A Wildfire Prediction Based on Fuzzy Inference System for Wireless Sensor Networks

A Wildfire Prediction Based on Fuzzy Inference System for Wireless Sensor Networks V.G. Gasull1, D.F. Larios1, J. Barbancho1, C. León1 and M.S. Obaidat2 1 Department of Electronic Technology, University of Seville, Seville, Spain 2 Department of Computer Science & Software Engineering, Monmouth University, W. Long Branch, NJ 07764, U.S.A. {vgasull,dflarios}@dte.us.es, {jbarbancho,cleon}@us.es, [email protected] Abstract. The study of forest fires has been traditionally considered as an important application due to the inherent danger that this entails. This phenomenon takes place in hostile regions of difficult access and large areas. Introduction of new technologies such as Wireless Sensor Networks (WSNs) has allowed us to monitor such areas. In this paper, an intelligent system for fire prediction based on wireless sensor networks is presented. This system obtains the probability of fire and fire behavior in a particular area. This information allows firefighters to obtain escape paths and determine strategies to fight the fire. A firefighter can access this information with a portable device on every node of the network. The system has been evaluated by simulation analysis and its implementation is being done in a real environment. Keywords: Fuzzy System, Wireless Sensor Networks, Forest Fire, Simulation. 1 Introduction Usually, a wireless sensor networks is composed of multiple nodes spatially distributed in an area. These nodes obtain information on the environment such as temperature, pressure, humidity or pollutants, and send this information to a base station. A wide variety of applications for such networks often apply some kind of supervision, event detection, tracking or control, among others [1]. In [2] a large-scale deployment of these networks has been used for the supervision of wildlife habitats. Most applications of sensor networks in forest fires are based on the detection of fires, such as [3] which uses a WSN based on a swarm-inspired system for detecting wildfires. Reference [4] shows an algorithm based on fuzzy logic for detecting events. In this case, the event is fire detection. The nodes are equipped with various sensors such as temperature, humidity, light intensity and carbon monoxide. In [5], a satellite monitoring system with a WSN is used to detect a forest fire. In [6] an implementation scheme of communication oriented WSN and monitoring computer is presented. The work in reference [7] reduces the consumption of the transmission using the information gathered by analyzing the Fire Weather Index (FWI) System. Reference [8] uses data from a WSN in the FARSITE simulator for fire detecting. Other papers are related to the use of WSN to improve the security on evacuations [9]. Reference [10] deals with the use of WSN to improve the information gathering for firefighters, in order to better perform when extinguishing a fire. Some other research work is related to the study of fire evolution [11] or also fire prevention [12]. Others do real experiments with WSNs in order to evaluate its robustness against real conditions of a wildfire [13]. Traditionally, most wireless sensor networks are used for monitoring meteorological variables, with the final purpose of detecting the occurrence of fires. Only a few distributed approaches are proposed, like the one in [14], where the remote nodes process information of multiple sensors (temperature and smoke) sending alarm to a base station if a node detects an incident. However, the scope of these networks can be increased by using distributed processing techniques and computational intelligence. In this paper, we propose a novel system for prediction of fire as well as the prediction its evolution. One advantage of this system is the real time processing of environmental variables, and then the information is available in real time, knowing at all times the current state of forests. Hence, if a fire brigade is currently operating in the area, it can choose the escape route or know which attacking side has to put out the fire. The proposed system, called ISFPWSN (Intelligent System for Fire Prediction using Wireless Sensor Networks), is based on a distributed processing, that transmits information to neighboring nodes and does not need a base station. Centralized algorithms could be a problem in real situations because with a fire, one or more nodes could be burned, so the path to the base station could disappear. The purpose of ISFPWSN is evaluating the risk of fire as 95% of wildfires are caused by humans [15]. It is necessary to consider sociological information, not only environmental conditions. Apart from the risk of fire, other goal of ISFPWSN is to offer information about the behavior and evolution of the fire in case of wildfire. It reduces the risk of the