Regional-scale risk assessment of forest fires induced by distribution lines via a hybrid approach

1.   Introduction

In recent years, forest fires induced by power distribution lines across forest regions have been frequent, and once accidents occur, they may affect power supply safety or cause extensive forest fires^[1–4]. For example, the forest fire in Xichang city, Liangshan Prefecture, Sichuan Province, on March 30, 2020 resulted in 19 victims and 3 injuries, with a total area of 3047.78 hectares of various land fires and a direct economic loss of 97.312 million CNY. The direct cause of the accident was that the drainage line of the 10 kV distribution line overlapped with the support frame hoop of the pole cross arm under the action of a wind force in a specific direction and formed a permanent ground discharge fault, resulting in the melting of the line body, insulation material burning, and further igniting the surrounding combustible materials to form a fire spread^[5]. The Dixie fire^[6] in California in 2021 resulted in three injuries, while a total of $637.4 million was spent on firefighting. The immediate cause of the incident was a large Douglas-fir tree that fell on a power line, causing electrical failure and releasing an electrical arc that ignited fuels on the forest floor over the next few hours. These fire accident cases demonstrate the critical importance of forest fire risk assessment for daily transmission line maintenance^[7]. The 10 kV distribution line across the forest region is the key component of the grid system, and risk assessment for stable operation is a very important guide for the daily operation and maintenance of operating companies.

The scientific basis for forest fire risk assessment began to take off and evolve in the 1910s^[8]. Many scholars have conducted research on forest fire risk assessment. Akinola and Adegoke^[9] assessed forest fire risk in four ecoregions of Missouri on the basis of several measurable environmental parameters, such as forest density and forest type, that affect vulnerability to forest fire risk and determined the risk level for each region and the corresponding risk management strategies. Novo et al.^[10] proposed a forest fire risk mapping method based on light detection and ranging analysis, with relevant parameters such as elevation, road distance, and fire weather indices used as indicators of the possibility of forest fires in the fire risk assessment phase. However, most current studies have focused only on the state of the forest itself, environmental conditions and human activities, and have not considered the impact of power systems across forest regions on forest fires.

Current research on fire risk assessment for forest power systems has focused on the establishment of indicator systems or mountain fire spread models for risk assessment of transmission lines^[11–13]. For example, Xiong et al.^[13] analyzed transmission line fault data and calculated the probability of mountain fires via a forest fire risk meteorological rating to establish a transmission channel risk assessment model with spatial and temporal distribution characteristics for mountain fires. However, transmission lines and distribution lines are very different in terms of voltage level, current carrying capacity, and line support. Transmission lines usually exceed 69 kV and have high current capacity. However, distribution lines are usually below 10 kV and have lower current carrying capacities. In addition, transmission lines use tall, long steel towers for line support, whereas distribution lines are generally supported by means such as wooden poles and track poles. This means that transmission lines tend to have higher strength line support and are far from the ground and forest canopy, whereas overhead distribution lines often cross forest regions, greatly increasing the possibility of contact with the tree canopy. Public documents released by PG&E^[14] indicate that distribution lines are three times more likely to cause wildfire ignitions per mile of line than transmission lines are. In summary, due to the different uses, design specifications and protection measures of transmission and distribution lines, there are differences in the level of risk between these two types of lines.

However, research on distribution lines across forest regions has focused mainly on lightning protection and relay protection schemes. For example, Zhu et al.^[15] calculated the lightning resistance level of 35 kV distribution lines under different operating conditions and established a lightning strike risk assessment model based on entropy weighting and the TOPSIS method. Liang et al.^[16] analyzed the types and causes of distribution line fault-induced forest fires and proposed a multilevel current protection scheme. In fact, forest fire accidents caused by overhead distribution lines are frequent^[1–4], but little research has been conducted on forest fires caused by distribution lines. Taylor and Roald^[17] assessed forest fire risk in terms of both hill fire potential and fire probability and derived hill fire potential and fire probability on the basis of historical fire data and line voltage ratings. However, for a single specific assessment region, historical fire data may be rare or difficult to obtain, which greatly reduces the accuracy of the assessment. Therefore, there is an urgent need to construct a risk assessment framework for forest fires induced by distribution lines, conduct fire risk assessments of overhead distribution lines across forest regions on the basis of regional characteristics and use more readily available data to reflect the true forest fire risk level in the assessment region.

