Fault management in renewable energy production

Abstract

A computerized method of fault prioritization for a distributed system of renewable energy production assets is described Fault notifications, each comprising an identification of an energy production asset with a fault and one or more fault parameters for that fault, are received. A fault prioritization is determined from the fault parameters and from at least one further parameter for each fault indicative of current projected possible operation of the renewable energy production asset affected by that fault on remediation of the fault. A method of fault remediation for a distributed system of renewable energy production assets comprising such a method of fault prioritization is also described, as is a computer system adapted to perform such a method of fault prioritization.

Description

[0001] FAULT MANAGEMENT IN RENEWABLE ENERGY PRODUCTION

[0002] TECHNICAL FIELD

[0003] The invention relates to fault management in renewable energy production.

[0004] BACKGROUND

[0005] In renewable energy production, it is normal for energy production assets, such as wind turbines, to be highlydistributed geographically, but centrally managed. Forwind parks and individualwind turbines, a service operator (who may be the turbine manufacturer) will typically monitor and maintain wind turbines. Often this will be in accordance with service contracts with various performance thresholds. The service operator will typically have a central management process, which may be for example running on a management server in a central facility, or on a virtual server in the cloud.

[0006] Where this type of management process exists, fault monitoring will typically involve the energy production assets indicating to the management process when a fault exists. The management process will then schedule fault remediation - for example, by remote reconfiguration and management, or by dispatching a repair crew to conduct a physical repair. Typically such management processes operate by using a FIFO (first in, first out) buffer so that faults are addressed in the order that they are received. This has the benefit of predictability, but it fails to take account of variation between faults - some faults may be more significant than others, and some energy assets may be more significant than others, and these factors may vary over time rather than be static. Consequently, such an approach to management of faults may be less than optimal in maximizing the effectiveness of the renewable energy production infrastructure.

[0007] It would be desirable to provide a solution to such problems in fault management in renewable energy production.

[0008] SUMMARY OF THE INVENTION

[0009] In a first aspect, the invention provides a computerized method of fault prioritization for a distributed system of renewable energy production assets, the method comprising: receiving a plurality of fault notifications, where each fault notification comprises an identification of an energy production asset with a fault and one or more fault parameters for that fault; and providing a fault prioritization determined from the fault parameters and from at least one further parameter for each fault indicative of current projected possible operation of the renewable energy production asset affected by that fault on remediation of the fault.

[0010] Using such an approach, a real-time prioritization of faults can be achieved, allowing resources to be directed to where they will provide the greatest overall value, rather than relying simply on the contents of a FIFO buffer, which may contain a number of non-urgent faults otherwise blocking remediation of the urgent faults.

[0011] In embodiments, the current projected possible operation of the renewable energy production asset affected by that fault on remediation of the fault may comprise projected current energy production for the renewable energy production asset in fault-free operation. Using this feature allows assets that would, if fault-free, be producing energy highly effectively or in significant volumes to be prioritized.

[0012] In embodiments, the renewable energy production assets may comprise wind turbines and / or wind turbine farms. If so, one of the at least one further parameters may be wind speed at the renewable energy production asset. Using wind speed in this way allows fault prioritization to be focussed on assets that apart from the fault, are in the best position to meet the requirements placed on them.

[0013] In embodiments, the fault parameters may comprise a time the fault occurred. If so, the fault prioritization may be determined from an age of the fault, using the time that the fault occurred.

[0014] In embodiments, the fault parameters may comprise a severity of the fault.

[0015] In embodiments, the method may further comprise providing a revised fault prioritization at a prioritization refresh rate such that the fault prioritization is substantially always current. In embodiments, the fault prioritization may also be determined from a prioritization of the renewable energy production assets, wherein the prioritization of the renewable energy production assets may be based on one or more asset parameters. Such asset parameters may comprise one or more of the following: asset efficiency; asset energy production; and asset availability. In such cases, the prioritization of the renewable energy production assets may be further based on renewable energy production asset performance against predetermined performance targets. When an approach involving prioritization of the renewable energy production assets is taken, the method may further comprise redetermining the energy production asset prioritization ata predetermined asset prioritization frequency. For cases where there is also a revised fault prioritization provided at a prioritization refresh rate, the asset prioritization frequency may be at least an order of magnitude lowerthan the prioritization refresh rate. Use of such an asset prioritization approach allows for a sophisticated approach to fault prioritization to be adopted - faults are assessed not only on the fault and on current conditions at the asset, but also on whether and to what extent the asset itself is a priority.

[0016] In a second aspect, the invention provides a computer system comprising at least one memory and a processor programmed to perform the method of the first aspect of the invention.