people exposed and improves the fire detection because it offers information about secure ways of escape and permits evaluation of a strategy for fire extinguishing. ISFPWSN is based on computational intelligence algorithms that use fuzzy inference systems. It is because a fuzzy system has many advantages for WSN applications [166], such as its simplicity, which permits execution on devices with limited capabilities, or its ability to manage imprecise and uncertain information. All of these characteristics allow us to obtain a robust system without a high computational load. The remainder of this paper is organized as follow. Section 2 presents the proposed system ISFPWSN. In Section 3, the simulator developed for testing the system is described. The results are presented in Section 4. Finally, Section 5 contains the concluding remarks and provides a discussion for future works. 2 Proposed System 2.1 Hardware Infrastructure of ISFPWSN The system is designed to be used in a wireless sensor network, such as the one illustrated in Figure 1. This Figure shows a common network made by anchor nodes, but in this case there is no base station. The portable device, which is carried by qualified personnel, acts as base station, gathering the information of the nodes. A brief description about them is given below. Fig. 1. The ISFPWSN network Anchor Nodes. Anchor nodes are made up of wireless sensors that form a wireless sensor network. These devices can communicate among each other using the appropriate routing protocols. The nodes execute all the algorithms of ISFPWSN. Every Anchor node obtains the next measurements of sensor which include: temperature, wind direction, wind velocity, daily rain and humidity. This information can be obtained from a small and cheaper weather station attached to each node, where the node can gather the information through a communication port, such as RS232-C or RS-485. It is interesting to provide power for these systems using renewable energy sources, such as solar panels or wind energy means. Portable Devices. The other devices used in the project are the measurement apparatus devices. They are small and portable devices used by the fire-fighters to obtain environmental information. These devices can access all information on the state of the fire and can help in case of fire to provide information about its future evolution. This information is useful both to the extenuation and to determine secured ways for an escape; increasing worker safety. 2.2 Descriptions of ISFPWSN Algorithms The nodes of the network execute their local processing algorithms to obtain the local estimates. Then they send these local estimates to other nodes in the network. Every node of the network gathers its information and executes the distributed processing algorithms. Subsequently, all the nodes share their information about the risk of fire or, in case of a forest fire, its evolution. ISFPWSN does not need a base station to gather the information. This increases the robustness of the network in case of wildfire because if some nodes were damaged in the fire, it is possible that they will not find a route to send the information to the base station. Figure 2 sums up a flowchart of the main algorithm of the system. This figure shows several processes which will be explained later. In order to save energy, extending the lifetime of the network, the refresh rate (i.e., the frequency at which measurements are taken from the environment and therefore communicating with neighboring nodes), of this algorithm will automatically adjust as a function of the risk of fire. In this approach, each node executes one or other processes depending if a fire has occurred. If there is a fire, it will run algorithms to study their evolution, Fire Behavior (FB) and Prediction of Direction and Velocity of Fire (PD) at the maximum refresh rate. On the contrary, if there is no fire, it will run the prediction process, Probability of Fire (PF) and Risk of Fire (RF). Fig. 2. Flowchart of the main algorithm If there is a wildfire all nodes are aware and they update their data and communications quickly; obtaining the desired information in the rest of the nodes in real time. In this case, a firefighter can access the data available to the system at any time, and reacts appropriately using the parameters displayed from the nodes in the network. Fig. 3. Processing types ISFPWSN is based on two main processes: a local process and a distributed process, as shown in Figure 3. These processes shared information between them as well as with the other nodes that form the network. One goal of our approach is that a firefighter can access this information anywhere on the network, due to the fact that all nodes share information concerning the estimation or the behavior of the fire. 2.3 Local Processing Local processing is executed on every node