Method of operating an electric power system, predictive system, supervisory control and data acquisition system, and electric power system

Abstract

Techniques useful for operating an electric power system using a Supervisory Control and Data Acquisition, SCADA, system (40) are provided. The SCADA system (40) comprises a first control and protection subsystem (62) and a second control and protection subsystem (65) that provide redundancy. The SCADA system (40) comprises a predictive system (70) operative to communicatively interface with the first and second control and protection subsystems (62, 65) to perform a predictive asset health assessment.

Description

[0001] METHOD OF OPERATING AN ELECTRIC POWER SYSTEM, PREDICTIVE SYSTEM, SUPERVISORY CONTROL AND DATA ACQUISITION SYSTEM, AND ELECTRIC POWER SYSTEM

[0002] TECHNICAL FIELD

[0003] Embodiments of the invention relate to systems and methods useful in association with electric power system. Embodiments of the invention relate in particular to systems and methods that allows a predictive asset health assessment to be performed for devices (e.g., assets, equipment and / or other devices) of an electric power system and / or devices of a Supervisory Control and Data Acquisition (SCADA) system associated with the electric power system. Embodiments of the invention relate in particular to such systems and methods that operative in association with a high voltage direct current (HVDC) system.

[0004] BACKGROUND

[0005] Systems that comprise primary system equipment and a Supervisory Control and Data Acquisition (SCADA) system for monitoring and coordinating devices for controlling and protecting the equipment are widely used. Such SCADA systems may be implemented in an automation control system, such as an automation control system for use in electric power system automation (e.g. substation automation systems) or other infrastructure automation systems. SCADA systems for electric power systems are complex and may comprise a significant number of devices and function instances executed on these devices.

[0006] Failure of equipment of an electric power system, such as a power grid equipment, and / or of a device of a control and protection system can have potentially catastrophic consequences. Thus, it is desirable to perform an asset health management operative to monitor assets of the electric power system and / or devices of the SCADA system to identify assts or other devices requiring maintenance prior to their failure.

[0007] Predictive techniques of performing an asset health assessment offer the benefit that maintenance work and / or control operations that allow maintenance to be performed can be implemented in a proactive manner, e.g., when a remaining useful life of an asset reaches or falls below a safety margin period. Predictive techniques of monitoring asset health are disclosed in, e.g., EP 3 923 214 Al, EP 3 923 213 Al, EP 3 923 101 Al.

[0008] Conventionally, maintenance work on primary system equipment of an electric power system or SCADA system devices typically requires that part of the electric power system be set to a maintenance mode for performing the maintenance. For systems such as a high voltage direct current (HVDC) system (e.g., a HVDC transmission system), it is desirable that downtimes of the primary system equipment, in particular the converter / inverter assets, be reduced or, preferably, eliminated.

[0009] In view of the above, there is a continued need in the art for improved techniques of performing predictive asset health assessment. There is in particular a need for techniques of performing predictive asset health assessment that are operative use in association with an electric power system having subsystem parts in the primary system that provide a backup in case of maintenance work.

[0010] SUMMARY

[0011] It is an object of the invention to provide methods and systems that provide enhanced techniques of performing predictive asset health assessment. It is in particular an object of the invention to provide methods and systems operative in association with a control and protection system of a Supervisory Control and Data Acquisition (SCADA) system in a manner that can reduce downtimes during maintenance.

[0012] According to exemplary embodiments, methods and systems as recited in the claims are provided.

[0013] According to an aspect of the invention, there is provided a method of operating an electric power system using a Supervisory Data Acquisition and Control, SCADA, system. The method comprises monitoring, by the SCADA system, the electric power system. The SCADA system comprises a first control and protection subsystem (e.g., an "A" or "active" subsystem) and a second control and protection subsystem (e.g., a "B" or "standby" subsystem). The first control and protection subsystem and the second control and protection subsystem implement control and protection function redundancy. The method comprises performing, by the first control and protection subsystem, control and protection functions for a first electric power subsystem of the electric power system. The method comprises performing, by the second control and protection subsystem, control and protection functions for a second electric power subsystem of the electric power system. The second electric power subsystem is a backup for the first electric power subsystem. The method comprises communicatively interfacing, by a predictive system, with the first control and protection system and the second control and protection system to obtain electric power system information for performing an asset health assessment. The method comprises using, by the predictive system, the obtained electric power system information to perform a predictive asset health assessment for at least one asset of the electric power system. The method comprises generating, by the predictive system, output based on the predictive asset health assessment.

