Computer-implemented method and system for preventive maintenance of network components

Abstract

The invention relates to a computer-implemented method for preventive maintenance of components of a network (10), wherein in the method the network (10) is divided into selected subsections, for which subsections dynamic, individual inspection intervals or renovation intervals are determined based on values of selected variables (12) describing the network (10) subsection. In the method - the variables (12) affecting the inspection interval or renovation interval are selected - the variables (12) are given a selected number of starting values differing from each other and the inspection intervals or renovation intervals corresponding to these starting values, - the impact of the dependence between the values of the selected variables (12) on the inspection interval or renovation interval is determined by a computational analysis, wherein the dependent variable is the inspection interval or renovation interval, and the variables (12) and / or combinations of the variables (12), such as sums and / or products of the variables (12), are used as independent variables, wherein the starting values are used in the computational analysis, and wherein the computational analysis is used to determine the inspection interval or renovation interval as a function of the variables (12), after which in the method - the values of the variables (12) are determined for a single subsection of the network (10) - an individual inspection interval or renovation interval is determined for the single subsection of the network (10) using the function determined by computational analysis, wherein the function is entered with the values of the variables (12) of the subsection in question, - a single subsection of the network (10) is inspected or renovated according to the specified inspection interval or renovation interval. The invention also relates to a corresponding system.

Description

[0001] COMPUTER- IMPLEMENTED METHOD AND SYSTEM FOR PREVENTIVE MAINTE-

[0002] NANCE OF NETWORK COMPONENTS

[0003] The invention relates to a computer-implemented method for the preventive maintenance ooff nneettwwoorrkk components, wherein the method comprises dividing the network into selected subsec- tions, for which dynamic, individual inspection intervals or renovation intervals are specified based on the values of the selected variables describing the network subsection . The in- vention also relates to a corresponding system.

[0004] Preventive maintenance of various networks, which here gener- ally refers to various linear infrastructures, such as elec- tricity network, domestic water network, railway network, road network, district heating network, telecommunication cable net- work, gas pipeline network or oil pipeline network, is imple- mented in the prior art solutions in principle on a time basis .

[0005] In other words, the network here refers to the network used to transfer the selected commodity, like electricity, water, oil, gas, heat, data or vehicles .

[0006] Such a network, like an electricity network managed by a network company, is a valuable asset and its special feature is the long lifetime of the components and equipment, which in prac- tice is tens of years . The goal of preventive maintenance of the electricity network is to keep the various components in operational condition, but on the other hand, the maintenance measures directed at them also affect their techno -economic lifetime . For example, in Finland , the Electricity Market Act ( 09 Aug 2013 / 588 § paragraph 51 ) obliges network companies to develop and maintain their network so that customers in urban development plan areas do not experience power outages of more than 6 hours . In rural areas, customers must not experience outages of more than 36 hours . In addition, the law obliges network companies to prepare for new loads and production fa- cilities.

[0007] Preventive maintenance of the electricity network is based on the periodic inspections of the electricity network subsections and the necessary corrective measures . As a state of the art, the inspection of the electricity network is most commonly done on a time basis, whereby the electricity network subsections are inspected according to a predetermined regular inspection interval. By moving from a time-based to a need-based inspec- tion interval, the maintenance of the electricity network can be made more efficient and investments can then be allocated elsewhere instead of inspections to improve supply security. Need-based, i.e. reliability-based, maintenance is based on the assumption that each network component has a defined priority value and its condition is known at some measurable level - for example, based on condition information obtained from inspec- tions. However, no effective solution has been presented in the prior art for determining the need-based inspection interval.

[0008] The time-based inspection model does not take into account different circumstances, but treats the entire operating area of the network company equally. However, from different per- spectives, the network is not equal. Part of the network is more significant in terms of failure, as a larger number of electricity users will be affected by the failure. An old and poorly maintained network is more prone to failure than a new one. There are also differences in electricity users; for ex- ample, the undisturbed use of electricity in hospitals is so- cially more significant than in secondary residences.

[0009] Preventive maintenance aims to ensure that a failure never occurs. It can be done, for example, through various inspec- tions, vegetation management or periodic maintenance. Time- based preventive maintenance is suitable for those components of the network that affect the operational reliability of the network - i.e. can cause an outage if they fail. The traditional time-based model applies, for example, to overhead wires, which according to electrical safety regulations must be inspected every five years or more frequently. Inspections can be done from the ground or from the air, e.g. with helicopters. The poles must also be inspected regularly after 20 years of ser- vice, but no minimum inspection cycle is specified.

[0010] In condition-based maintenance, the network component is in- spected at certain intervals to determine its condition. If deficiencies are detected during the inspection, the component can either be serviced or replaced. On the other hand, it can also be an inspection wherein it is stated that a component is working correctly - for example, the correct operation of pro- tection relays can be checked with measurements.

[0011] The problems of time-based preventive maintenance are common to all networks.

[0012] At the prior art the following publications are known:

[0013] - YUMBE, Y. et al. Optimization Method for Inspection Schedul- ing of Power Distribution Facilities. In IEEE Transactions on Power Delivery, July 2013, Vol. 28, no. 3, 1558-1565, <DOI : 10.1109 / TPWRD.2013.2253806>

[0014] - US 2014324506 Al

[0015] - YUMBE, YY.. et al. Evaluation of Optimization Method for In- spection Scheduling of Power Distribution Facilities Using Maintenance Data Accumulated by Power Utility. In IEEE Trans- actions on Power Delivery, April 2017, Vol. 32, no. 2, 696-702, <DOI : 10.1109 / TPWRD.2016.2618936>

[0016] In the publications mentioned above, efforts have been made to optimise the preventive maintenance of the electricity network by defining the scheduling of inspection visits . However, these publications do not provide a workable solution that would take into account the effect of the dependence between the values of several non-linear variables on the inspection interval.

[0017] The purpose of this invention is to provide a new computer- implemented method and system for performing preventive mainte- nance of the network on reliability basis, and for improving the quality of network maintenance. The characteristic features of the method according to the invention are set out in the attached claim 1, and the characteristic features of the system are set out in the attached claim 21.

[0018] In the computer-implemented method according to the invention, for preventive maintenance of network components, the network is divided into selected subsections, for which dynamic, indi- vidual inspection intervals or renovation intervals are speci- fied based on the values of the selected variables describing the network subsection. In the method, the variables that af- fect the inspection interval or renovation interval are se- lected, the variables are given a selected number of starting values differing from each other and the inspection intervals or renovation intervals corresponding to these starting values, the impact of the dependence between the values of the selected variables on the inspection interval or renovation interval is determined by computational analysis, wherein the dependent variable is the inspection interval or renovation interval and the independent variables are variables and / or combinations of variables, such as sums and / or products of variables, wherein the starting values are used in the computational analysis, and wherein the computational analysis determines the inspection interval or renovation interval as a function of the variables, after which the values of the variables for a single subsection of the network is determined, an individual inspection inter- val or renovation interval is specified for a single subsection of the network using the function determined by computational analysis, into which function the values of the variables of the relevant subsection are entered, a single subsection of the network is inspected or renovated according to the specified inspection interval or renovation interval.