To address this issue, this paper proposes a novel assessment framework for the regional risk of forest fires induced by distribution lines. This framework is based on the Analytic Hierarchy Process (AHP), Bayesian network (BN) and Fussel-Vesely (FV) importance metric, where the AHP method is used for scoring against a single route, and the BN and FV importance metric are used for fire probability outputs and hazardous event identification against the region. Regional risk assessment can provide more comprehensive and generalized guidance to local jurisdictions than can risk assessment for a single line. The regional assessment methodology in this assessment framework not only assesses the regional risk but also identifies the poorly performing lines in the region and makes different targeted recommendations for the region as well as for single lines. The assessment framework provides the probability value of the forest fire risk of distribution lines in the region, as well as the lines that need to be focused on and the targeted measures that need to be taken. In this work, field research data and historical operation and maintenance data are used as basic data, which are easier to obtain than the historical fire data used in previous methods.

2.   Materials and methods

The assessment framework proposed in this paper is shown in Fig. 1. The assessment framework is strongly generalized to assess the risk of forest fires induced by overhead distribution lines crossing forest regions in different assessment regions and outputs accurate probability values of fires. First, the assessment region is determined, the risk factors are identified, and the indicator system is constructed in step 1. Next, in step 2, field data research and historical operation and maintenance data collection are conducted on the basis of the constructed risk indicator system, and on the basis of these data, scores are calculated in step 3 via the AHP, and lines with poor overall or secondary indicator score profiles are identified among them. The research data in step 2 are then transformed into input data for the BN, the probability of each hazard event in the assessment is derived via Bayesian inference in step 4, and a criticality analysis is performed via the FV importance metric to derive the root nodes that have a greater impact on the probability of forest fires. Finally, targeted recommendations are made in step 5 for high-risk regions and major lines that need to be corrected on the basis of the results of the analysis. These steps are described in detail in Sections 2.1 to 2.5 below.

Figure  1.  The assessment framework for the regional risk of forest fires induced by distribution lines.

DownLoad: Full-Size Img PowerPoint

2.1   Step 1: risk indicator system construction

The process of forest fires induced by distribution lines is caused mostly by line operation faults that generate high temperatures or electric sparks, which ignite surrounding combustible materials and thus start a fire. Combining the research data and accident cases^{[5, 6]} and summarizing and analyzing the typical risk factors identified in previous studies^{[10, 13, 18–23]}, we identified the risk factors that induced forest fires. Twenty-nine typical risk factors are identified, which influence the probability of a forest fire, and some of them have been mentioned by previous studies many times. According to the possibility and danger of the disaster, the exposure and vulnerability of the carrier, and the level of prevention and control, these factors are classified as surrounding fire hazard conditions, component structure health conditions, line operation conditions, technical protection measure conditions, and management conditions. The risk indicator system is constructed on the basis of the risk identification results, as shown in Table 1.

Table  1.  Risk indicator system.

Secondary indicators	Tertiary indicators	Weights
Surrounding fire hazards condition	1. The horizontal safety distance is too small	0.105
	2. The vertical safety distance is too small	0.090
	3. The width of surrounding road is too narrow	0.035
	4. Excessive flammable sundries around the line	0.130
Component structure health condition	5. Tower exceeded its service life	0.010
	6. Tower does not meet the requirements	0.010
	7. Wires and cables exceeded their service life	0.025
	8. Wires and cables do not meet the requirements	0.040
	9. Insulator exceeded its service life	0.015
	10. Insulator does not meet the requirements	0.020
Line operation condition	11. Overheating of line components	0.090
	12. Line (of electricity) leakage	0.060
	13. Too high grounding resistance	0.035
	14. Line tripping	0.045
Technical protection measures condition	15. Failure of Fault Indicator	0.010
	16. Failure of On-column Load Switch	0.025
	17. Failure of On-column Circuit Breaker	0.030
	18. Failure of Overvoltage Protector	0.020
	19. Failure of Bird Repellent	0.020
	20. Failure of Disconnector	0.010
	21. Failure of Drop Fuse	0.025
	22. Failure of Lightning Arrester	0.040
Management condition	23. Insufficient implementation of the production post system	0.010
	24. Insufficient implementation of employee learning system	0.010
	25. Insufficient implementation of technical data system	0.010
	26. Insufficient implementation of information and warehouse management system	0.010
	27. Insufficient implementation of operational analysis system	0.020
	28. Insufficient implementation of equipment defect management system	0.030
	29. Insufficient implementation of the safety activity system	0.020