[0014] Various effects and advantages are associated with the method. By providing first and second control and protection subsystems that provide a redundancy of control and protection functions by performing control and protection functions for different electric power subsystems of the electric power system (that again provide a backup at the primary system equipment level), a change-over can be performed in case maintenance is required. For illustration, if one or several devices of the first control and protection subsystem are set to maintenance mode, a change-over logic can cause the second control and protection subsystem and the second electric power subsystem with which it is associated to operate, both with regard to electric power conversion and / or transmission in the primary system (second electric power subsystem) and with regard to control and protection (second control and protection subsystem), in a manner that allows seamless continued operation. The operations normally performed by the first electric power subsystem and the first control and protection subsystem are implemented in a redundant manner in the second electric power subsystem and the second control and protection subsystem, thereby reducing downtimes. The predictive system is operative to perform the predictive asset health assessment that allows maintenance to be implemented in a pro-active manner, thereby synergistically contributing to the reduction or elimination of electric power system downtimes.

[0015] The first and second subsystems may be comprised by or otherwise associated with a same station (e.g., a station at a local end or a station at a remote end of a line). Each of the two stations at opposite ends of a line (local end and remote end) may have its respective first and second control and protection subsystems and first and second electric power subsystem.

[0016] The predictive system is communicatively coupled with both the first control and protection subsystem and the second control and protection subsystem. Thereby, the predictive system is enabled to use both the part of the electric power system information obtained from the first control and protection subsystem and the part of the electric power system information obtained from the second control and protection subsystem to perform the predictive asset health assessment. This makes it possible to perform the predictive asset health assessment by combining the part of the electric power system information obtained from the first control and protection subsystem and the part of the electric power system information obtained from the second control and protection subsystem.

[0017] Performing the predictive asset health assessment may comprise detecting a drift from the part of the electric power system information obtained from the first control and protection subsystem and the part of the electric power system information obtained from the second control and protection subsystem. The drift may be determined based on a change in asset state of an asset of the first control and protection system or of the first electric power subsystem, as determined from a drift in data included in the part of the electric power system information obtained from the first control and protection subsystem as compared to data included in the part of the electric power system information obtained from the second control and protection subsystem. Thereby, parameters that do not only depend on a current asset state but take into account a drift in change in asset condition that allows the predictive asset health assessment to be performed may be used to determine and generate the output.

[0018] Performing the predictive asset health assessment may comprise determining remaining useful life (RUL), determining an end of life, etc.

[0019] Thereby, parameters that do not only depend on a current asset state but take into account a drift in change in asset condition that allows the predictive asset health assessment to be performed may be used to determine and generate the output.

[0020] The output may be indicative of a future degradation of at least one asset as predicted based on the detected drift, the RUL, or the end of life.

[0021] Thereby, the predicted future degradation of the asset to be taken into account for generating and providing the output.

[0022] The obtained electric power system information may comprise first configuration data obtained from the first control and protection subsystem and second configuration data obtained from the second control and protection subsystem. The predictive system may use the first configuration data and the second configuration data to perform the predictive asset health assessment and generate the output.

[0023] Thereby, consistency of the configuration data used by the predictive system to perform the predictive asset health assessment with the configuration data of the first and second control and protection systems is attained. Use of the first and second configuration data as obtained from the first and second control and protection systems by the predictive system also make the predictive system operative to discriminate asset degradation that may warrant maintenance from a change in hardware that has been performed and that can result in an abrupt change in the received power system information.

[0024] The first configuration data may relate to a configuration of the first electric power subsystem. The second configuration data may relate to a configuration of the second electric power subsystem.

[0025] Thereby, the predictive system is operative to take into account the configuration of the primary system (i.e., the components that carry the high currents or have the high voltages in the primary power system) for performing the predictive asset health assessment.

[0026] The predictive asset health assessment may comprise discriminating, by the predictive system, an asset degradation from a hardware change. The predictive system may identify the hardware change based on the first configuration data and the second configuration data.

[0027] Thereby, the predictive system is operative to discriminate asset degradation that may warrant maintenance from a change in hardware that has been performed and that can result in an abrupt change in the received power system information. The obtained electric power system information may comprise operation data of the electric power system. The predictive system may use the operation data to perform the predictive asset heath assessment and generate the output.

[0028] Thereby, the predictive system is operative to use the operation data to perform the predictive asset health assessment, which is particularly useful for detecting, e.g., drifts.

[0029] The operation data may comprise first sensor output of sensors of the first control and protection subsystem and second sensor output of sensors of the second control and protection system. The predictive system may be operative to perform the predictive asset health assessment based on the first sensor output and the second sensor output, in particular based on a time-dependent evolution of the first sensor output and the second sensor output.

[0030] Thereby, the predictive system is operative to use the first and second sensor outputs to detect drifts or otherwise perform a predictive asset health assessment.

[0031] The predictive system may be operative to perform the predictive asset health assessment based on an inconsistency between first sensor output and the second sensor output, in particular based on a time-dependent evolution of the inconsistency.