[0019] In this way, the need-based inspection interval or renovation interval of the electricity network or other network can be determined in practice and the preventive maintenance of the electricity network can be enhanced. In the reliability-based model according to the invention, the inspection of the network subsection can be done less often in a subsection that is classified as non-critical, has a small number of failures and is in good condition. With the preventive maintenance according to the invention, financial benefits are achieved by optimising the inspection intervals of the network subsection. For exam- ple, subsections, which are in good-condition and less criti- cal, can be inspected less often than in the regular time-based method. On the other hand, inspections can be performed on the certain network subsections more often than the time-based method would determine, if the reliability-based inspection interval according to the invention is shorter than the time- based interval, whereby the added preventive maintenance pre- vents network failure more effectively. In addition, the method has an ecological benefit, because thanks to the method the total number of inspections on the network can be reduced in the long term reducing the total emissions caused by the in- spections .

[0020] The function determined by the method according to the inven- tion can be used for the entire network as part of maintenance planning. The function defined in this way is easily modifiable and supports the addition of new data. One key advantage of the preventive maintenance implemented in accordance with the re- liability-based model according to the invention is a cost- reducing effect compared to the time-based model of the prior art.

[0021] The method according to the invention is particularly suitable for inspecting overhead wires in the electricity network. Ac- cording to the failure statistics of the energy industry (2020) , the biggest cause of outages experienced by customers is with- out exception a failure in overhead wires. The method according to the invention can also be applied to the preventive mainte- nance of other components of the electricity network.

[0022] In the medium voltage network, the inspection visit covers every cable span in the subsection, including the poles, In addition to overhead wires, especially the condition of poles should be monitored. The condition of the poles is visually assessed when the overhead wires are inspected. Regarding the poles, in addition to the visual general condition, the points to be checked are: top rotting, pole caps, tilting, insulation damages, wire fixings, insulation damages, stayed supports, safety signs and warning tapes, and the condition of the earth- ing. After approximately 20 to 30 years, the poles should be inspected for rotting.

[0023] The inspection interval can be determined, for example, by classifying different network subsections based on criticality, failure rates and condition information. By defining the ex- amined variables and their weights, the inspection needs for the entire network can be calculated. Based on the calculated results, the inspections can be carried out as such or logis- tically rationalised by making changes . So it is partly a lo- gistical method for optimising the inspection of the network subsections.

[0024] With the computational model of the invention, dynamic inspec- tion entities are created for the network subsections priori- tising the most important objects in the network in terms of the security of use and society for more frequent inspection intervals. The network subsections in better condition, which do not have significantly critical elements, can be inspected less often. The dynamism of the computational model is due to the fact that the network data is constantly changing and the method reacts to changes when the data is updated. For example, new critical customers may join part of the network affecting the criticality review. In the same way, the network gets older every year and, for example, in the electricity network, new rotten poles are revealed during inspections.

[0025] The computational model according to the invention is built so that variables can be added to and removed from it. That is, as the model develops, new variables can be introduced into it making the computational model even more accurate.

[0026] The method can be used to determine the inspection interval or renovation interval. Renovation refers here broadly in addition to the inspection to all measures related to the maintenance of the network, such as service, maintenance, renovation and refurbishment. Renovation interval means that single subsec- tions of the network can be assigned individual index values that describe reliability and renovation need, on the basis of which, for example, the resources used can be directed to the renovation of the subsections with the greatest need for reno- vation. In this way, the method can be used to determine the reliability of the network subsections and to renovate the subsections with the weakest reliability. When referring above and in the following to the inspection interval, it can generally also mean the renovation interval, because the de- termination of the inspection interval and the renovation in- terval are equivalent in the terms of the method according to the invention.

[0027] The step of the method, wherein the variables are given a selected number of starting values differing from each other and the inspection intervals or renovation intervals corre- spending to these starting values, refers to expert calibra- tion, wherein the computational model of the method is cali- brated based on the need for preventive maintenance assessed by the network administrator. In other words, inspection in- tervals corresponding to combinations of the values of the selected variables are determined based on research data or mathematical analysis or data clustering related to each vari- able, and this expert calibration is used to generate the func- tion used to calculate the inspection interval .

[0028] There is no lower or upper limit to the number of variables used in the method, but the method can use a selected number of variables describing the electricity network subsection.

[0029] The number of variables can be at least 2, preferably at least 3. In this way, the accuracy of the method can be increased.

[0030] Preferably, a variable describing the condition information of the network subsection is used as one variable, which is a function of the properties of the components of the network subsection. In this way, the method can take into account the impact of the condition of the network physical components on the need for preventive maintenance, whereby objects with poorer condition can be inspected more often and objects with better condition can be inspected less often. Preferably, a variable describing the failure frequency of the network subsection is used as one variable, which is a function of the number of failures detected by the sensor in the sub- section. In this way, the method can take into account the impact of failures detected by sensors in the network on the need for preventive maintenance, whereby failure-sensitive ob- jects can be inspected more often and more reliable objects can be inspected less often. The variable that describes the fail- ure frequency can be, for example, the failure data detectable by a sensor from the network protective devices, storable and readable by memory tools, more precisely the failure frequency in a selected subsection of the network.

[0031] Preferably, a variable describing the topology of the network subsection is used as one variable, which is a function of the objects of use and the network type of the subsection. In this way, the method can take into account the impact of the number of socially or otherwise important objects and the number of the objects of use affecting on the subsection on the need for preventive maintenance. In other words, topology here refers to a criterion defined for the importance of the operation of the selected network subsection, which may be based on the number of the objects of use or the importance of the objects of use, to which the network subsection delivers the selected commodity.

[0032] Regression analysis can be used as a computational analysis. In this way, the impact of the dependence between variables on the inspection interval can be determined by statistical meth- ods using the starting values determined as expert calibration. Instead of regression analysis, the method can also use other known mathematical and / or statistical analysis methods.

[0033] Preferably, the least squares method is used in the regression analysis to determine the function based on the starting val- ues . In this way, the dependence of the variables used in the function on the starting values can be determined in a compu- tationally reliable and efficient way, and the method can eas- ily be made dynamic, whereby easy modifiability is achieved in terms of variable changes . Instead of the least squares method, any other suitable mathematical method can be used, such as maximum likelihood methods, Bayesian Estimation, Method of Mo- ments, Quantile Regression, Empirical Likelihood, Principal Component Analysis or Partial Least Squares.

[0034] A neural network can also be used as a computational analysis. For example, as a neural network, a forward-connected multi- layer network can be used, for which the above-mentioned vari- ables are used as input information. The neural network is trained with selected teaching material, i.e. expert calibra- tion, to take into account the connection between the values of the variables. The neural network can be any suitable prior art neural network software or library adapted to receive a selected number of starting variables. A neural network can be used to identify non-linear dependencies between variables in an alternative way.

[0035] Alternatively, the computational analysis can use any prior art solution to analyse the non-linear dependence between varia- bles.

[0036] The number of starting values of the variables and inspection intervals corresponding to the starting values given for the computational analysis can be 1-40%, preferably 5-20%, of all possible combinations of the starting values of the variables. In this way, the computational model can be calibrated with a small amount of work to match the selected variables, whereby the function determined by the method ccaann easily be used dynamically when the variables used in the model change.

[0037] Preferably, the determined individual inspection intervals or renovation intervals of the network subsections are visualised using a geographic information system. In this way, the indi- vidual inspection intervals of the network subsections can be visualised in a map-based system facilitating the logistic planning of the network operator in terms of predictive mainte- nance .