 | Show Table

DownLoad: CSV

Surrounding fire hazards condition. Most of the safety hazards in the surrounding environment of overhead distribution lines across forest regions are trees and flammable debris left by human activities. When the distance between the conductor and trees is too small or the width of the fire escape is too narrow, it may cause line failure and thus induce forest fires and affect subsequent rescue^{[24, 25]}. When there is too much flammable debris around a line, the probability of forest fires increases.

Component structure health condition. 10 kV overhead distribution lines are composed mainly of wires and cables, insulators and towers. Each component needs to adopt a model and material suitable for the AC voltage value of the distribution line. If the material or model of the component does not meet the requirements, the failure of the whole distribution line and each component may be aggravated^[26]. In addition, when the service life of each component in the distribution line reaches a fixed number of years, its failure probability increases. When the main structure of the distribution line fails, it may produce electric sparks and cause forest fires^[27].

Line operation condition. In the daily operation of 10 kV overhead distribution lines, line faults may be caused by their own performance or by external influences. The operating parameters mainly include the line temperature, leakage, grounding resistance value, and tripping condition. When the line is overheated, it will lead to an increase in line resistance, which will cause the line to burn up in serious cases^[26]. When leakage occurs in a line, it may cause a safety accident, and when leakage is severe, it can produce electric sparks and high temperatures, resulting in forest fires. The main role of the grounding resistance is to introduce the lightning current that hits the distribution line to the Earth. When the value of the grounding resistance is too high, it may lead to the inability to discharge normally while causing some of the phase voltage to rise, resulting in the burning of the line^[24]. The tripping situation of the line reflects the overall operating condition; when the line operating condition is poor, the tripping rate also increases, and the possibility of high temperature and electric sparks generated by the line also increases^[28].

Technical protection measures condition. Equipment and facilities related to the protection technology of 10 kV distribution lines across forest regions include fault indicators, on-column load switches, on-column circuit breakers, overvoltage protectors, bird repellents, disconnectors, drop fuses, and lightning arresters. Overload, lightning and bird nesting threaten the operational safety of distribution lines. The role of this technical equipment is to reduce the possibility of generating high temperatures, electrical sparks, and line failures, as well as to reduce damage and the severity of consequences after a fault occurs. For example, lightning arresters can effectively improve the level of lightning resistance of distribution lines and reduce lightning accidents on distribution lines. The use of bird repellents can reduce tripping accidents caused by bird damage. Therefore, when technical protection fails, the possibility of high temperatures and electric sparks on the line increases, thus triggering forest fires.

Management condition. The departments to which the distribution lines belong should develop relevant management systems to ensure the safe and stable operation of the distribution lines^{[28, 29]}. These management systems include production post systems, employee learning systems, technical data systems, information and warehouse management systems, operational analysis systems, equipment defect management systems, and safety activity systems. When the system is not implemented adequately, it may lead to an increased likelihood of line failure and thus an increased probability of induced forest.

2.2   Step 2: data collection

In this step, the main method is field research to collect data on the actual condition of each line in the region and to collect historical operation and maintenance data from the power companies and other departments responsible for the assessment region. The field research data include surrounding fire hazard conditions and technical protection measures, such as vertical and horizontal distances between tree barriers and lines and the installation of safety devices. Historical operation and maintenance data include the component structure health condition, line operation condition, and management condition, such as grounding resistance values, line tripping, and system implementation.