[0032] Thereby, the predictive system is operative to use the sensor outputs of the various control and protection subsystems as a reference for each other, facilitating the detection of drift in asset state that is indicative of a degradation or that otherwise influences the RUL or end of life.

[0033] The operation data may comprise first control or protection device output of a control or protection device of the first control and protection subsystem and second control or protection device output of a control or protection device of the second control and protection system. The predictive system may be operative to perform the predictive asset health assessment based on the first control or protection device output and the second control or protection device output, in particular based on a time-dependent evolution of the first control or protection device output and the second control or protection device output.

[0034] Thereby, the predictive system is operative to use the first and second control or protection device outputs to detect drifts or otherwise perform a predictive asset health assessment.

[0035] The predictive system may be operative to perform the predictive asset health assessment based on an inconsistency between first control or protection device output and the second control or protection device output, in particular based on a time-dependent evolution of the inconsistency.

[0036] Thereby, the predictive system is operative to use the sensor outputs of the various control and protection subsystems as a reference for each other, facilitating the detection of drift in asset state that is indicative of a degradation or that otherwise influences the RUL or end of life.

[0037] The operation data may comprise one or several of current, voltage, phasor measurements, temperature measurements, cooling fluid temperatures, fluid flow rates, cooling fluid flow rate. The operation data may comprise at least a cooling fluid inlet temperature and a cooling fluid outlet temperature of a cooling system of a converter / inverter asset of the electric power system.

[0038] Thereby, trends in quantities related to complex primary system equipment may be detected.

[0039] The operation data may comprise online operation data. The predictive system may be operative to process the online operation data as it becomes available from the first and second control and protection subsystems.

[0040] Thereby, online data processing may be performed to perform the predictive asset health assessment. This is particularly efficient in association with techniques capable of processing timeseries data.

[0041] The predictive system may be operative to remove (e.g., by filtering) part of the data obtained from the first and second control and protection subsystems prior to the online data processing. A data filter may be provided between the predictive system and each of the first and second control and protection subsystems to remove data not required or not useful to perform the predictive asset health assessment.

[0042] The operation data may comprise offline data (also referred to as batch data). The predictive system may be operative to process the offline operation data as retrieved from a storage system.

[0043] Thereby, batch data processing may be performed to perform the predictive asset health assessment. This allows aggregation of information acquired over, e.g., longer time periods and / or use of data acquired from other SCADA systems in the past.

[0044] The predictive system may be operative to store a subset of the operation data in a storage system of the predictive system or accessible to the predictive system. The predictive asset health assessment may be performed based on the stored subset of the operation data.

[0045] Thereby, the predictive system is operative to aggregate data for use in performing the predictive asset health assessment and / or to perform the predictive asset health assessment for an asset at a later time (depending on, e.g., processing bandwidth availability of the SCADA system and / or criticality of the electric power system). Thereby, the predictive asset health assessment may be performed at times at which it does not consume SCADA system processing resources that might be required otherwise in view of system criticality.

[0046] The method further may comprise the following operations performed by the predictive system: selecting the subset of the operation data for storing in the storage system based on a relevance for performing predictive asset health assessment and storing the selected subset in the storage system.

[0047] Thereby, the predictive system is operative to aggregate data for use in performing the predictive asset health assessment and / or to perform the predictive asset health assessment for an asset at a later time. The predictive system may be operative to perform the asset health assessment for an asset comprising by a plurality of components. Performing the predictive asset health assessment for the asset may comprise establishing a time-dependent degradation of several components of the plurality of components and establishing a time-dependent degradation of the asset based on the timedependent degradation of the several components.

[0048] Thereby, the techniques disclosed herein may be applied to complex assets having several components, the degradation of which can be determined on component level. The asset having the plurality of components may be provided, with identical operation, in each of the first and second electric power subsystems.

[0049] The predictive system may be operative to perform the asset health assessment for both a first asset of the first electric power subsystem, the first asset comprising a first plurality of components, and a second asset of the second electric power subsystem, the second asset comprising a second plurality of components having a construction and operation identical to the first plurality of components. The predictive asset health assessment for the first asset may comprise establishing a time-dependent degradation of several first components of the first plurality of components as compared to a time-dependent degradation of several second components of the second plurality of components and establishing the time-dependent degradation of the first asset based on the timedependent degradation of the first several components. Alternatively or additionally, the predictive asset health assessment for the second asset may comprise establishing a time-dependent degradation of several second components of the second plurality of components as compared to a time-dependent degradation of several first components of the first plurality of components and establishing the time-dependent degradation of the second asset based on the time-dependent degradation of the second several components.

[0050] Thereby, the techniques disclosed herein may be applied to complex assets having several components, the degradation of which can be determined on component level.