[0017] In a third aspect, the invention provides a method of fault remediation for a distributed system of renewable energy production assets, the method comprising: receiving and prioritizing notifications of a plurality of faults each associated with one of the energy production assets according to the method of the first aspect; and remediating the plurality of faults according to the determined fault prioritization.

[0018] BRIEF DESCRIPTION OF THE DRAWINGS

[0019] One or more embodiments of the invention will now be described, byway of example only, with reference to the accompanying drawings, in which:

[0020] Figure 1 is a schematic diagram of a renewable energy production infrastructure with a central management facility;

[0021] Figure 2 is a flow diagram illustrating a method of fault prioritization according to an embodiment of the invention; Figure 3 illustrates an exemplary computing environment for implementing an embodiment of the invention;

[0022] Figure 4 is a flow diagram illustrating in detail an asset prioritization process used in embodiments of a fault prioritization method according to the invention and

[0023] Figure 5 is a flow diagram illustrating in detail a specific embodiment of a fault prioritization method according to the invention.

[0024] DETAILED DESCRIPTION

[0025] Figure 1 is a schematic diagram illustrating a renewable energy production infrastructure with a central management facility. In embodiments described in detail here, this renewable energy production infrastructure comprises a plurality of wind turbines and wind parks. However, in other embodiments, the energy production infrastructure may comprise other types of renewable energy source (solar, hydroelectric, geothermal) - the infrastructure may comprise only one type of energy production asset, or may comprise a variety of types. The present invention may also be applied to infrastructures in which energy production assets are predominantly renewable, butwhere othertypes of energy production asset may also be present.

[0026] Here, the renewable energy production infrastructure comprises a plurality of wind turbines 1 and a plurality of wind farms 2 (each comprisinga plurality of wind turbines). In principle, a wind farm 2 could be considered as a single asset or as a plurality of assets, with each wind turbine being a single asset - certain types of fault may affect the wind farm 2 as a whole (for example, a power connection failure affecting the whole wind farm 2), whereas other faults will be in individual turbines. In the detailed specific example that follows, individual wind turbines are each treated as a single asset, though they may share characteristics with other wind turbines in the same wind farm. These energy production assets may provide power to one or more power grids 3 (shown as a single power grid for convenience) and here are managed from a common management facility 4. This common management facility 4 is shown as being a single control centre having a management server 5 adapted for management of the infrastructure - this management functionality could of course be disaggregated or virtualised as for most serverbased computing systems. The management server 5 communicates with the wind turbines 1 and wind farms 2 over an appropriate communications infrastructure 6 (this may be a dedicated communications infrastructure, or simply the public Internet) - the management server 5 receives data from the energy production assets relating to their performance to monitor them, and the server may provide control information to the energy production assets when required. A part of the monitoring role of the management server 5 is in receiving and managing fault notifications from the energy production assets. For a wind turbine, for example, when a fault or defect is detected, energy production stops and an alarm signal is sent - where the wind turbine is centrally managed, this may be sent to the management server 5 of the common management facility 4. For a wind turbine, such alarm signals will typically identify the wind turbine affected, the time the fault occurred, and some kind of fault code -they may also provide details of factors relatingto energy production, such as a current wind spend and a generated power (active power) at the time of the fault. These alarms are handled by a remote operations team at the common management facility 5 accessing an alarm handling system provided on the management server 5, which conventionally provides alarms to be addressed in age order through a First In First Out (FIFO) buffer. The operations team address these faults in order - they determine whether the fault or defect can be fixed remotely, in which case they will do so, and if the fault or defect cannot be fixed remotely then notifications are issued to the site or to a local team to address the fault physically on site.

[0027] The present inventors have appreciated that by a process of fault prioritization, faults remediation can be achieved in a way that contributes to much better management of resources. Figure 2 illustrates a method 20 of fault prioritization, implemented on a computing system such as the management server 5, according to an embodiment of the invention. A plurality of fault notifications are received 21 , as in previous arrangements. Each fault notification comprises a notification of an energy production asset, such as a wind turbine, with a fault, along with one or more fault parameters forthat fault. These parameters may for instance include the time that the fault occurred and the severity of the fault.

