549608905092] |
[2] | Y. Jacquemart, J.-N. Marquet, and P. Pruvot, "A comprehensive analysis of mid- term voltage stability", IEEE Trans. Power Syst., vol. 10, pp. 1173-1182, 1995. [http://dx.doi.org/10.1109/59.466539] |
[3] | Z. Bouzoubaa, A. Soulhi, and J. El Alami, "Optimization of maintenance methods to improve the availability of the national electrical network", J. Innovat. Technol. Educat., vol. 3, pp. 23-35, 2016. [http://dx.doi.org/10.12988/jite.2016.612] |
[4] | L.A. Wehenkel, Automatic Learning Techniques in Power Systems, Kluwer Academic Publishers, 1997. |
[5] | S. Henry, C. Lebrevelec, and Y. Schlumberger, "Defining operating rules against voltage collapse using a statistical approach: The EDF experience", In: Electric Power Engineering, 1999. PowerTech Budapest 99. International Conference, 1999, pp. 114-126. [http://dx.doi.org/10.1109/PTC.1999.826461] |
[6] | G. Gela, A.E. Lux, H. Kientz, D.A. Gillies, J.D. Mitchell, and P.F. Lyons, "Application of portable protective gaps for live work on compact 550 kV transmission lines", IEEE Trans. Power Deliv., vol. 11, pp. 1419-1429, 1996. [http://dx.doi.org/10.1109/61.517500] |
[7] | S.D. Kaminaris, B.C. Papadias, and A.V. Machias, "Substation maintenance using fuzzy sets theory and expert system methodology", In: Athens Power Tech, APT 93, Joint International Power Conference, 1993, pp. 611-613. [http://dx.doi.org/10.1109/APT.1993.673871] |
[8] | MATLAB User’s Guide, The MathWorks, Inc., Natick, MA, 1994-2016. |
[9] | H.T. Nguyen, and M. Sugeno, Fuzzy Systems: Modeling and Control, Kluwer Academic Publisher: Massachusetts, 1998. |
[10] | A. Soulhi, S. Guedira, and N.-E. El Alami, "Decision-making automation fuzzy decision-making in industry", In: Proceedings of the 8th WSEAS International Conference on Artificial Intelligence, Knowledge Engineering & Data Bases, 2009, pp. 181-185. |
[11] | A. Soulhi, S. Hayat, S. Hammadi, and P. Borne, "New strategy for the aid decision-making based on the fuzzy inferences in the traffic regulation of an urban bus network", In: Systems, Man, and Cybernetics, 1999. IEEE SMC'99 Conference Proceedings. 1999 IEEE International Conference, 1999, pp. 12-16. |
[12] | A. Soulhi, "Artificial Intelligence Contribution to decision-making aid in Management of the urban collective transport systems", M.S. thesis, University of Lille 1, Villeneuve-d'Ascq, France, 2000. |
[13] | S.D. Kaehler, Fuzzy logic – an introduction, part 2, Available at http://www.seattlerobotics.org/encoder/mar98/fuz/fl_part2.html, 2006. |
[14] | A.B. Tee, M.D. Bowman, and K.C. Sinha, "Application of fuzzy logic to condition assessment of concrete slab bridges", Transport. Res. Record., pp. 22-30, 1988. |
[15] | P. Borne, J. Rozinoer, J.Y. Dieulot, and L. Dubois, Introduction à la commande floue., Edition Technip: France, 1998. |
[16] | S. Hayat, R. Hartani, S. Sellam, and B.B. Meunier, "Modelling of an automatic subway traffic control by fuzzy logic and networks theory", In: IFAC Symposium on Transportations Systems’94, Tianjin, China, IFAC Publication, Pergamon Press Ltd. 24-26, 1994, pp. 205-210. |
[17] | A. Soulhi, S. Hayat, S. Hammadi, and P. Borne, "Scheduling and decision-making in the traffic exploitation of the urban bus networks", In: Conference ISAS’99, 1999, pp. 8-12. |