2.3   Step 3: AHP analysis

The main purpose of this step is to sift out lines with relatively low overall scores or secondary indicator scores to provide the basis for analysis in step 5. The methodology used in this step is the analytic hierarchy process (AHP). AHP is a combined qualitative and quantitative analysis method that allows quantitative processing of problems that are difficult to analyze quantitatively and have complex conditions to obtain a decision solution that meets the requirements of decision makers^[30]. The process relies on the decision maker’s judgment of the relative importance among the indicators, forming a judgment matrix and conducting consistency tests, calculating the indicator weights via arithmetic averaging^[31], and finally scoring the assessment results via a scoring criterion. In this step, the score is first calculated via the AHP on the basis of the risk indicator system established in step 1 and on the data collected in step 2. It is necessary to use expert consultation to obtain the judgment matrix of the relative importance of the indicators, and the commonly used basic judgment scale is the 1–9 value scale proposed by Saaty^[32], as shown in Table 2, in which a_{i, j} indicates the relative importance relationship between risk indicators i and j, and the higher the a_{i, j}, is, the greater the relative importance. In this study, two experts were consulted. One is an engineer with more than a decade of experience at the State Grid Anhui Electric Power Company, and the other is a professor who has studied forest fires for more than a decade. After the expert judgment matrix is obtained, the indicator weights w_i are calculated via Eq. (1)^[33], where n is the number of indicators.

Table  2.  The 1–9 value scale.

ai,j	Definition
1	Equal importance
3	Indicator i is slightly more important than indicator j
5	Indicator i is more important than indicator j
7	Indicator i is very more important than indicator j
9	Indicator i is extremely more important than indicator j
2, 4, 6, 8	Intermediate values between the two adjacent judgments

 | Show Table

DownLoad: CSV

After the indicator weights are obtained, Eq. (2) is used to calculate the line score.

where S is the line score and C_i is the score of indicator i. The scores of each indicator are obtained by combining the field research information and the scoring criterion. Table 3 shows the scoring criteria of some indicators. The detailed scoring criterion mainly refers to the current standards and regulations^{[24–29, 45, 46]} and the long-term working experience of engineers in the power industry. The detailed scoring criteria for all the indicators are given in Table S1 in Supporting informaiton.

Table  3.  Detailed scoring criteria for several indicators.

Indicators	Detailed scoring criterion	Corresponding Score
Tower exceeded its service life	1~50 years, the state is stable.	100
	Less than 1 year, it is slightly easier to failure due to manufacturing defects or improper installation.	50
	More than 50 years, it is easier to failure.	0
Tower does not meet the requirements	Iron tower or steel pole	100
	Cement pole	50
	Wooden pole	0
Wires and cables do not meet the requirements	Fluoroplastics and other insulating materials, with good mechanical properties, aging resistance and high temperature resistance.	100
	Cross-linked polyethylene and other insulating materials, with good mechanical properties and high temperature resistance.	50
	Bare wires	0
Insulator does not meet the requirements	Ceramic insulators with fire resistance	100
	Glass insulator, may burst	50
	Synthetic insulators, not resistant to high temperatures	0

 | Show Table

DownLoad: CSV

When the line score is out of 100, a higher score represents a better line condition, which is equal to a lower risk. In addition to calculating the line score, the scores of secondary indicators can also be calculated. The way is to calculate only the tertiary indicators under the calculated secondary indicator by Eq. (2) to obtain the secondary indicator score and standardize it. After the total score of all the lines and the scores for each secondary indicator are derived, the lines with relatively low total scores and relatively low scores for some secondary indicators can be identified.

2.4   Step 4: Bayesian network analysis

The main purpose of this step is to derive regional fire probabilities as well as key measures through the Bayesian network and Fussel-Vesely importance metrics. These findings provide the basis for the analysis in step 5. BN is a probabilistic-based graphical model whose main role is to calculate the probability distribution of a set of variables on the basis of observations of the variables and prior knowledge^[34]. Bayesian networks consist of a directed acyclic graph (DAG) model and a conditional probability table (CPT). The DAG and CPT provide qualitative and quantitative descriptions of the dependencies between variables, respectively. The advantage of Bayesian networks over traditional methods of determining causality is that independence between variables is easily identified and isolated^[35].