[0028] These fault notifications are then prioritized 22 - this is on the basis of the fault parameters, but also on the basis of at least one further parameter indicative of current projected possible operation (or operational effectiveness) of the asset on remediation of the fault. Possible operation, or operational effectiveness, reflects how effectively an asset would perform its core function - power production - if fault-free and fully operational. Renewable assets will typically vary in current effectiveness over time. For a wind turbine, such a factor may be the wind speed at the turbine location - if the wind speed is too high or too low for the turbine to function effectively, the priority may be lower than if the wind speed was in the preferred operating range for the turbine, as fault remediation could then lead to greater saved power production. Similar considerations would apply to other types of renewable energy production assets - for example, there may not be an immediate need to address a fault in a solar cell array early in the evening, as it would not produce power until the following morning.

[0029] Faults will then be assigned for remediation 23 from the list as resources become available. The prioritization 22 will change as new faults appear, but it may also change as various parameters used in the prioritization may change overtime. For example, prioritization may be dependent on the age of the fault- while this is a factor determined from a fault parameter (the time of the fault), it will increase over time, typically leading to a higher prioritization for the fault.

[0030] As will be described in detail below with reference to an implementation of such a process in wind farm management, other factors may also be involved in this prioritization process. One significant factor may be the prioritization of the energy production assets themselves.

[0031] Figures 3 to 5 illustrate in more detail an implementation of this approach in the management of wind turbines and wind farms. Figure 3 illustrates relevant management server functionality, Figure 4 illustrates an asset prioritization process used in certain embodiments of the invention, and Figure 5 illustrates a fault prioritization process accordingto an embodiment of the invention using the asset prioritization process of Figure 4.

[0032] Figure 3 shows a management server adapted to implement fault prioritization according to an embodiment of the invention. The following functional processes are implemented in a computing environment 30 of the management server 5, implemented by a suitably programmed processor 32 with access to a memory 33 running under a suitable operating system 31 : these processes comprise an asset prioritization process 34, a fault management process 35, and a management interface process 36 to allow operator control. The asset prioritization process accesses an asset database 37, and the management server 5 has a network interface 38 through which the management server 5 connects to the individual assets 1 and receives fault notifications - connection through the network interface 38 may also be used to obtain further information from the individual assets orfrom their location, and it may be used to provide control signals to individual assets or their location for remote remediation of faults or otherwise. In embodiments, such remediation of faults may be achieved through the management server s, or may be implemented through other remote access systems. While this functionality is here shown as provided in one management server 5, this may not be the case in practice - the functionality may be split between multiple servers, or partly virtualised. Operator access to the management interface process 36 may be through a client / server interface 39.

[0033] As previously noted, it has been realised that appropriate prioritization of faults can be used to improve the overall effectiveness of management of the renewable energy production asset infrastructure - for the purpose of the discussion of this embodiment, wind turbine will be used in place of energy production asset as this embodiment relates to wind turbine and wind farm management. As noted with respect to Figure 2, the time (and hence age) and severity of the fault are significant factors here, as are conditions determining the current effectiveness of the wind turbine which determines the energy production lost by leaving the wind turbine out of operation. However, a further factor that can be considered is a prioritization of wind turbines themselves. Such a process will be described with reference to Figure 4.

[0034] Figure 4 illustrates an asset prioritization process 34 used in certain embodiments of the invention - in this case, each asset is a wind turbine in the total managed assemblage of wind turbines. This begins with extraction 41 of operational data from all managed wind turbines by the asset prioritization process. This is followed by a feature engineering step 42 to convert the operational data into inputs for a statistical ranking process 43. This feature engineering step 42 comprises a set of data transformation rules that enable the raw data extracted in the extraction step to be converted into meaningful input variables for the subsequent statistical ranking process 43.

[0035] In these initial steps, operational data is obtained that is particularly relevant to determining whether one wind turbine should be addressed before another. Certain factors may relate to the energy production capacity of the wind turbine (for example, a production profile, which may be measured as possible production divided by turbine availability in time), and others may relate to a turbine history (such as an amount of lost capacity over time). Other factors may relate to commitments made in relation to a wind turbine ora wind farm -these may involve any difference between a guaranteed availability and a historical actual availability for an asset, or any other contractual guarantee in relation to wind turbine performance.

[0036] Using these inputs, the statistical ranking process 43 generates a ranking of the wind turbines. Ranking here is provided bya multi-criteria decision aid such as a multi-feature ranking algorithm - in this embodiment, this uses a PROMETHEE II approach to provide a complete ranking of alternatives. The PROMETHEE (Preference Ranking Organization Method for Enrichment Evaluations) approach was originally developed by Brans (see Brans etal, “Howto select and how to rank projects: The PROMETHEE method”, European Journal of Operational Research, 24(2), pp:228-238, February 1986) and a full review of PROMETHEE methodologies and applications can be found in Behzadian et al, “PROMETHEE: A comprehensive literature review on methodologies and applications”, European Journal of Operational Research, 200(1 ), pp:198- 215, January 2010. By using outranking relations, the PROMETHEE II approach enables the creation of a comprehensive ran king amongst potential alternatives. This method performs better than known alternatives and can be used to integrate many criteria. PROMETHEE II can be implemented as an O(n2) algorithm and it offers robustness in terms of ranking consistency.