of the network. Nodes evaluate the environment with the information provided by each sensor. The local processing offers a partial solution of the global state of the system since this method only uses local information to get the results. This proposed processing is basically a fusion and data aggregation algorithm. Within local processing there are two algorithms; an algorithm to obtain the probability of fire occurrence (PF) and an algorithm to obtain the behavior of fire (BF). Probability of Fire Occurrence (PF). This algorithm determines the risk of fire in the environment in the neighborhood of the node. The output of the PF algorithm is in [0-100] range, which indicates the probability of ignition of a forest fire after normalization. This algorithm is divided up into two blocks: an environmental processing and a sociological processing. Both methods are based on a fuzzy logic engine (Figure 4), and have as outputs fuzzy sets with functions summarized in Table 1. Fig. 4. Algorithm of the probability of fire Table 1. Fuzzy sets of the outputs of environmental and sociological processing Name Membership function Very low Trapezoid (-20,0,10,17) Low Triangle (13,25,37) Medium Triangle (35,47,59) High Triangle (57,69,81) Very high Trapezoid (79,90,100,120) Environmental Processing. This processing method obtains the fire likelihood as a function of the environmental variables. This is the classic assumption considered in other fire simulators, such as Behave Plus and FARSITE [17]. These simulators base their approximations on the Rothermel model [18] or the FWI index [19]. To determine the input functions of the fuzzy logic system, a preliminary study of the importance of the climate variables in the generation and evolution of a fire has been done. Therefore, due to historical data of wildfires, temperature and the fuel moisture has been considered as input variables. These magnitudes appear as the most relevant ones. Fuel moisture is the dead fine fuel moisture. Within it, other variables such as humidity, shading from the sun, slope and terrain exposure has been considered. The fuzzy sets of the inputs variables are shown in Table 2. The parameters of fuzzy logic engine have been obtained by ANFIS (Adaptive Neuro Fuzzy Inference Systems) techniques which allows us obtain more accurately membership functions. This fuzzy logic engine has the knowledge base described in Table 3. Because there are 25 rules, this table only shows a small sample sufficient for comprehension. Table 2. Fuzzy sets of the inputs of the environmental processing Name Set Membership function Temperature Very-low Triangle (-62, -40, -18) Low Triangle (-40, -18, 4) Medium Triangle (-18, 4, 26) High Triangle (4, 26, 48) Very-high Triangle (26, 48, 709 Fuel Moisture Very-low Triangle (-5.5, 1, 7.5) Low Triangle(1, 7.5, 14) Medium Triangle (7.5, 14, 20.5) High Triangle (14, 20.5, 27) Very-high Triangle (20.5, 27, 33.5) Table 3. Base of knowledge of the fuzzy logic engine Temperature F. Moisture Output Low High Very low Low Medium Low Low Low Medium Medium High Low Medium Medium Medium Medium Low High High High Medium High Medium High High Low Very High Figure 5 shows the probability of fire occurrence. This surface shows the probability for different values of temperature and fuel moisture used in simulations. Temperatures are in [-40 to 49] ºC ranges and fuel of moisture [1 to 27] % range. As it can be seen, when temperature increases and fuel of moisture decreases, the probability of fire occurrence increases, reaching very-high values (around 90-100%). Sociological Processing. T the sociological variables, s As an example, Figure 6 1992-2002. As it can be se e a great percentage of these w F Because of this, it is n e risk of fire. As example, w as important as the environ m generally not consider this k Sociological behavior i s niques. Fuzzy logic is a go o this information is the ma i with a fuzzy logic engine. The input variables are d marized in Table 4. These f a region. In this case, we h a fires that took place during occurred during this period . Sixty percent (67%) of fo tember and the time of det e A low season for park vis i high season considered bet w T his processing obtains the fire likelihood as a functio n s uch as weekends or holydays. 