[0051] The predictive system may be operative to determine an end of life of the asset having the plurality of components based on the earliest end of life of any of the several components.

[0052] The operation data may comprise measurements from measurement instrumentation associated with the several components of the asset.

[0053] Thereby, the techniques disclosed herein may be applied to complex assets having several components with associated measurement instrumentation.

[0054] Performing the predictive asset health assessment and / or generating the output may comprise using a data-driven technique.

[0055] Thereby, the predictive asset health assessment and / or the generation of the output may be performed in a manner that does not rely exclusively on expert experience in setting up rules for determining asset degradation. Thereby, asset health assessment can be performed in a more objective manner.

[0056] Performing the predictive asset health assessment and / or generating the output may comprise using an artificial intelligence (Al) model. The Al model may have an input operative to receive Al model input that is based on at least part of the obtained data and an Al model output operative to provide a predictive asset health assessment or to otherwise provide Al model output from which the predictive system can determine the output indicative of the asset health assessment.

[0057] Thereby, the predictive asset health assessment and / or the generation of the output may be performed in a data-driven manner that does not rely exclusively on expert experience in setting up rules for determining asset degradation.

[0058] The Al model may be operative to process timeseries data.

[0059] Thereby, the predictive asset health assessment can detect trends, such as drifts in asset state, in an efficient manner.

[0060] The Al model may comprise long short term memory (LSTM) cells and / or gated recurrent units (GRUs) to process timeseries data.

[0061] Thereby, the predictive asset health assessment can detect trends, such as drifts in asset state, in an efficient manner.

[0062] The Al model may comprise at least one attention mechanism.

[0063] Thereby, the predictive system is operative to apply a processing technique that allows a local or global context of datapoints (such as an evolution of measurements indicative of a trend in asset state) to be considered in performing the asset health assessment.

[0064] The Al model may comprise at least one self-attention mechanism.

[0065] Thereby, the predictive system is operative to apply a processing technique that allows a local or global context of datapoints (such as an evolution of measurements indicative of a trend in asset state) to be considered in performing the asset health assessment.

[0066] The Al model may comprise a stack of self-attention mechanisms.

[0067] Thereby, the predictive system is operative to apply a processing technique that allows a local or global context of datapoints (such as an evolution of measurements indicative of a trend in asset state) to be considered in performing the asset health assessment.

[0068] The Al model may comprise an Al-transformer model.

[0069] Thereby, the predictive system is operative to apply a processing technique that allows a local or global context of datapoints (such as an evolution of measurements indicative of a trend in asset state) to be considered in performing the asset health assessment.

[0070] The Al model may comprise a recurrent neural network (RNN) and / or a convolutional neural network (CNN). Thereby, techniques suitable to detect the similarity or dissimilarity of the obtained electric power system information as compared to previous degradation events (encoded in the parameters of the

[0071] RNN or CNN) may be detected to perform the predictive asset health assessment.

[0072] Generating the output may comprise controlling, by the predictive system or by at least one of the first and second protection and control systems, at least one interface to provide the output.

[0073] Thereby, the result of the predictive asset health assessment is used to automatically perform a control operation.

[0074] The output may comprise output indicative of one, several, or all of a degradation, RUL, end of life of at least one asset.

[0075] Thereby, the predicted change in asset condition can be determined and output.

[0076] The at least one interface may comprise a human machine interface (HMI).

[0077] Thereby, the predicted change in asset condition can be determined and provided to a user (such as an electric power system operator and / or maintenance engineer).

[0078] The predictive system may be operative to process degradation, RUL, and / or end of life of at least one asset to determine timing information for at least one scheduled maintenance operation. The output may comprise the timing information for the at least one scheduled maintenance operation. Providing the output may comprise providing the timing information.

[0079] Thereby, the predictive system may be operative to not only identify a degradation in asset state but to also determine when maintenance is required.

[0080] The HMI may comprise a SCADA system HMI. The SCADA system, e.g., the predictive system of the SCADA system, may be operative to control the SCADA system HMI to provide a dedicated maintenance display portion including the output (e.g., the degradation, RUL, end of life of at least one asset and / or the timing information for a scheduled maintenance of at least one asset).

[0081] The output may comprise control commands that cause a transition to a state for performing maintenance.

[0082] Thereby, the risk of unintentional downtimes is reduced. Maintenance can be performed in a manner which is safer for maintenance engineers.

[0083] The control commands may comprise control commands that cause control or protection devices of one of first and second control and protection system to transition to a maintenance mode.

[0084] Thereby, the risk of unintentional downtimes is reduced. Maintenance can be performed in a manner which is safer for maintenance engineers.