[0037] In embodiments of this statistical ranking process 43, different features are assigned different weights to enable the asset priority to be determined. These weights may be positive or negative - beneficial or non-beneficiaL An increase in a beneficial variable will boost the ranking- for example, a history of production loss will make it a high priority for an asset to stay in service with as little further downtime as possible. Conversely, non-beneficial indices contribute inversely to the ranking process, such that an increase in such variables leads to a decrease in rank. This may be the case, for example, where an asset has substantially outperformed what is required of it.

[0038] The result of the process is an asset priority ranking list 44 of all wind turbines for use in fault prioritization. As will be described below, fault prioritization may be a “real-time” process, but that need not be the case for establishing this assent priority ranking list. The priority ranking list 44 will need to change over time - the operational data will change, and the assemblage of wind turbines itself may change. However, the process of extracting operational data, feature selection and statistical ranking for a large number of assets (potentially tens of thousands for a managed wind turbine ecosystem) may not be suited to real-time calculation, and the sensitivity of the fault prioritization to changes in the asset priority ranking list may not justify such rapid cycling. An appropriate frequency for refreshing the asset prioritization process may be, for example, weekly, which would be a typical frequency for refreshing a centralised list of assets. It however may be changed at greater or lesser frequency, either when the composition of the list of serviced assets changes or otherwise, but it does not need to be “real-time”. Figure 5 shows operation of the fault management process 35, showing how it interacts with the asset prioritization process 34 and the management interface process 36. The fault management process 35 receives fault notifications 51 - these will include an identification 511 of the asset (or assets) affected by the fault, and they will generally include the fault parameters 512 (though it is possible that some fault parameters may be obtained by interrogation by the fault management process 35 rather than simply provided with the fault notification 51 ). For a received fault notification 51 , the fault management process will obtain 52 necessary inputs for fault prioritization, which here includes the current asset priority ranking list 44 from the asset prioritization process and the parameter indicative of current projected effectiveness of the asset - in the case of a wind turbine, this would typically be current wind speed. These inputs are then used to generate 53 a fault prioritization list 54 using a ranking score. This fault prioritization list 54 is accessed 55 by the management interface process 36, which operators use to address faults from the fault prioritization list 54 - when an operator addresses a fault, this is removed from the fault prioritization list 54. An operator either remediates the fault by a remote management operation 56, or by scheduling site maintenance 57 (if necessary for the fault type or if a remote management operation does not succeed).

[0039] The process of generating a fault prioritization list will now be described in more detail An exemplary fault prioritization function for use for this purpose is indicated below: where x is the rank of a wind turbine (among n alternatives) in the current asset priority ranking, a is the age of the alarm / fault notification, w is the wind speed and f(s) is a fault severity function. These functions are weighted by control weights mi, m2, m3and m4, which sum to 1. Each function will now be described in more detail.

[0040] Priority Ranking Function

[0041] This function weights the fault prioritization function to favour wind turbines that are ranked more highly in the asset prioritization process. This function may be, for example,

[0042] 100(x — n)

[0043] 1 — n More complex functions are possible here - for example, instead of an asset ranking, an asset prioritization score could be used so that more weight is given to assets where factors very strongly suggest prioritization of the asset.

[0044] Alarm Age Scoring Function

[0045] The time an alarm is raised is a parameter of the fault. For fault prioritization, the parameter to be used directly will generally be the age a of the fault, which can of cause be derived from the alarm time. A service level agreement (SLA) between an asset owner and an asset manager will typically have a clause to which alarm age is releva nt -for exam pie, a permitted length of downtime p over a particular evaluation period. If this limit p (expressed in minutes) and the alarm age a is given in seconds, an appropriate function may be the following:

[0046] ( 100 * a r - , a < 60p f a) = 60 * p

[0047] ( 100, a > 60p

[0048] The intention here is that if the alarm age reaches the SLA compliance limit, it should be prioritized very highly- there is a monotonic increase in f(a) up until this point, at which point the function reaches 100 and stays there. At this point the wind turbine downtime has become problematic - it affects the expectations (which may be contractually enforced by SLA) of the wind turbine owner.