[0038] Preferably, at least one of the variables used in the method is assigned a value during the inspection of the network sub- section, which value is stored in the memory tool and read from the memory tool to determine the next inspection interval or renovation interval. In this way, feedback is provided to the method, whereby the method takes into account the impact of network aging and condition changes on the need for preventive maintenance.

[0039] Preferably, a flying device, such as a helicopter or drone, and a recording device, such as a camera or laser scanner, is used to inspect the network subsection. In this way, the condition of the network subsection can be easily checked and documented during the inspection visit as part of preventive maintenance.

[0040] Preferably, the network is an electricity network. Preventive maintenance of the electricity network benefits significantly from the reliability-based determination of inspection inter- vals in terms of costs and reliability, among other things.

[0041] Preferably, one of the variables is a variable describing the condition information of the electricity network subsection, which is a function of the characteristics of the components of the electricity network subsection, such as poles, like rotting and / or age of the poles. In this way, the method can take into account the impact of the condition of the network physical components on the need for preventive maintenance, whereby objects with poorer condition can be inspected more often and objects with better condition can be inspected less often.

[0042] Preferably, one of the variables is a variable describing the failure frequency of the network subsection, which is a func- tion of the number of failures detected by the sensor in the subsection and the power and energy of the electricity trans- mission in the subsection. In this way, the method can take into account the impact of failures detected by sensors in the electricity network on the need for preventive maintenance, whereby failure-sensitive objects can be inspected more often and more reliable objects can be inspected less often, The variable that describes the failure frequency can be, for ex- ample, the failure data detectable by a sensor from the network protective devices, storable and readable by memory tools, more precisely the failure frequency in a selected subsection of the network.

[0043] Preferably, one of the variables is a variable describing the topology of the network subsection, which is a function of the objects of use and the cable type of the subsection. In this way, the method can take into account the impact of the number of socially important objects and the number of the objects of use affecting the subsection on the need for preventive mainte- nance. A parameter describing the cable type can be, for exam- ple, the division of the electricity network into trunk and branch lines. If the electricity is cut off from the trunk line, the branch lines connected to it are also cut off, whereby method preferably emphasises the need for inspection of the trunk lines. Preferably, one of the variables is a variable describing the vegetation status in the electricity network subsection, In this way, the method can take into account the impact of the vegetation growth rate on the need for preventive maintenance, whereby subsections with nutrient-rich soil and fast-growing plants can be inspected more often than other subsections.

[0044] Preferably, 1 to 10 disconnector distances are used as a sub- section of the electricity network. In this way, the electric- ity network can be divided into easily manageable subsections. Alternatively, for example, a cable outlet or pole spacing can be used as the subsection.

[0045] The network can be one of the following: domestic water network, railway network, road network, district heating network, tele- communication cable network, gas network, oil pipeline network. The networks in question are linear infrastructures that bene- fit significantly from reliability-based preventive mainte- nance, e.g. in terms of maintenance costs and network relia- bility.

[0046] Preferably, the inspection interval or renovation interval of a single network subsection and the detected failures are rec- orded, and the recorded information is used when giving the variables a selected number of starting values differing from each other and the inspection intervals or renovation intervals corresponding to these starting values. In this way, a feedback is obtained for the calibration of the computational analysis, which can be used to monitor the functionality of the previous calibration and the function determined with it. In this way, the determination of the inspection interval or renovation in- terval can be optimised by taking into account the impact of the previous inspection interval or renovation interval on the observed failure frequency.

[0047] The system according to the invention for the preventive mainte- nance of network components comprises a computer, comprising software tools, a processor and memory tools, which are adapted to define dynamic, individual inspection intervals or renova- tion intervals for network subsections based on the values of selected variables describing the network subsection. The sys- tem comprises aa memory structure stored on computer memory tools, which comprises a selected number of starting values given to selected variables differing from each other and the inspection intervals or renovation intervals corresponding to these starting values, a first program code that can be arranged on the computer memory tools for execution by the processor, and which computer can be adapted to determine, by executing the first program code on its processor, the impact of the dependence between the values of the selected variables on the inspection interval or the renovation interval by means of a computational analysis, wherein the dependent variable is the inspection interval or renovation interval and the independent variables are variables and / or combinations of variables, such as sums and / or products of variables, wherein starting values are used in the computational analysis, and wherein the compu- tational analysis determines the inspection interval or reno- vation interval as a function of the variables, and a second program code that can be arranged on the computer memory tools for execution by the processor, and which computer can be adapted to determine, by executing the second program code on its processor, an individual inspection interval or renovation interval for a single subsection of the network using a function determined by computational analysis, to which function the values of the variables in the subsection in question have been entered. The advantages of the system according to the inven- tion correspond to the advantages of the method according to the invention described above.

[0048] Preferably, the system comprises display tools and a geographic information system to display the specified inspection interval or renovation interval. In this way, the individual inspection intervals or renovation intervals of the network subsections can be visualised in a map-based system, which facilitates the logistic planning of the network operator in terms of predic- tive maintenance.

[0049] Preferably, the system comprises a sensor for determining the value of at least one variable describing the network subsec- tion. In this way, measurable information is obtained about the condition of the network subsection, which can be used to de- termine the need for preventive maintenance inspections or the need for renovation.

[0050] The method according to the invention described above can there- fore be applied to the electricity network and other linear infrastructures, such as the reliability-based maintenance of the domestic water network, railway network, road network, dis- trict heating network, telecommunication cable network, gas pipeline network or oil pipeline network. These infrastructures are collectively referred to as the network. Common to the method used for the maintenance of these networks is that, as described above, the network is divided into small entities, i.e. subsections, that are meaningful from the point of as- set / condition management, and the selected analytical method is used to determine their reliability. Only the details re- lated to the selection of the variables to be used in the calculation and the determination of the value of the variables depend on the applicable network, the method and system accord- ing to the invention being applicable as such to any network. Preferably, for any network, the method uses as one variable the condition information of the network subsection, i . e . the variable describing the technical condition, as a second vari- able the failure frequency of the network subsection, i.e. the variable describing failure, and as a third variable the to- pology of the network subsection, i.e. the variable describing distribution importance . The risk caused by environmental con- ditions can be used as the fourth variable.

[0051] Examples of networks to which the method according to the in- vention can be applied and, in the case of such networks, the factors to be advantageously taken into account in the varia- bles used in the method, are listed below: o Domestic water network

[0052] ■ Technical condition

[0053] • Material / type of pipes used

[0054] • Age

[0055] • Possible inspection / measurement results

[0056] ■ Importance of distribution

[0057] • Is it a trunk line or a branch line?

[0058] • Does it transport water to a particularly important location? Hospital or similar.

[0059] ■ Failure

[0060] • Leaks

[0061] • Other mechanical failures

[0062] ■ Risk caused by environmental conditions o Railway network

[0063] ■ The method can be used either for condition classifica- tion, or if sufficient condition classification already exists, it can be used to prioritise inspections and other measures.

[0064] ■ Technical condition

[0065] • The condition of the sleepers • The condition of the rails

[0066] • Age

[0067] • Possible inspection / measurement results

[0068] ■ Importance of distribution

[0069] • How important is this connection?