[0085] The control commands may comprise control commands that cause the first electric power subsystem to transition to a state in which maintenance can be performed on the first control and protection subsystem and that cause the second electric power subsystem to transition to an energycarrying state. Thereby, the risk of unintentional downtimes is reduced. Maintenance can be performed in a manner which is safer for maintenance engineers.

[0086] The SCADA system may comprise the predictive system.

[0087] Thereby, the predictive system can be an integral part of the SCADA system. This is desirable to afford an operator of the electric power system with full control over the data used by the predictive system and to mitigate safety concerns that the operator might have if the predictive system were to operate remotely from the operator premises.

[0088] The SCADA system is or may comprise a high voltage direct current (HVDC) system SCADA system.

[0089] Thereby, the risk of downtime of the HVDC system is reduced by virtue of the effects and advantages discussed in association with the methods according to various embodiments. In particular, the synergies of providing redundancy at the control and protection level and the primary system level while adding the predictive asset health assessment reduce the risk of the HVDC system being unavailable for, e.g., DC power transmission.

[0090] The SCADA system may comprise control and protection devices for HVDC power system assets. The HVDC power system assets may comprise one, several, or all of: an inverter / converter asset; a cooling system for an inverter / converter asset; a transformer comprising three single-phase transformers or a single 3-pahse transformer operated under control of the control and protection devices.

[0091] Thereby, the synergies of providing redundancy at the control and protection level and the primary system level while adding the predictive asset health assessment reduce the risk of the HVDC system being unavailable for, e.g., DC power transmission.

[0092] The SCADA system may be operative to monitor primary system devices.

[0093] Thereby, the commissioning of the SCADA system is facilitated when the SCADA system is operative to monitor primary system devices (such as assets, equipment, and / or other devices of a primary system of the electric power system).

[0094] The SCADA system may be operative to monitor secondary devices that perform control and protection functions. Monitoring the secondary devices may comprise monitoring hardware usage of the secondary devices that perform control and protection functions.

[0095] Thereby, the commissioning of the SCADA system is facilitated when the SCADA system also monitors the secondary devices.

[0096] The SCADA system may comprise a change-over logic (which may be provided in a redundant manner in the first and second control and protection subsystems or which may be provided separately therefrom) to cause a transition between fully operative and backup modes. The change-over logic may be operative to cause a transition to maintenance mode based on the output of the predictive system. Thereby, the synergies of providing redundancy at the control and protection level and the primary system level while adding the predictive asset health assessment reduce the risk of the HVDC system being unavailable for, e.g., DC power transmission.

[0097] The electric power system may be or may comprise an HVDC system.

[0098] Thereby, the synergies of providing redundancy at the control and protection level and the primary system level while adding the predictive asset health assessment reduce the risk of the HVDC system being unavailable for, e.g., DC power transmission.

[0099] According to another aspect, there is provided predictive system for use with a first control and protection subsystem and a second control and protection subsystem of a Supervisory Control and Data Acquisition, SCADA, system, wherein the first control and protection subsystem and the second control and protection subsystem implement control and protection function redundancy, the first control and protection subsystem being operative to perform control and protection functions for a first electric power subsystem of an electric power system, the second control and protection subsystem being operative to perform control and protection functions for a second electric power subsystem of the electric power system, wherein the second electric power subsystem is a backup for the first electric power subsystem. The predictive system comprises at least one interface operative to communicatively interface the predictive system with the first control and protection system and the second control and protection system to obtain electric power system information for performing an asset health assessment. The predictive system comprises at least one processing circuit operative to use the obtained electric power system information to perform a predictive asset health assessment for at least one asset of the electric power system, and generate output based on the predictive asset health assessment.

[0100] Various effects and advantages are associated with the predictive system. By providing a predictive system operative to communicatively interface with first and second control and protection subsystems that provide a redundancy of control and protection functions by performing control and protection functions for different electric power subsystems of the electric power system (that again provide a backup at the primary system equipment level), a change-over can be performed in case maintenance is required. The predictive system is operative to perform the predictive asset health assessment that allows maintenance to be implemented in a pro-active manner, thereby synergistically contributing to the reduction or elimination of electric power system downtimes.

[0101] According to another aspect, there is provided a Supervisory Control and Data Acquisition (SCADA) system for an electric power system. The SCADA system comprises a first control and protection subsystem and a second control and protection subsystem, wherein the first control and protection subsystem and the second control and protection subsystem implement control and protection function redundancy, the first control and protection subsystem being operative to perform control and protection functions for a first electric power subsystem of an electric power system, the second control and protection subsystem being operative to perform control and protection functions for a second electric power subsystem of the electric power system, wherein the second electric power subsystem is a backup for the first electric power subsystem, and the predictive system of any aspect or embodiment disclosed herein.