[0049] Wind Speed Function

[0050] If the asset is not currently capable of being productive, faults may be given a lower priority. For a wind turbine, the simplestway of implementing this is by deprioritizing wind turbines where the current local wind speed is belowthe cut-in speed for power provision from the turbine. This may be expressed as follows, where wcis the cut-in wind speed:

[0051] 100, iv > ivc100, w is unavailable 0, iv < wc More complex arrangements are possible. For example, wind turbines where the wind speed is too high for power generation operation could also be deprioritized, or there could be a more complexfunction than a binaryon-off to cover different power production capabilities atdifferent windspeeds. Ratherthan just a current wind speed, the wind speed w could be an averaged wind speed over a period of time, or a predicted wind speed for a period of future operation in which production would be lost if the fault is not remediated.

[0052] Alarm Severity Function

[0053] Different faults have different severity - typically, most faults are of low severity whereas a small class of faults have high severity and need to be addressed rapidly. Consequently, the alarm severity function f(s) may be a log transform function, as this is effective to deal with a sparse and skewed alarm priority distribution. An exemplary alarm severity function is the following: 0, s = 0 100 i - logio S, O < s < ns og10nswhere ns s the number of turbines under active remote monitoring. This function gives appropriate weight to the more severe alarm priorities.

[0054] The result of this process is that each asset will have a fault prioritization score y, and the fault prioritization list 54 is created y ordering the faults according to their fault prioritization score. This prioritization list 54 is presented to the operators through the management interface process 36. Faults are then removed from the prioritization list by the operators when they are processed by remote management operations or by scheduled site maintenance.

[0055] The fault prioritization list 54 is advantageously presented as an accurate reflection of the position in real-time - it therefore needs to be regenerated frequently. This is straightforward - only a small proportion of the whole population assets will have faults requiring attention at any one time, and the fault prioritization score function is very straightforward to calculate. There is no great burden if, for example, the fault prioritization list 54 is re-determined each time a fault notification is received or a fault removed from the list, and additionally refreshed relatively frequently (for example, every second or every few seconds). Using this approach, priority can be given where it is particularly needed - for example, alarms from turbines with a critical health status can be prioritized, and their overall downtime reduced and performance improved. This will improve the historical performance of these turbines, and will lead to them being lowered in the asset priority ranking list 44, and reduces any risk that obligations of an asset operator to an asset owner may not be met.

Claims

CLAIMS1 . A computerized method of fault prioritization fora distributed system of renewable energy production assets, the method comprising: receiving a plurality of fault notifications, where each fault notification comprises an identification of an energy production asset with a fault and one or more fault parameters forthat fault; and providing a fault prioritization determined from the fault parameters and from at least one further parameter for each fault indicative of current projected possible operation of the renewable energy production asset affected by that fault on remediation of the fault.

2. The method of claim 1 , wherein current projected possible operation of the renewable energy production asset affected by that fault on remediation of the fault comprises projected current energy production for the renewable energy production asset in fault-free operation.

3. The method of claim 1 or claim 2, wherein the renewable energy production assets comprise wind turbines and / or wind turbine farms.

4. The method of claim 3, wherein one of the at least one further parameters is wind speed at the renewable energy production asset.

5. The method of any preceding claim, wherein the fault parameters comprise a time the fault occurred.

6. The method of claim 5, wherein the fault prioritization is determined from an age of the fault, using the time that the fault occurred.

7. The method of any preceding claim, wherein the fault parameters comprise a severity of the fault.

8. The method of any preceding claim, further comprising providing a revised fault prioritization ata prioritization refresh rate such that the fault prioritization is substantially always current.

9. The method of any preceding claim, wherein the fault prioritization is also determined from a prioritization of the renewable energy production assets, wherein the prioritization of the renewable energy production assets is based on one or more asset parameters.

10. The method of claim 9, wherein the asset parameters comprise one or more of the following: asset efficiency; asset energy production; and asset availability.

11. The method of claim 10, wherein the prioritization of the renewable energy production assets is further based on renewable energy production asset performance against predetermined performance targets.

12. The method of any of claims 9 to 11 , further comprising redetermining the energy production asset prioritization at a predetermined asset prioritization frequency.

13. The method of claim 12 where dependent on claim 8, wherein the asset prioritization frequency is at least an order of magnitude lower than the prioritization refresh rate.

14. A computer system comprising at least one memory and a processor programmed to perform the method of any of claims 1 to 13.

15. A method of fault remediation for a distributed system of renewable energy production assets, the method comprising:receiving and prioritizing notifications of a plurality of faults each associated with one of the energy production assets according to the method of any of claims 1 to 13; and remediating the plurality of faults according to the determined fault prioritization.

Images