The DAG consists of directed edges and nodes, X₁, X₂ and Y are nodes, and the directed edges between nodes represent the dependencies between nodes, where the starting nodes of the directed edges are the parents, which are X₁ and X₂, and the end nodes are the children, which are Y. The CPT contains the conditional probabilities of all nodes in the Bayesian network. As shown in Eq. (3), the conditional independence assumption and marginalization can be applied to each state probability value of a node to obtain the edge probability of that node^[36].

After the risk factors for forest fires caused by overhead distribution lines across forest areas are determined, a Bayesian network is constructed according to the risk indicator system. The probability of the root node is transformed from the data collected in step 2 by frequency statistics, and this probabilistic statistical transformation method has been used in many studies^[37–40]. The scoring data and statistical results of the research data are provided to the experts, and the interdependence of the risk factors is obtained on the basis of the expert system to determine the input parameters of the subnode event CPT in the Bayesian network. This step allows us to obtain the probability of forest fires induced by distribution lines of the assessment region and the probability of occurrence of each subnode.

To better prevent accidents and mitigate applicable measures of consequences, the most critical root nodes leading to top events need to be identified through critical analysis.

The FV importance metric describes the contribution of root nodes to top events and can also be used to identify critical nodes. For the root event $ {x}_{i} $, the FV can be derived from Eq. (4), which represents the ratio of the probability of occurrence of any cut set containing the root event $ {x}_{i} $ to the probability of the top event. The identified critical root nodes have a greater impact on the probability of forest fires and therefore need to be the main direction of correction in regional rectification.

2.5   Step 5: obtaining the main rectification measures

On the basis of the results of the analysis in steps 3 and 4, region-wide rectification can be carried out in the assessment region for hazardous events (root nodes) with a high impact on forest fire probability, and additional rectification measures can be targeted for low-scoring lines on the basis of the scores. The materials and methods section may contain subheadings.

3.   Results and discussion

The forest resources in Anhui Province are basically concentrated in the southern part of Anhui Province. The existing forestland area in southern Anhui is 266,700 hectares and has grown rapidly. The region presents a topographic landscape with a staggered distribution of low mountains, hills and basins; abundant precipitation; and rich vegetation. The complex environment, overload of combustible materials and many fire risk points present serious challenges to forest fire prevention in the region^{[41, 42]}.

In this study, a total of twenty 10 kV overhead distribution lines across forest regions under the State Grid Anhui Electric Power Company in southern Anhui were selected and numbered 1−20, with the basic profile of each line, such as overhead length, across forest regions, and the insulation rate shown in Table 4. The research area is divided into three regions, A, B and C, according to regional characteristics such as line density and geomorphological features, among which 11 lines are located in region A, 4 lines are located in region B, and 5 lines are located in region C. Each line is divided into several assessment units for field research, and each assessment unit is scored on the basis of the scoring rules via the hierarchical analysis method. The score is based on a percentage system, with higher scores representing better conditions for 10 kV overhead distribution lines across forest regions.

Table  4.  Line information.

Region	Line Number	Total length of the line (km)	Overhead line length (km)	Insulation rate	Maximum grounding resistance (Ω)	Length across forest (km)
Region A	1	31.66	28.01	100%	12	3.0
	2	9.13	9.14	0%	7	2.1
	3	21.89	21.89	0%	8	1.4
	4	21.88	21.74	100%	6	1.5
	5	11.27	11.27	8%	6	1.8
	6	17.67	17.67	0%	5	1.8
	7	15.24	15.24	100%	10	2.2
	8	34.98	34.98	9.2%	11	22.0
	9	65.52	63.52	26.5%	10	30.0
	10	49.43	49.40	100%	9	6.0
	11	17.61	16.61	88.8%	11	2.0
Region B	12	22.24	19.28	100%	110	1.0
	13	23.29	23.08	100%	160	22.0
	14	51.37	49.10	100%	100	2.0
	15	20.51	20.51	0.1%	10	5.4
Region C	16	51.51	46.21	100%	100	8.0
	17	79.03	77.66	100%	100	20.0
	18	16.81	16.81	6.7%	100	12.0
	19	4.24	3.35	6.7%	100	12.0
	20	32.59	32.59	60%	100	26.0