[0070] • Speeds

[0071] • How critical the connection is in terms of freight and passenger traffic

[0072] ■ Failure

[0073] • Train delays

[0074] • Detected failures or condition observations

[0075] ■ Risk caused by environmental conditions o Road network

[0076] ■ The road network has already been classified, but the method can bring added value to the process of condition classification and inspections. On the other hand, measures and / or sections already identified as weak still need to be prioritised. For this purpose, the method produces added value.

[0077] ■ Technical condition

[0078] • Age

[0079] • Rut depth, profile

[0080] • Drainage

[0081] ■ Importance of distribution

[0082] • Road classes

[0083] • Driving speeds

[0084] • Traffic volumes

[0085] ■ Failure

[0086] • Frost damages

[0087] • Accident susceptibility

[0088] ■ Risk caused by environmental conditions o District heating network

[0089] ■ Technical condition

[0090] • Material / type of pipes used • Technical lifetime

[0091] • Mpul / channel structures

[0092] ■ Importance of distribution

[0093] • Trunk line or branch line

[0094] • Does it feed heat to important locations? o E.g. hospital, health centre, assisted living facility, retirement home

[0095] • Booster station

[0096] • Critical pipe sections for heat supply

[0097] • Supply lines for heat production o Connection lines of the district heating plant

[0098] ■ Failure

[0099] • Are there leaks / additional water consumption

[0100] • Damage frequency

[0101] • Can leaks be predicted? o Sensors installed in wells

[0102] • Leak monitoring cables for installation in pipe elements

[0103] ■ Risk caused by environmental conditions o Telecommunication cable network

[0104] ■ Technical condition

[0105] • Technical lifetime is about 50 years; installation method has only little impact on the cable

[0106] • Used installation materials and components and work methods

[0107] • Technical structure (land, air and water cables) , the cable is selected according to the purpose of use

[0108] ■ Importance of distribution

[0109] • Trunk connection or branch connection

[0110] • Does it feed locations that are important to soci- ety?

[0111] • Resilience, the network is secured or unsecured

[0112] ■ Failure

[0113] • How many failures have there been, material and environmental failures, i.e. failure frequency

[0114] • How many failures have there been caused by exter- nal agents, weather (winds and tree falls, water, ice and lightning) , traffic and machinery damages

[0115] • Failure duration

[0116] • Automatic failure repair process

[0117] ■ Risk caused by environmental conditions o Gas network

[0118] ■ Technical condition

[0119] • Technical lifetime

[0120] • Materials and components used

[0121] • Condition improvement measures made afterwards

[0122] ■ Importance of distribution

[0123] • Trunk connection or branch connection

[0124] • Does it feed locations that are important to soci- ety?

[0125] ■ Failure

[0126] • Leaks

[0127] • Amount of repair needed

[0128] • Other mechanical failures

[0129] ■ Risk caused by environmental conditions o Oil pipeline network

[0130] ■ Technical condition

[0131] • Technical lifetime

[0132] • Materials and components used

[0133] • Condition improvement measures made afterwards

[0134] ■ Importance of distribution

[0135] • Trunk connection or branch connection

[0136] • Does it feed locations that are important to soci- ety?

[0137] ■ Failure

[0138] • Leaks

[0139] • Amount of repair needed

[0140] • Other mechanical failures ■ Risk caused by environmental conditions

[0141] The examples above do not limit the application of the method, but the method and system according to the invention can be applied to the maintenance of any corresponding existing net- work.

[0142] Below are examples of how the values of the selected variables can be determined for different networks and how the values of each variable can be taken into account in the expert calibra- tion.

[0143] In the district heating network, leaks combined with the age of the pipeline, technical lifetime and material are largely determining factors. The importance of the distribution scales the reliability up or down. On the other hand, important con- nections themselves require a higher need for action.

[0144] In the road network, for example, high traffic volume and speed combined with poor road surface condition and relatively high accident rate indicate low reliability and high maintenance needs. Correspondingly, if the road is technically in good condition with relatively low traffic volume and driving speed, and there are not many accidents and reliability is high, a maximum long intervention interval ccaann be defined for the measures, taking into account the sector.

[0145] In the railway network, poor condition combined with driving speed and train delays cause low reliability and short-term maintenance needs. The importance of the connection scales re- liability, i .e. the need for measures up or down. On the other hand, important connections need to be maintained more often. In the telecommunication network, age and the material used, combined with failure rraattee,, are significant factors in determining reliability. The importance and resilience of the connection scales reliability in relation to other factors.

[0146] In the gas pipeline network and the oil pipeline network, the age of the network and the components used in it, combined with failure rate and repair needs, are largely determining factors. The sensors produce information about the need for repair. The importance of distribution scales reliability up or down. On the other hand, important connections themselves require a higher need for action.

[0147] In the case of different networks, the selected values of the variables used in the computational model can be determined with one or more sensors installed in the network, examples of which are given below.

[0148] Temperature, air humidity, water level height and vibration can be used as sensors in the district heating network.

[0149] In road traffic, at least optical fibre, optical sensor, under- road induction loop, camera equipped with image recognition technology can be used as a sensor. In addition, the condition of the road surface can be assessed with camera and laser scanning technologies installed in the vehicle.

[0150] In the railway network, laser scanners and image-based sensors can be used to assess the condition of the track.

[0151] Various water quality sensors, such as water pH value, elec- trical conductivity, temperature and ultrasound sensors can be used in the domestic water network to measure water quality and flow rate.

[0152] In the telecommunication cable network, sensors can be active devices in the network and failure reports from customers. Failure reports from the sensors go directly to the control room, and customer failure reports end up in the control room from customer service.

[0153] In the gas network and the oil pipeline network, pressure sen- sors, gas detectors, and volume flow and temperature sensors can be used to determine the value of one or more selected variables .

[0154] However, the method according to the invention is particularly suitable for the preventive maintenance of electricity network components, due to the massive volume of the electricity net- work, the long lifetimes of the electricity network, and to the fact that the electricity network has been built over several decades using different methods. The different scenarios are distributed heterogeneously throughout the electricity network, which makes it particularly useful to have the reliability of the electricity network analysed and the condition of the elec- tricity network subsections inspected using the method accord- ing to the invention.

[0155] The invention is described in detail in the following by re- ferring to the accompanying drawings depicting some applica- tions of the invention, in which

[0156] Figure 1 shows the steps of a method according to the inven- tion, wherein the computational model used to deter- mine the inspection interval is created,

[0157] Figure 2 shows the steps of another method according to the invention, wherein the computational model used to determine the inspection interval is created,

[0158] Figure 3 shows the steps of a method according to the inven- tion, wherein the determined computational model is used in the preventive maintenance of a single sub- section of the electricity network,

[0159] Figure 4 shows a visualisation of the inspection intervals determined by a method according to the invention for the electricity network subsections.

[0160] Figure 1 shows the steps of a method according to the invention, wherein the computational model, or function, used to determine the inspection interval of the network 10 subsections is cre- ated. The same example can also be applied to determine the renovation interval . In this example, network 10 is an elec- tricity network, but the presented method is suitable for other networks 10 as well. In the computational model, a calculated inspection interval can be determined for all parts of network 10 by subsections. Here, a disconnector distance is used as the electricity network subsection. Alternatively, for example, a cable outlet or pole spacing can be used as a subsection. In this model, trunk lines, critical customers, failure frequency and power are taken into account through the criticality index. The condition index takes into account the condition of the electricity network through the rotting and age of the poles.