6 shows the causes of forest fires in Spain during the pe r e n, man causes approximately 95% of wildfires. Moreo v w ildfires are produced during holidays. F ig. 5. Probability of fire occurrence e cessary to consider sociological variables to estimate w e consider that the number of visitor to the natural par k m ental variables. Despite its importance, fire simulator s k ind of information, because it is difficult to assess it. s uncertain and it is not easy to model with classic t e o d approach to evaluate this kind of behavior. Conside r i n objective of sociological processing. It is impleme n Fig. 6. Causes of wildfires d ay of week, month and hour time. Its fuzzy sets are s u f uzzy sets can be obtained from the past fire informatio n a ve considered a Spanish report with the information o f the years 1992 – 2002 [15]. Figures 7 and 8 show the . fo rest fires occurred in the months of July, August and S e ction is 73% for fall time slots between 12:00 and 20 i ts is considered between November and April, while w een June and September. Low visit hours represent ti m n of r iod v er, the k is s do e chr ing n ted u mn of f the fire S ep- :00. the m es when it is cold during the day (0:00 – 10:00h), while high visit hours are the hours when it is hot during the day (13:00 -19:00h); medium visit hours are all other times. For days of a week, all weekdays were considered low, Saturday or before holidays were considered high and on Sunday or holidays were medium. Table 4. Fuzzy sets of the inputs of the sociological processing Hours Low 0:00-10:00 h. Triangle Medium 10:00-13:00 and 19:00-23:00.h High 13:00-19:00 h Months Very-low Jan. Feb. Nov.Dec Triangle Low Mar. Apr. Medium May. Oct. High Jun. Sep Very-high Jul. Aug Days of Week Low Working day Gaussian Medium Sunday High Saturday Fig. 7. Month of detection Fig. 8. Detection time The proposed fuzzy logic engine has 3 input variables; hours, months and days of week and 1 output variable; probability of ignition for sociological variables. This base of knowledge is made up 13 rules. Some of these rules are listed below: ─If hour is Low then Probability of Ignition (PI) is Very-low. ─If months Low then PI is Very-low. ─If hour is Low and month is high then PI is Medium. ─If hour is High and month is Medium then PI is Medium. ─If Day of Week is Medium then PI is High. Figure 9 shows the different values of the probability of ignition of sociological variables. This area is a surface of probabilities for the input variable hours and months. It is important to mention that this fuzzy system is designed to apply to Spain, where the dry months are between April and November. For other regions, it would be necessary to change the fuzzy logic engine. Fire Behavior (FB). This algorithm obtains the direction of the fire front and its velocity in an area around the node. It does not consider the topology of the terrain; it is used on the distributed algorithm for prediction of direction and velocity of the fire. This algorithm considers the nodes as isolated devices with their local information of wind direction and velocity, temperature, humidity and daily rain. This algorithm is executed only in case of occurrence of fire. Fig. 9. Probability of ignition for sociological variables The velocity of propagation of the fire is calculated using the formula below: V = a(U + 1)b (1) Where V is the velocity of propagation of fire (m/min), U is the wind velocity (m/s), a=0.233 and b=1.332 for a “Pinnus Pinea” forest, as we considered this in this work. The direction of propagation can be obtained as follows. If wind velocity = 0 Km/h: The fire follows a radial pattern, with the same velocity on all direction. For other cases, the fire front has the direction of the wind and will trace an ellipse. This algorithm also implements an alarm system for risk of great fires, according to the rule of 30% [15]. This rule says that a great forest fire can be produced with the conditions described on Table 5. Table 5. Rule of 30% Parameter Value Temperature > 30% Humidity < 30% Wind velocity > 30km/h Days without rain > 30 When two or more values are above the threshold, the system considers that there are critical values (Figure2). Instead of taking the system back to the continuous capture of data from the environment, it verifies previously if a fire occurred. 2.4 Distributed Processing The distributed processing is executed on each node of the network with the information gathered from the broadcast messages sent by the others nodes. This permits checking the global state of the environment without the need for a base station that Fig. 17. Histogram e 5 Conclusions an d In this work, our ISFPWS sensor networks. In order t o we use computational inte l network. One aim of the p r mitting the development of The proposed algorith m cooperate with the wildfire act in two ways: first, it c a work gives useful informat i In case of fire, the syste m fire fronts. This informatio the fire and how to design e The results obtained wi t puts from the Behave simu l Currently, we are desig n in a real environment in o r currently working to dev e algorithms proposed in our Acknowledgements. This w Ciencia y Empresa, Junta d e (reference number P07-TI C 5705). The authors would l i Council for Scientific Re s (Doñana Natural Space, Al m References 1. Akyildiz, I.F., Su, W., S a survey. Computer Netwo r rro r Fig. 18. Dimensional error d Future Work N system is presented. This