[0102] Various effects and advantages are associated with the SCADA system. By providing a predictive system operative to communicatively interface with first and second control and protection subsystems that provide a redundancy of control and protection functions by performing control and protection functions for different electric power subsystems of the electric power system (that again provide a backup at the primary system equipment level), a change-over can be performed in case maintenance is required. The predictive system is operative to perform the predictive asset health assessment that allows maintenance to be implemented in a pro-active manner, thereby synergistically contributing to the reduction or elimination of electric power system downtimes. Integration of the predictive system into the SCADA system affords an operator of the electric power system with full control over the data used by the predictive system and to mitigate safety concerns that the operator might have if the predictive system were to operate remotely from the operator premises.

[0103] The first and second subsystems may be comprised by or otherwise associated with a same station (e.g., a station at a local end or a station at a remote end of a line). Each of the two stations at opposite ends of a line (local end and remote end) may have its respective first and second control and protection subsystems and first and second electric power subsystems.

[0104] According to another aspect, there is provided an electric power system, comprising primary power system equipment and the SCADA system of any aspect or embodiment disclosed herein, wherein the SCADA system is operative to perform control and protection functions for the primary power system equipment.

[0105] Thereby, the effects of the SCADA system discussed herein are leveraged in association with an electric power system.

[0106] The SCADA system and / or the predictive system may be operative to perform the method of any aspect or embodiment disclosed herein. Vice versa, the methods disclosed herein may be performed by or using the predictive system, the SCADA system, and / or the electric power system according to any aspect or embodiment disclosed herein.

[0107] The electric power system may be or may comprise an HVDC system, in particular an HVDC transmission system.

[0108] According to another aspect of the invention, there is provided machine-readable instruction code comprising machine-readable instructions which, when executed by at least one processing circuit, cause the at least one processing circuit to perform the method according to an aspect or embodiment of the invention.

[0109] The effects attained by the machine-readable instruction code correspond to the effects disclosed in association with the methods, predictive systems, and SCADA systems according to various embodiments.

[0110] According to another aspect of the invention, there is provided non-transitory storage medium having stored thereon machine-readable instruction code comprising machine-readable instructions which, when executed by at least one processing circuit, cause the at least one processing circuit to perform the method according to an aspect or embodiment of the invention.

[0111] The effects attained by the non-transitory storage medium correspond to the effects disclosed in association with the methods, predictive systems, and SCADA systems according to various embodiments.

[0112] Various effects and advantages are attained by embodiments of the invention. For illustration, the methods and processing provide enhanced techniques of performing predictive asset health assessment. The methods and system are operative in association with a control and protection system of a Supervisory Control and Data Acquisition (SCADA) system in a manner that can reduce downtimes during maintenance.

[0113] The processing systems and methods can be used in association with a HVDC power transmission system or subsystems thereof, such as a power system substation (e.g., an inverter / converter, components comprising power semiconductor devices, such as power semiconductor devices having a control gate (e.g., thyristors or Insulated Gate Bipolar Transistors (IGBTs)), a valve hall comprising valves including thyristors or IGBTs), without being limited thereto.

[0114] BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Embodiments of the invention will be described with reference to the drawings in which similar or identical reference signs designate elements with similar or identical configuration and / or function.

[0116] Figure 1 is a schematic diagram of an electric power system comprising a Supervisory Control and Data Acquisition (SCADA) system.

[0117] Figure 2 is a schematic representation of the SCADA system and first and second electric subsystems.

[0118] Figure 3 is a block diagram of the processing system.

[0119] Figure 4 is a flow chart of a method.

[0120] Figure 5 is a block diagram of a predictive system of the SCADA system.

[0121] Figure 6 is a schematic representation of a data-driven processing that can be performed by the predictive system. Figure 7 is a schematic representation of a data-driven processing that can be performed by the predictive system.

[0122] Figure 8 is a schematic representation of a data-driven processing that can be performed by the predictive system.

[0123] Figure 9 is a schematic representation of processing operations that can be performed by the predictive system.

[0124] Figure 10 is a schematic representation of the SCADA system and first and second electric subsystems.

[0125] Figure 11 is a flow chart.

[0126] Figure 12 shows control and protection device(s) of the SCADA system and associated primary system equipment.

[0127] Figure 13 is a flow chart.

[0128] DETAILED DESCRIPTION OF EMBODIMENTS

[0129] Embodiments of the invention will be described with reference to the drawings. In the drawings, similar or identical reference signs designate elements with similar or identical configuration and / or function.