[0161] In the first step 40 of the method, the variables 12 affecting the inspection interval used in the method are selected, which are examined for the electricity network subsection. In this application form, the computational model uses three variables 12 : condition information, failure frequency and topology. Here, the values of the variables 12 are composed of several examined network data, which have different weight values in forming the inspection needs.

[0162] First variable 12, failure frequency:

[0163] In terms of operational reliability, failures that have already occurred in the electricity network indicate the condition and susceptibility to failure of the disconnector distance in ques- tion. A failure here means an unplanned outage of the electric- ity network.

[0164] The failure frequency parameter combines the number of failures and the maximum power of the disconnector distance. Power has an effect on outage costs, and on the other hand, it also tells about electricity users: for example, secondary residences typ- ically have lower power compared to permanent residences. The failure frequency can also take into account, for example, the amount of energy transferred in the network subsection.

[0165] Here, regarding the number of failures, the numbers of failures in cable outputs are examined. Assigning the exact location to a single disconnector distance is also possible, provided that accurate location information is documented for all failure locations.

[0166] In addition to the amount of failures, the parameter examines the maximum power passing through the disconnector distance. In practice, disconnector distances of higher power are recom- mended for a more frequent inspection cycle even with a lower number of failures. Here, in terms of powers, prioritisation is only done for disconnector distances of above 0.5 MW. Dis- connector distances of powers lower than that are considered equal, whereby only the number of failures in the cable output is considered. A finer gradation can also be used for the power limits to refine the computational model.

[0167] Here, the value of the failure frequency variable is generated according to the table below.

[0168]

[0169] In the table above, the value of the failure frequency variable is presented as a recommended inspection interval. However, it is not necessary for the value to be in the unit of time, but here also a unitless index number can be used for the value, which is used later in the computational analysis, which fi- nally gives the inspection interval in the time unit as a function of several variables.

[0170] The number of failures for each disconnector distance is the number of failures of the relevant cable output, but is pro- portional to the total number of poles in the cable output. Thus, the model considers long and short cable outputs more equally. For example, if the cable output of 100 poles has had

[0171] 3 failures, the proportion of failures is 3%. Assuming pole spacing of 50 meters, a 3% ratio would be the same as 3 failures per 5 kilometres.

[0172] With regard to failure data, it is possible to further develop the contribution of the computational model by taking into account more detailed failure data and topological effects. In the computational model, it is possible to take into account, for example, the location information of the cable. For exam- ple, it is possible to record for network subsections whether the subsection in question is located in a forest, on the side of the road or in a field, and the location information can be used to further emphasise the probability of failure.

[0173] Second variable 12, topology:

[0174] Topology variable 12 combines the weighting of critical cus- tomers and the examination between trunk and branch lines. The weighting of the variable 12 in question is high, because fail- ures affecting trunk lines have typically a larger area of influence than failures in branch lines. Equally important is the weighting of the critical customers. For example, uninter- rupted electricity distribution to hospitals is of great im- portance to society.

[0175] In the computational model, the criticality of trunk lines is emphasised higher, because their role in terms of supply reli- ability is central . If the trunk line fails, the branch lines will also have no electricity. In the case of a single branch line, the effects of failure outage are typically limited to that branch line only, and electricity can be restored to other sections of the cable output.

[0176] Here, the value of the topology variable is generated according to the table below.

[0177]

[0178] In the table above, the value of the topology variable is presented as a recommended inspection interval . However, it is not necessary for the value to be in the unit of time, but here also a unitless index number can be used for the value, which is used later in the computational analysis, which finally gives the inspection interval in the time unit as a function of several variables.

[0179] The criticality of the disconnector distance is at its extreme when it is a trunk line and critical customers have been con- nected to it. If it is a trunk line without critical customers or a branch line that has connected critical objects, the in- spection interval is at a neutral level . Third variable 12, condition information:

[0180] Here, variable 12, which includes condition information, exam- ines the rotting and the age of the poles. The poles are subjected to rot inspections or more precise strength measure- ments, depending on the type of pole. A climbing ban can be used as a comprehensive reference value for all inspections. During rot inspections, the online information system calcu- lates the climbing ban by comparing the length and base thick- ness of the pole with the measured amount of rot. Based on the calculation, the online information system sets the pole to a climbing ban or allows climbing with support. In strength meas- urements, the climbing ban is set manually by the inspector based on the results of the measuring device. The variable considers poles that are allowed to be climbed with support as equivalent as those that are completely prohibited. In many cases, the poles of the disconnector distance in poor condition have both climbing prohibition attributes, which is why it is worth considering both in order for the model to reliably find the weakest disconnector distances.

[0181] In a situation, wherein less than 10% of the poles in the disconnector distance have not been assigned a climbing ban, the method only looks at the age of the poles. The installation time of the pole can be found on every pole in the network, with the exception of single documentation errors. With the age inspection of the poles, the condition information variable gives a recommended inspection interval for each disconnector distance, even when no climbing ban has been specified.

[0182] The condition information variable can also take into account, for example, any batch of components that proves to be unreli- able. If, for example, in the case of the electricity network, an insulator batch that has proven to be bad has been used, a greater need for inspection, i.e. a shorter inspection inter- val, can be determined for the subsections comprising this batch.

[0183] Here, the value of the condition information variable is gen- erated according to the table below.

[0184] In the table above, the value of the condition information variable is presented as a recommended inspection interval. However, it is not necessary for the value to be in the unit of time, but here also a unitless index number can be used for the value, which is used later in the computational analysis, which finally gives the inspection interval in the time unit as a function of several variables.

[0185] The comparison value takes into account the length of the dis- connector distance by looking at the number of poles. The ratio is calculated by dividing the comparison value by the number of poles in the disconnector distance. For example, if there are 50 poles in the disconnector distance and 10 of them are prohibited from climbing or climbing is allowed only with sup- port, the ratio is 20%.

[0186] The age of the poles does not directly indicate to the poor condition or the degree of rotting of the poles. The recommended lifetime of the poles is a maximum of 50 years, and the actual technical lifetime may be even lower. On the other hand, well- saturated poles can remain in good condition for well over 50 years. In the majority of cases, the age of the poles also indicates the age of the other components of the network, be- cause other components of the overhead wire network are typi- cally built at the same time as the poles. Thus, through the age of the poles, the age information of disconnectors, trans- formers, insulators and conductors is taken into account in the majority of cases. Over the years, various components have been used in the construction of overhead wires, the technical life- time of which may vary significantly. In some old lines, for example, wooden perches have been used, the lifetime of which are probably not as long as steel perches.

[0187] In addition to these three variables 12, also a fourth variable 12 can be used, which is illustrated in Figure 1 with a dashed line. As such, there is no lower or upper limit to the number of variables 12, but the method can use a selected number of variables 12 describing the electricity network subsection. However, the accuracy of the method can be increased by using at least two, preferably at least three variables 12.

[0188] In the second step 50 of the method, the variables 12 are given a selected number of starting values differing from each other and the inspection intervals corresponding to these starting values. It is an expert calibration, wherein the computational model of the method is calibrated based on the need for pre- ventive maintenance assessed by the network administrator.