system is based on wire l o obtain the probability of fire and fire behavior predic t l ligence schemes in each of the nodes that belong to r oposed wireless sensor networks is to be economical, p the system in a huge area with a low cost. m s employ collaborative processing techniques in orde r fighting based on fuzzy logic processing. This system a n determinate the risk of fire and, in such a case, this n i on to the firefighters. m gives information about the location and direction o f n can be used to study in an effective way how to co n e scape routes in order to enhance workers’ safety. t h the proposed system have been compared with the o l ator, which proved the accuracy of our system. n ing a real prototype version of the system, to be develo r der to validate the proposed algorithms. The authors e lop neural network system to verify the fire beha v scheme. w ork has been supported by the Consejería de Innovac i e Andalucía, Spain, through the excellence project ART I C -02476) and project eSAPIENS (reference number T i ke to thank the Doñana Biological Station – EBD (Spa n s earch CSIC) and the Doñana Biological Reserve I C m onte, Huelva) for their collaboration and support. a nkarasubramaniam, Y., Cayirci, E.: Wireless sensor network s r ks 38(4), 393–422 (2002) l ess t ion the p err to can n etf the n trol o utped are v ior i ón, I CA T ICn ish C TS s : A 2. Mainwaring, A., Polastre, J., Szewczyk, R., Culler, D., Anderson, J.: Wireless sensor networks for habitat monitoring. In: Proceedings of the International Workshop on Wireless Sensor Networks and Applications, pp. 88–97 (2002) 3. Ramachandran, C., Misra, S., Obaidat, M.S.: A probabilistic zonal approach for swarminspired wildfire detection using sensor networks. International Journal of Communication Systems 21(10), 1047–1073 (2008) 4. Manjunatha, J., Verma, A.K., Srividya, A.: Multi-Sensor data fusion in cluster based wireless sensor networks using fuzzy logic method. In: Proceedings of the International Conference on Industrial and Information Systems (2008) 5. Zhang, J., Li, W., Yin, Z., Liu, S., Guo, X.: Forest Fire Detection System based on Wireless Sensor Network. In: 4th IEEE Conference on Industrial Electronics and Applications, pp. 520–523 (2009) 6. Li, W., Zhang, J., Zhang, J., Xia, Z., Wang, G.: The monitoring of host computer for forest fire detection system based on wireless sensor network. In: 5th International Conference on Wireless Communications, Networking and Mobile Computing, pp. 1–4 (2009) 7. Hefeeda, M., Bagheri, M.: Wireless sensor networks for early detection of forest fires. In: IEEE International Conference on Mobile Adhoc and Sensor Systems, pp. 1–6 (2007) 8. Kosucu, B., Irgan, K., Kucuk, G., Baydere, S.: FireSenseTB: A wireless sensor networks testbed for forest fire detection. In: 6th International Conference on Wireless Communications and Mobile Computing: Connecting the World Wirelessly, pp. 1173–1177 (2009) 9. Cherniak, A., Zadorozhny, V.: Towards adaptive sensor data management for distributed fire evacuation infrastructure. In: 11th International Conference on Mobile Data Management, pp. 151–156 (2010) 10. García, E.M., Serna, M.A., Bermúdez, A., Casado, A.: Simulating a WSN-based wildfire fighting support system. In: International Symposium on Parallel and Distributed Processing with Application, pp. 896–902 (2008) 11. Lei, Z., Lu, J.: Distributed coverage of forest fire border based on WSN. In: 2nd International Conference on Industrial and Information Systems, pp. 341–344 (2010) 12. Obregón, P.D.P., Sondón, S., Sañu, S., Masson, F., Mandolesi, P.S., Julián, P.M.: System Based on Sensor Networks for Application in Forest Fire Prevention. In: Micro-Nanoelectronics, Technoloy and Applications, pp. 61–65 (2009) 13. Antoine-Santoni, T., Santucci, J.F., de Gentili, E., Silvani, X., Morandini, F.: Performance of a protected wireless sensor network in a fire: analysis of fire spread and data transmission. Sensors 9(8), 5878–5893 (2009) 14. Liu, S., Tu, D., Zhang, Y.: Multiparameter fire detection based on wireless sensor network. In: IEEE International Conference on Intelligent Systems, pp. 203–206 (2009) 15. Plan INFOCA 2003. Cap I-VIII, http://www.juntadeandalucia.es 16. Marin-Perianu, M., Havinga, P.: D-FLER – A Distributed Fuzzy Logic Engine for Rule-Based Wireless Sensor Networks. In: Ichikawa, H., Cho, W.-D., Satoh, I., Youn, H.Y. (eds.) UCS 2007. LNCS, vol. 4836, pp. 86–101. Springer, Heidelberg (2007) 17. Fire: Public domain software for the wildlandfire community (2008), http://www.fire.org 18. Rothermel, R.C.: How to predict the spread and intensity of forest and range fires. US Department of Agriculture, Forest Service, General Technical Report, INT-143 (1983) 19. Canadian Forest Fire Danger Rating System (CFFDRS), Equations and FORTRAN program for the Canadian forest fire weather index system (1985)