[0130] Embodiments relate to methods and predictive systems operative to perform a predictive asset health assessment. The predictive system is operative to communicatively interface with first and second control and protection (C&P) subsystems of a Supervisory Control and Data Acquisition (SCADA) system for an electric power system. The first and second control and protection subsystems are respectively operative to be operatively coupled to the electric power system such that the first control and protection subsystem performs control and protection functions for a first electric power subsystem of the electric power system and that the second control and protection subsystem performs the same control and protection functions (i.e., other instances of the same functions but executed on different devices) for a second electric power subsystems of the electric power system. The first and second electric power subsystems provide redundant implementations of primary system equipment. This may be implemented in such a manner that a converter / inverter asset can continue operation in a seamless manner when a control or protection device of the first control and protection subsystem is set to maintenance mode, for example, with the second electric power subsystem then carrying the currents that are otherwise carried by the first electric power subsystem. Thus, one of the first and second electric power subsystems provides a redundant implementation (also referred to as backup), but now on the primary system level, for the other of the first and second electric power subsystems. As used herein, "predictive asset health assessment" refers to determining a prediction for a future asset state or degradation of at least one asset. The predictive asset health assessment may comprise determining a quantity indicative of asset health or asset degradation for a plurality of times in the future, e.g., as a prediction of an asset health or asset degradation evolution. The predictive asset health assessment may comprise determining a remaining useful life (RUL), an end of life, and / or another quantity indicative of asset health or asset degradation. The asset health assessment may be performed for at least one, preferably for several and more preferably for all, assets to be monitored of the first electric power subsystem and for at least one, preferably for several and more preferably for all, assets to be monitored of the second electric power subsystem.

[0131] The techniques disclosed herein utilize first and second control and protection subsystems, with the predictive system being operative to communicatively interface with the first and second control and protection subsystems. The first and second control and protection subsystems implement redundancy of the control and protection functions performed. For every instance of a control or protection function performed by the first control and protection subsystem to provide control or protection of a first asset of the first electric power subsystem, the second control and protection system may be operative to perform another instance of the same control and protection function to provide control or protection of a second asset of the second electric power subsystem, with the second asset being a backup for the first asset. In particular, the second asset may have a configuration, interconnections, and operation in the second electric power subsystem that is identical to a configuration, interconnections, and operation of the first asset in the first electric power subsystem.

[0132] The provision of first and second control and protection subsystems as disclosed in the present application may use, for example, the Hitachi Energy Modular Advanced Control for HVDC (MACH™) control system. According to embodiments disclosed in detail herein, the system is enhanced by the predictive system. Thus, the methods and systems disclosed herein address the fact that conventional SCADA systems are operative to perform on continuous operation monitoring, control and protection in the case of faults. The SCADA system may also have alarm and event handling capabilities that generate alarms in cases of any abnormalities based on the logics and to be notified to an operator screen via a human machine interface (HMI). Events take care of, e.g., causing trip actions in case of faults to protect the system, so that it can be attended to by the personnel at site before putting it back to operation. The predictive system and methods disclosed herein provide, in addition, the capability of monitoring degradations, aging or other aspects related to the maintenance aspects.

[0133] Conventional maintenance techniques are time based or corrective approaches. The predictive systems and methods disclosed herein provide an approach to deduce inputs for predictive maintenance harnessing information from the first and second control and protection subsystems. This may be implemented with providing the predictive system as an additional information handling layer in the SCADA system. Data-driven methods may be employed for an analysis of data that can be collected in parallel to the existing channels but with a particular view to predictive asset health assessments. Thus, there are provided automated techniques for predictive maintenance.

[0134] While embodiments will be described in detail primarily in association with SCADA systems for high voltage direct current (HVDC) power system, in particular HVDC transmission systems, the embodiments are not limited thereto.

[0135] As used herein, the term "SCADA system" encompasses in particular a control and protection system, e.g., a control and protection system of a HVDC system.

[0136] The techniques disclosed herein may involve the application of artificial intelligence (Al) models. The Al-based techniques may comprise the use of machine learning (ML) techniques and / or deep learning (DL) techniques. As used herein, the term Al is intended to encompass ML techniques and DL techniques, without being limited thereto. The techniques may comprise or may be supervised, semisupervised , unsupervised or self-supervised techniques to process data captured at the installation site, using data previously captured at the test site.

[0137] The techniques disclosed herein may use subsystems, in the primary and / or secondary system parts, which are referred to as "first" and "second" subsystems herein. Such subsystems are also referred to as "A" and "B" subsystems or as "active" and "standby" subsystems in the art. The "first" and "second" subsystems may in particular be subsystems of or for a same station (e.g., a station at a local end or at a remote end). Thus, the first" and "second" subsystems are distinguished from subsystems provided in separate stations (e.g., stations at a local and remote end). It will be appreciated that the techniques disclosed herein with relation to the "first" and "second" subsystems of a station may equally be applied to each one of the two stations. e., the techniques disclosed herein may be applied to a first station (which may have its first and second subsystems) and a second station provided on an opposite end of a line as the first station, with the second station optionally having its own first and second subsystems.