 | Show Table

DownLoad: CSV

3.1   Field research and AHP analysis results

In the field research, data collection and shooting were carried out for each evaluation point of the lines. The following are two pictures from the field research. In Fig. 2a, the horizontal safety distance between the line and the tree is too small; in Fig. 2b, the authors detected the overheating point of the line.

Figure  2.  Hidden danger pictures in field research: (a) the horizontal safety distance between the line and the tree is too small; (b) the overheating point of the line is detected. (Photographed by the authors)

DownLoad: Full-Size Img PowerPoint

In addition, based on the AHP method, the indicator weight results are obtained through expert consultation and Eq. (1) and are shown in Table 1. Combining the results of the indicator weights and the research data, the line and secondary indicator scores were obtained via Eq. (2). The scores of each secondary indicator and the analysis of the results are given in Sections 3.1.1 to 3.1.5.

3.1.1   Component structure health condition

The main structure scores of the 20 transmission lines obtained through field research are shown in Fig. 3a. The main reason for the low score of the main structure of the line is that the wire- and cable-type materials and insulator-type materials do not meet the requirements. Data analysis shows that cables 8 and 9 have the lowest scores, because these cables are bare wires and synthetic insulators, which are not resistant to high temperature. In the process of operation, a bare wire is more prone to aging and failure, especially under the influence of stormy weather. A line that experiences a lightning strike prompts line flashover, and tree branches and leaves cause two-phase line shortening and corona discharge and obvious leakage current, with a high risk of hidden danger. Under fire thermal effect conditions, high temperatures cause insulator string melting or loss of mechanical strength and electrical strength; the lower the degree of insulation loss is, the heavier the wire is, and grounding when an electric spark (arc) may also trigger a new burning point again.

Figure  3.  The secondary indicators score. (a) Component structure health condition score; (b) operation condition score; (c) surrounding fire hazard condition score; (d) technical protection measures condition score.

DownLoad: Full-Size Img PowerPoint

Field research has revealed that all line towers use cement poles, which are not the best choice for distribution line towers, whereas composite towers have better performance indicators in terms of strength, stiffness, fatigue resistance and aging resistance^[43]. According to the data provided by the power supply company to which the line belongs, all major components of the line are in a stable operation period of 1−50 years, and the risk is relatively low.

A further study revealed that among the 20 lines investigated, 6 lines used synthetic insulators, all of which were in region A. Most of the insulators in the remaining regions were fiber-reinforced polymers or ceramic insulators. Nine lines were insulated at 100%, and nine others were insulated at less than 10% (Table 4), with a higher frequency of bare conductors in region A.

3.1.2   Operation condition

On the basis of field research and historical operation and maintenance data, all 20 lines were free of overheating and leakage. Trip rates of 19.58 times/(100 km·year) and 13.13 times/(100 km·year) for Lines 1 and 2 in region A, respectively, were too high. The rest of the trip rates are low.

According to the standard^[24] on the requirements of the maximum grounding resistance, i.e., 10 kV distribution lines without lightning conductors, reinforced concrete poles in residential regions should be grounded, metal pipe poles should be grounded, and grounding resistance should not exceed 30 Ω. Research has revealed that a total of eight lines of grounding resistance values do not meet the requirements of the specification (Table 4); these lines are region B (12−14) and region C (16−20) company subordinate lines. Among them, five lines in region C have grounding resistance values that are too high.

3.1.3   Surrounding fire hazards condition

Tree obstructions are a major problem in power distribution line accidents^[44]. It is also necessary to consider the width of combustible materials and fire roads around the line, so the perimeter safety hazard indicators include the horizontal and vertical safety distances of the distribution line to the tree barrier, the width of the perimeter road, and the presence of combustible materials.