[0189] The starting values are generated by entering manually the expected final results of the different variations of the val- ues of the selected variables 12 into the table. The more cases entered, the more precisely the calculation follows the desired final result. The values in the table can be changed, whereby the calculation will be modified accordingly. Easy adaptability enables changes to be made if, for example, when industry reg- ulations or the network company' s own policies change.

[0190] In determining the starting values of the computational analy- sis, feedback can also be used, wherein the inspection interval of a single subsection of the network 10 and the detected failures are recorded, and the recorded data is used when giving the variables 12 a selected number of starting values differing from each other and the inspection intervals corresponding to these starting values. In other words, the effectiveness of the used inspection interval on the failure frequency a single subsection of the network 10 can be determined and this data can be used to determine the starting values, which can be modified at any time during the application of the method. For example, if the inspection interval is long and there are many failures in the subsection, the inspection interval correspond- ing to the values of the variables in that subsection can be shortened, i.e. the inspection can be carried out more often in a shorter time interval. In this case, the computational model can be modified so that the model is adapted to give the values of the variables in question a shorter inspection in- terval .

[0191] An example of the starting values of the parameters is shown in the table below, wherein KT is the condition information, T is the topology, V is the failure frequency and TT is the inspection need, i.e. the inspection interval .

[0192] The expected final results of the table - i .e. the inspection years - have been determined based on an expert' s assessment. Here, the variables KT, T and V as well as TT can be given integer values between 1 and 5, corresponding to inspection intervals of 1 and 5 years. KT, T and V can be the unitless index numbers described above, while the unit of TT is a year.

[0193] In the third step 60 of the method, the effect of the dependence between the values of the selected variables 12 on the inspec- tion interval is determined by regression analysis, wherein the dependent variable is the inspection interval and the independ- ent variables are variables 12 and / or combinations of variables

[0194] 12, such as the sums and / or products of the variables 12, wherein the starting values determined above are used in the regression analysis, and which regression analysis is used to determine the inspection interval as a function of variables 12. Instead of regression analysis, any other known mathemati- cal and / or statistical computational analysis can also be used, like a neural network.

[0195] In order to increase the accuracy of the computational model, different combinations of variables 12 are used here as inde- pendent parameters of the regression analysis. For example, the combination of topology and failure frequency can have a dif- ferent weight value than if the parameters were separate. With combinations, for example, the combination of topology and con- dition information can give a different need for inspection than the combination of topology and failure frequency would give.

[0196] The variables 12 and combinations of variables 12 used as in- dependent variables in this regression analysis are: - KT

[0197] T

[0198] - V

[0199] KT+T

[0200] T+V

[0201] KT+V

[0202] KT+T+V

[0203] - KT*T

[0204] T*V

[0205] - KT*V

[0206] - KT*T*V

[0207] - KTA1.24

[0208] - TA1.24

[0209] - VA1.24

[0210] - Constant The purpose of combinations is to refine the calculation. By creating different combinations for variables 12, it is possi- ble to have different shapes for the curve formed by the cal- culation instead of just a straight line. The function defined by the constant term is adjusted to give non-zero inspection intervals for all the values of the variables 12. In the re- gression analysis, a value other than 1.24, such as 2, can be used as the power of the variables 12, but it has been found that the power 1.24 provides better accuracy for the calcula- tion. In other words, the individual inspection interval of each subsection of the electricity network = f (KT, T, V) .

[0211] Weight values are calculated for independent variables with regression analysis. Regression analysis and the above pre- sented starting values are used for the calculation. The start- ing values are calculated on a case-by-case basis and they may differ from those shown in the table above.

[0212] Here, the independent variables have been entered as a function into Python' s Optimization Curve Fit tool with the starting values tabulated above, whereby weight values are calculated for the independent variables using the least squares method. It is clear that other known computational tools can also be used to calculate weight values.

[0213] The function used in the regression analysis, and also the final function used to calculate the inspection interval, is of the form wherein p is the weight value and TTiis the independent vari- able. In other words, for each of the independent variables above, of which there are 15 here, weight values are calculated by regression analysis and the final inspection interval is obtained by summing each weighted value of the independent variable together.

[0214] Alternatively, the reliability of the disconnector distances can also be examined through the reliability index. The relia- bility index for disconnector distances can be calculated using the equation below.

[0215] The weight values obtained by the least squares method for the independent variables are shown in the table below.

[0216] Therefore, the inspection interval for a single subsection of the electricity network can be determined by assigning a value to each term in the table above, i.e. independent variable, weighting the value by the weight value described in the table above, and summarising the weighted terms. The results of the function are rounded to even numbers corresponding to inspec- tion intervals of 1 to 5 years. With this computational model, an inspection interval of 1 to 5 years is determined for the electricity network subsection, which is in accordance with the current regulations of the Finnish authorities. Other inspection intervals can also be used, for example extending to inspection intervals of more than 5 years, to meet regionally valid authority regulations or the need for network maintenance.

[0217] After calculating the weight values, an optional correction step can be performed if the inspection interval calculated by the model for some starting value differs from the inspection interval used as the starting value.

[0218] Predefined variations must follow the same logic so that all possible variations can be calculated reliably. If the prede- fined cases, i.e. the values entered as expert calibration, are in any way inconsistent with the inspection intervals given by the computational model, the calculation tools can be used to make the adjustments automatically. Therefore, the table en- tered as expert calibration can be updated if necessary after the creation of the computational model. The computational model determined in this way is used to calculate the inspection interval for each subsections of the electricity network 10.

[0219] With the regression analysis described above, which is here a four-dimensional regression analysis, the computational model of the method according to the invention can be determined in step 70 of the method. The computational model comprises the calculation of the inspection interval of the electricity net- work 10 subsection as a function of the selected variables 12.

[0220] One of the most important criteria of the computational model is its adaptability in future aspects. The method according to the invention enables the addition of new variables as quickly as possible, which is why the weight values are calculated here using the least squares method. The computational model allows new independent variables to be entered easily into the calcu- lation of weight values, so that the user does not have to go through every possible situation manually.

[0221] Alternatively, the computational model determined as described above can also be used to determine the renovation interval.

[0222] The numeric value obtained as the output of the function can then describe the renovation interval in years, or the numeric value can be used as an index number that describes the need for renovation of the network subsection. In this case, the index number can be used to direct resources to the network subsections with the greatest need for renovation, In this case, a short renovation interval or a small index number given by the function can mean a high need for renovation.

[0223] Figure 2 shows the steps of another method according to the invention, in which the computational model used to determine the inspection interval of the electricity network 10 subsec- tions is created. Figure 2 differs from Figure 1 only in that, instead of regression analysis, a neural network is used as a computational analysis to determine the non-linear dependence of the selected variables.

[0224] In general, neural network solutions are technically able to solve the problem of combining non-linear variables. However, the critical factor is the background work, wherein the influ- encing factors, threshold values and their key weighting are determined, in other words, the expert calibration described above, which is used for training the neural network. Selected variables 12 are used as inputs to the neural network, for example the above described failure frequency, topology and condition information. The structure of the neural network can be variable. For exam- pie, the ready-made scikit-leam library can be used, and more precisely, the neural network created with its MLP regression feature, i.e. multi-layer perceptron regression feature. This library is quick to implement. Other similar libraries include e.g. Tensorflow and pytorch. It is also possible to program a self-made corresponding neural network program code. The number of layers and neurons can be retrieved through data iteration. The activation function to be used is determined for each cell, which can be, for example, a ReLU function, i.e. rectified linear unit activation function, more precisely the function f (x) = max (0,x) .

[0225] The selection of the neural network architecture determines the mathematical function, to which the parameters of that function are determined during the training phase of the neural network, including parameters such as weight values and other factors related to the neural network architecture. The function pro- duced by the neural network, which is used to calculate the inspection interval of a single subsection, is not typically written open, but the function produced by the neural network taught by expert calibration can be understood herein as an algorithm, to which is given as input data, i.e. independent variables, the values of the selected variables 12, which var- iables 12 correspond to the variables 12 used in the expert calibration, and which algorithm produces an inspection inter- val corresponding to the values of the variables 12 in question.

[0226] The parameters for forming the neural network are obtained by expert calibration of the model . Expert calibration can be replaced by another mathematical or data analysis method as well, wherein calibration data is generated by analysis rather than by expert analysis. Calibration data is used as the input data of the neural net- work, which can be produced either traditionally by expert calibration, by mathematical methods or, for example, by clus- tering data into certain groups.

[0227] The scikit-learn library used in the example can be asked, for example, with the following program code to create a neural network: from skleam.neural_network import MLPRegressor file = "D: \PTVY.xslx*

[0228] X = np. array (read_excel (file) [ : , 0: 3] ) y = np. array (read_excel (file) [ : , -1] ) regr = MLPRegressor (hidden_layer_si zes= (8, 4) , random_state=l , max_iter=1000) . fit (X, y)

[0229] The above term file = "D: \PTVY. xslx" comprises the expert cal- ibration, i.e. the memory structure stored on the computer memory tools, which comprises the selected number of starting values differing from each other given to the selected varia- bles 12 and the inspection intervals corresponding to these starting values. Expert calibration can be carried out in the case of the neural network in the same way as in the case of the regression analysis described above.

[0230] Figure 3 shows the use of the computational model to determine the inspection interval for a single subsection of the network 10.

[0231] In step 45 of the method, the values of the variables 12 are determined for a single subsection of the network 10. The value of the variables 12 can be determined automatically, for exam- ple, based on the measurement data of a sensor in the network 10, which measurement data is stored in the database with memory tools and which measurement data is read from the database and entered into the computational model, or the measurement data can be based, for example, on a manual measurement performed during an inspection visit.

[0232] In step 55 of the method, the values of the variables 12 are entered into the computational model defined above, i . e . into the function determined by regression analysis or a neural network or other suitable computational analysis, In other words, with the computational model, an individual inspection interval is determined for a single subsection of the network 10 using the function defined above, into which the values of the variables 12 of the subsection in question are entered.

[0233] In step 65 of the method, an individual inspection interval is thus obtained for a single subsection of the network 10 by using the computational model defined above.

[0234] The computationally determined inspection interval can be vis- ualised by a map-based system in alternative step 75 of the method.

[0235] After this, an inspection of a single subsection of the network 10 is performed as part of preventive maintenance according to the specified inspection interval in step 85 of the method.

[0236] In alternative step 95 of the method, a new value can be de- termined for at least one variable 12 during the inspection visit, which new value can be used in the determination of the next inspection interval.

[0237] Similarly, in step 65, the renovation interval can be deter- mined, and step 85 can be a timed renovation.

[0238] Figure 4 shows the network 10, which is an electricity network, divided into subsections, the network comprising trunk lines 100 and branch lines 110. Individual inspection intervals have been calculated for the electricity network subsections, which comprise an inspection performed every 1 to 5 years on a single subsection as part of the preventive maintenance of the elec- tricity network 10. The individual inspection intervals of the subsection are visualised in Figure 4 with different lines. Alternatively, different colours can be used for visualisation, for example, corresponding to different inspection intervals,

[0239] In addition, a map base can be added to the visualisation according to Figure 4 to show the location of the electricity network in the terrain.

[0240] The presentation of inspection needs is done, for example, in the Arc-Gis program, wherein each disconnector distance is vis- ualised with a different line according to the inspection needs (Figure 4) .

[0241] An alternative to visualisation is, for example, tabulation of the inspection intervals and inspection of the network subsec- tions according to the tabulation.

[0242] Correspondingly, the map base according to Figure 4 can show the need for renovation of the network 10 subsections or an index based on the reliability of the network 10 subsections.

[0243] The system according to the invention for the preventive mainte- nance of the components of the network 10 comprises a computer comprising software tools, a processor and memory tools, which are adapted to determine dynamic, individual inspection inter- vals or renovation intervals for the network 10 subsections based on the values of the selected variables 12 describing the network 10 subsections.

[0244] The system comprises - a memory structure stored on the computer memory tools, which comprises a selected number of starting values differing from each other given to the selected variables 12 and the inspection intervals or renovation intervals corresponding to these start- ing values,

[0245] - a first program code that can be arranged on the computer memory tools for execution by the processor, and which computer can be adapted to determine, by executing the first program code on its processor, the impact of the dependence between the values of the selected variables 12 on the inspection interval or the renovation interval by means of a computational analy- sis, wherein the dependent variable is the inspection interval or renovation interval and the independent variables are vari- ables 12 and / or combinations of variables 12, such as sums and / or products of variables 12, wherein starting values are used in the computational analysis, and wherein the computa- tional analysis determines the inspection interval or renova- tion interval as a function of the variables 12, and

[0246] - a second program code that can be arranged on the computer memory tools for execution by the processor, and which computer can be adapted to determine, by executing the second program code on its processor, an individual inspection interval or renovation interval for a single subsection of the network 10 using a function determined by computational analysis, to which function the values of the variables 12 in the subsection in question have been entered.

[0247] The computational model described above can be combined with vegetation management. For the coming years, it is possible to predict the growth rate for the different pole spacing of the network, when the vegetation and soil types are known.

[0248] The growth rates of vegetation can be modelled computationally with its own calculation model . It is known at the prior art, that each different soil and vegetation type has its own cal- culated growth rate. The effect of vegetation on the inspection of the electricity network subsection is described, for exam- ple, in the article Siipilehto J. , Sauvula-Seppälä T. , Lehtonen M. , Nykänen M. , Hynynen J. (2022) . Maximum length of vesas using quantile regression. Journal of Forestry, volume 2022, article 10704. https: / / doi.org / 10.14214 / ma.10704.

[0249] In the computational model, the inspection intervals are be- tween 1 and 5 years, in which case it is essential to take into account such disconnector distances, in which the vegetation grows above the threshold values in less than five years, In practice, the disconnector distance that would have an inspec- tion interval of, for example, five years, but the need for vegetation clearance would be every four years, should be in- spected earlier due to the rapid growth rate.

[0250] The vegetation status can be taken into account in the compu- tational model as its own variable 12. IInn pprraaccttiiccee,, however, the vegetation status would then guide the determination of inspection times, whereby it would be done either a year before the anticipated need for vegetation clearance or the following year after the vegetation clearance. Typically, in precision clearance, clearance needs are identified according to the re- sults of inspections, whereby it would be natural to perform the inspection before vegetation management measures.

[0251] If the current status and future growth of the vegetation are taken into account in the model as such, it has a dominant effect on other variables 12. The vegetation must not hit the line in any situation, which would affect the computational model as an absolute parameter. In other words, the contribu- tion of other variables 12 would be very small in those areas where the vegetation grows very quickly. Then the inspection would have to be scheduled in relation to the vegetation clear- ance, and the network reliability-based model would not be fully exploited.