In this research, it was found that all 20 lines have a large risk of hidden tree barriers. Tree barrier accidents can have a serious impact on distribution lines, which can lead to short-circuiting and burn the lines. At the same time, flammable debris was found around the lines, and debris with a low ignition point may become a fuse for forest fires caused by problems such as line heating and lightning strikes, which are hidden dangers that threaten the safe operation of the lines themselves. Tree barrier accidents consider mainly the distance between the conductor and street trees. We selected the horizontal and vertical safety distances as assessment index specification requirements^[24], in which the vertical distances for the maximum arc drop situation of the bare wires and insulated wires are 1.5 and 0.8 m, respectively, and the horizontal distances for the maximum wind deflection situation are 2.0 and 1.0 m, respectively.

According to the statistics of all the assessment units of the field research, the problem of the horizontal vertical safety distance among the external influencing factors is more serious, with more than 60% not meeting the required distance. In addition, 31.55% of the regions have a perimeter road width of less than 4 m. The situation is better in region C, where three lines have a perimeter road width that meets the requirements of the standard^[25]. In the case of forest fires, effective fire rescue accessibility cannot be achieved, and fire rescue forces cannot arrive at the scene to control the fire in the first place.

3.1.4   Technical protection measures condition

Line protection technology can respond proactively to reduce the consequences of accidents and the degree of impact after a fault occurs and is the last line of defense for line safety. Owing to the different line management methods and operations in each region, the application of protection technology to the line is related to its region, and the score is shown in Fig. 3d. Line 1 in region A scores the lowest, and the line only has fault indicators, lightning arrester and disconnect switches, and other protection technologies are missing; region A has five lines (2−6) that do not have fault indicators and overvoltage protectors; region A has another five lines (7−11) that do not have fault indicators, overvoltage protectors, bird repellents, column load switches, and disconnect switches; regions B and C have better protection technology, and the highest 4 lines (12−14, 20) are in regions B and C, and the problem is not set up as bird repellents.

3.1.5   Management condition

Because of the special operating environment of the distribution lines across the forest region itself, as well as the importance of ensuring power supply, it is vital to ensure its safe and stable operation, and the implementation of a complete management system can improve the stability of the lines. In addition to Line 1, the management techniques of each line can be implemented in place as needed, and the systems can be implemented in place. In the review of management information, Line 1 failed to implement the technical information system, operation analysis system, materials and warehouse management system. In the process of operation, 10 kV distribution lines are affected by subjective and objective factors, and there is irresistible risk. Strengthening the management of the line can find out the hidden danger in time and solve the problem effectively.

3.2   Bayesian network construction and analysis

Bayesian inference is performed via GeNIe software, and finally, a BN assessment model of forest fires induced by 10 kV overhead distribution lines across forest regions is obtained, as shown in Fig. 4.

Figure  4.  Bayesian network assessment model for region C.

DownLoad: Full-Size Img PowerPoint

According to the Bayesian network, the annual forest fire probability and the annual occurrence probability of each subevent in southern Anhui and the subregion shown in Fig. 5 can be obtained. The annual probabilities of “forest fire accidents” in the four regions are 4.34%, 5.7%, 9.98% and 5.88%, respectively. Among them, the annual probability of “forest fire accidents” in regions A and B is low, and the main difference between them lies in “technical protection failures”, with 86.82% in region A and only 49.61% in region B. The annual probability of “forest fire accidents” in region C is high, mainly because the probability of occurrence of various risk factors is high in the regional ranking, the probability of “component structure failure” is 35.69%, and those of other regions are 18.12%, 12.21% and 21.6%, respectively.

Figure  5.  Probability of events in the overall region and subregions.

DownLoad: Full-Size Img PowerPoint

According to the above results, the annual probability of “forest fire accidents” in southern Anhui is low, but there is still much room for improvement. The probability of “technical protection failure” in each region is generally high, and there are “excessive surrounding safety hazards” in some regions. Therefore, targeted risk investigation and rectification should be carried out in southern Anhui, and the rectification should be aimed mainly at the lines in region C. At the same time, the main structure of the line should be investigated and replaced in region C.