[0252] In the present application, the vegetation status does not affect the reliability-based computational model of the dis- connector distances, i.e. the vegetation status is not used as a variable 12. Instead, the vegetation is treated as a separate element according to the computational model presented in the aforementioned article. Vegetation clearance needs can be pro- cessed in parallel with network inspection needs, allowing the user to optimise the calculated inspection times in relation to vegetation clearance times, if necessary. The predictability of vegetation clearance needs and the dynamism of inspection intervals together create a model that optimises maintenance planning. The vegetation growing area is cleared and inspected at optimal intervals, but operational safety is not substan- tially impaired. However, it is possible that the vegetation status is used as one variable 12 in the computational model according to the invention.

Claims

Claims1. A computer- implemented method for preventive maintenance of components of a network (10) , wherein the method comprises dividing the network (10) into selected subsections, for which subsections dynamic, individual inspection intervals or reno- vation intervals are determined based on values of selected variables (12) describing the network (10) subsection, charac- terised in that in the method the variables (12) affecting said inspection interval or renovation interval are selected,- the variables (12) are given a selected number of starting values differing from each other and the inspection intervals or renovation intervals corresponding to these starting values,- the impact of the dependence between the values of the se- lected variables (12) on the inspection interval or renovation interval is determined by a computational analysis, wherein the dependent variable is the inspection interval or renovation interval, and said variables (12) and / or combinations of said variables (12) , such as sums and / or products of said variables (12) , are used as independent variables, wherein said starting values are used in the computational analysis, and wherein the computational analysis is used to determine said inspection interval or renovation interval as a function of said variables(12) , after which in the method- the values of said variables (12) are determined for a single subsection of the network (10) ,- an individual inspection interval or renovation interval is determined for the single subsection of the network (10) using said function determined by computational analysis, wherein the function is entered with the values of the variables (12) of the subsection in question,a single subsection of the network (10) is inspected or renovated according to the specified inspection interval or renovation interval.

2. The method according to claim 1, characterised in that the number of the said variables (12) is at least 2, preferably at least 3.

3. The method according to claim 1 or 2, characterised in that as one said variable (12) , a variable (12) describing the con- dition information of the network (10) subsection is used, which is a function of the properties of the components of the network (10) subsection.

4. The method according to any one of claims 1 to 3, charac- ter ieed in that as one said variable (12) , a variable (12) describing the failure frequency of the network (10) subsection is used, which is a function of the number of failures detected by the sensor in the subsection.

5. The method according to any one of claims 1 to 4, charac- terised in that as one said variable (12) , a variable (12) describing the topology of the network (10) subsection is used, which is a function of the objects of use and network type of the subsection.

6. The method according to any one of claims 1 to 5, charac- terised in that a regression analysis is used as the computa- tional analysis.

7. The method according to claim 6, characterised in that the least squares method is used in the regression analysis to determine said function on the basis of said starting values.

8. The method according to any oonnee of claims 1 to 7,characterised in that a neural network is used as the computa- tional analysis.

9. The method according to any one of claims 1 to 8, charac- terised in that the number of starting values of the variables (12) and the inspection intervals corresponding to the starting values to be given for the computational analysis is 1-40%, preferably 5-20%, of all possible combinations of the starting values of the variables (12) .

10. The method according to any one of claims 1 to 9, charac- terised in that the determined individual inspection intervals or renovation intervals of the network (10) subsections are visualised using a geographic information system.

11. The method according to any one of claims 1 to 10, charac- terised in that for at least one of the variables (12) used in the method a value is assigned during the inspection or reno- vation of the network (10) subsection, which value is stored in the memory, and is read from the memory to determine the next inspection interval or renovation interval.

12. The method according to any one of claims 1 to 11, charac- terised in that a flying device, such as a helicopter or a drone, and a recording device, such as a camera or a laser scanner, are used to inspect the network (10) subsection.

13. The method according to any one of claims 1 to 12, charac- terised in that the network (10) is an electricity network.

14. The method according to claim 13, characterised in that as one said variable (12) , a variable (12) describing the condi- tion information of the electricity network subsection is used, which is a function of the characteristics of the electricitynetwork subsection components, such as poles, such as the rot- ting and / or age of the poles.

15. The method according to claim 13 or 14, characterised in that as one said variable (12) , a variable (12) describing the failure frequency of the electricity network (10) subsection is used, which is a function of the number of failures detected by the sensor in the subsection and the power and energy of the electricity transmission in the subsection.

16. The method according to any one of claims 13 to 15, char- acterised in that as one said variable (12) , a variable (12) describing the topology of the electricity network (10) sub- section is used, which is a function of the objects of use and the cable type of the subsection.

17. The method according to any one of claims 13 to 16, char- acterised in that as one said variable (12) , a variable (12) describing the vegetation status in the electricity network (10) subsection is used.

18. The method according to any one of claims 13 to 17, char- acterised in that 1 to 10 disconnector distances are used as the said electricity network (10) subsection.

19. The method according to any one of claims 1 to 12, charac- terised in that the network (10) is one of the following: domestic water network, railway network, road network, district heating network, telecommunication cable network, gas network, oil pipeline network.

20. The method according to any one of claims 1 to 19, charac- terised in that the inspection interval and detected failures of a single subsection of the network (10) are recorded, andthe recorded data are used to assign to the variables (12) a selected number of starting values differing from each other and the inspection intervals or renovation intervals corre- spending to these starting values.

21. A system for the preventive maintenance of the components of the network (10) , the system comprising a computer compris- ing software tools, a processor and memory tools adapted to determine dynamic, individual inspection intervals or renova- tion intervals for the network (10) subsections based on values of selected variables (12) describing the network (10) subsec- tion, characterised in the that the system comprises- a memory structure stored on the computer memory tools, which comprises a selected number of starting values differing from each other given to the selected variables (12) and the inspec- tion intervals or renovation intervals corresponding to these starting values,- a first program code that can be arranged on the computer memory tools for execution by the processor, and which computer can be adapted to determine, by executing the first program code on its processor, the impact of the dependence between the values of the selected variables (12) on the inspection inter- val or renovation interval by means of a computational analy- sis, wherein the dependent variable is the inspection interval or renovation interval and the independent variables are said variables (12) and / or combinations of said variables (12) , such as sums and / or products of said variables (12) , wherein said starting values are used in the computational analysis, and wherein the computational analysis determines said inspection interval or renovation interval as a function of said variables(12) , and- a second program code that can be arranged on the computer memory tools for execution by the processor, and which computercan be adapted to determine, by executing the second program code on its processor, an individual inspection interval or renovation interval for a single subsection of the network (10) using said function determined by computational analysis, to which function the values of the variables (12) in the subsec- tion in question have been entered.

22. The system according to claim 21, characterised in that the system comprises display tools and a geographic information system to display the specified inspection interval or renova- tion interval.

23. The system according to claim 21 or 22, characterised in that the system comprises a sensor for determining the value of at least one variable describing the subsection of said network (10) .

24. The system according to any one of claims 21 to 23, char- acterised in that the system comprises tools for inspecting the network (10) subsection according to the inspection interval determined and read from the memory tools.

Images