3.3   Critical analysis

To better determine the applicable measures to prevent forest fires caused by distribution lines in southern Anhui, it is necessary to determine the most critical risk factors leading to forest fires. The FV value of the root node of each region, as shown in Fig. 6, can be obtained via Eq. (4).

Figure  6.  FVs of the root nodes in the overall network and subregions.

DownLoad: Full-Size Img PowerPoint

3.4   Regional and line-additional rectification measures

According to the FV value of the root node in each region, we can determine the most critical risk factors in each region. Combined with the key analysis results and the scores of each line in Section 3, the most effective rectification measures can be proposed for subregions and lines with low scores. The rectification measures for each region are shown in Table 5.

Table  5.  Regional rectification measures.

Region	Regional rectification measures	Extra measures of lines
A	During daily maintenance, the flammable sundries around the distribution line shall be cleaned in time, and the surrounding roads shall be widened. (Factor 3 and 4)	Line 1: (1) Investigate the cause of tripping and reduce the tripping rate. (2) Install line protection equipment. (3) Improve and implement the missing management system.
		Line 2: (1) Investigate the cause of tripping and reduce the tripping rate. (2) Replace insulated conductor and disc porcelain insulator.
		Line 8: Replace insulated conductor and disc porcelain insulator.
		Line 9: Replace insulated conductor and disc porcelain insulator.
		Line 11: Install line protection equipment.
B	(1) During daily maintenance, the flammable sundries around the distribution line shall be cleaned in time, and the surrounding roads shall be widened. (Factor 3 and 4) (2) Check the excavation of the grounding body of the tower, and relay the grounding ray[28]. (Factor 13) (3) Install line protection equipment especially bird repellent. (Factor 19)	Line 12: No extra measures.
C	(1) During daily maintenance, the flammable sundries around the distribution line shall be cleaned in time, and the surrounding roads shall be widened. (Factor 3 and 4) (2) Check the excavation of the grounding body of the tower, and relay the grounding ray[28]. (Factor 13)	Line 16: Remove trees that exceed the safety distance requirements.
		Line 17: Remove trees that exceed the safety distance requirements.

 | Show Table

DownLoad: CSV

4.   Conclusions

In this study, a novel assessment framework for the regional risk of forest fires induced by distribution lines is proposed. After the assessment framework is proposed, this paper identifies the risk factors for forest fires caused by 10 kV overhead distribution lines in the southern Anhui region and scores the assessment units by combining survey data and hierarchical analysis. Moreover, the annual probability of forest fire accidents on distribution lines in each region is derived via GeNIe based on Bayesian network. The following conclusions are drawn:

(I) The annual probability of forest fires induced by overhead distribution lines in the southern Anhui forest region is low, at 5.88%. The annual fire probabilities in regions A and B are 4.34% and 5.7%, respectively, but the annual fire probability in region C is relatively high, accounting for 9.98%. During daily operation and maintenance and risk investigations in southern Anhui, the lines in region C should be rectified.

(II) A lack of technical protection devices and excessive surrounding security risks are common and urgent problems in the southern Anhui forest region. Too many surrounding security risks are also key risk factors for regional forest fires. In daily operation, maintenance and risk investigations in southern Anhui, these two problems should be rectified.

On the basis of the assessment results, this paper proposes targeted measures for links that urgently need to be rectified in various regions, which is highly important for risk management and control of forest fires in southern Anhui. The assessment framework uses the low-scoring lines identified by the AHP and the regional analysis results of the Bayesian network to identify not only the hazardous events in the assessment region that require focused attention and region-wide corrective actions but also additional problems with the low-scoring lines in the region. Compared with previous methods, this assessment framework does not rely on historical fire data but uses more readily available field research data and historical operation and maintenance data, increasing the difficulty of assessment. Moreover, compared with the risk assessment of a single line, the assessment of regional risk is more instructive and universal. The assessment framework proposed in this paper is also applicable to other regions with different landforms and can provide a reference and ideas for the risk management and control of forest fires in other forest regions.