[0138] The methods and processing systems disclosed herein are particularly useful in association with HVDC systems. The HVDC system may comprise or may be a HVDC power transmission system. A power transmission line of the HVDC system may have a length of at least 100 km, at least 200 km, at least 300 km, or at least 400 km. HVDC transmission systems offer various advantages, in particular for power transmission over longer distances. HVDC transmission systems are gaining popularity for, e.g., power transmission from a renewable energy source to an area that may be remote from the location of the renewables energy source. Thus, the techniques disclosed herein are particularly useful in association with a transition to more environmentally friendly electric power systems that comprise renewables energy source and HVDC transmission systems.

[0139] As used herein, the term "power" refers to electric power unless explicitly stated otherwise. Figure 1 is a schematic representation of an electric power system 10 comprising a predictive system 70 operative to perform a predictive asset health assessment and generate and provide output based on the predictive asset health assessment. The electric power system 10 comprises a primary system (i.e., a system that carries the currents and has the voltages for electric power transmission performed as part of an international, national, or regional power grid). The primary system comprises a high voltage direct current (HVDC) system 20. The primary system may comprise a first AC system 11, a first transformer 13 (e.g., a step-up transformer 13), a second transformer 14 (e.g., a step-down transformer), and a second AC system 15, with the HVDC system 20 being operative to provide DC power transmission between the first and second AC systems 15.

[0140] The HVDC system 20 may comprise first and second converters 21, 22 located on opposite ends relative to a DC transmission line 23, 23'. The converters 21, 22 are also referred to as converter / inverter assets in the art, as they can perform not only inversion but also other conversion functions. Each or both of the converters 21, 22 may be or may comprise a line-commutated current-sourced converter (LCC). The LCC may comprise thyristor valves. Each or both of the converters 21, 22 may be or may comprise a Voltage Source Converter (VSC). The VSC may comprise Insulated Gate Bipolar Transistor (IGBT) valves. The inverter / converter 21, 22 may have other configurations that may include power semiconductor devices (e.g., power semiconductor devices having a control gate).

[0141] The HVDC system 20 may comprise equipment controllable to protect the HVDC system 20. This equipment may comprise grounding switches 24, 25, 24', 25' for DC protection and / or AC / DC circuit breakers (CBs).

[0142] The HVDC system 20 may provide a redundant system implementation having first and second subsystems of the primary system that are redundant implementations of each other. This may be implemented in such a manner that continuous converter operation can be ensured.

[0143] There does not have to be a redundant implementation for all primary system components. Redundancy may be provided for at least part of the primary system components, with the respective implementations being referred to as first / active / "A" and second / backup / "b" subsystems of the primary system herein. The first and second primary system subsystems may be comprised by a same station, i.e., they may be both arranged at a same line relative to an HVDC transmission line or cable.

[0144] As will be described in more detail herein, different first and second control and protection subsystems may be associated with the different first and second electric power subsystems (i.e., the different redundant parts of the primary system) to provide redundant implementations of the control and protection functions performed (i.e., to run different instances of the same control or protection functions), with the different first and second control and protection subsystems operating during field operation to ensure availability of the SCADA systems to control and operate main station functions. The electric power system 10 comprises a SCADA system 40. The SCADA system 40 comprises a plurality of SCADA system devices 41-43, 45-49. The SCADA system devices may comprise control and protection devices 41-43, 45-47 and sensors 48, 49. The control and protection devices 41-43, 45-47 may be communicatively coupled, via physical connections or signal transmission, and / or via wireless links (such as wireless point-to-point links) to the sensor devices 48, 49 to receive measurements therefrom. The control and protection devices may also be communicatively coupled, via wired or wireless links, to a human machine interface (HMI) of the SCADA system 40.

[0145] As mentioned above, redundancy is implemented both at the primary system level (i.e., the components carrying the large currents or having the large voltages of the transmission grid) and at the secondary system level (i.e., in the SCADA system). The SCADA system 40 thus is operative such that for any instance of a control or protection function performed for the first electric power subsystem (primary system level) by the first control and protection subsystem (secondary system level), there is a corresponding instance of the same control or protection function performed for the second electric power subsystem (primary system level) by the second control and protection subsystem (secondary system level). The instances are preferably executed on different physical devices to implement the redundancy.

[0146] The control and protection devices 41-43, 45-47 and a central control and protection system may be operative to perform control and protection functions, such as valve control of converter valves, grounding switch control, AC / DC CB control, without being limited thereto.

[0147] The SCADA system 40 comprises a communication system 60. The communication system 60 may comprise wired communication links and / or wireless communication links. For illustration, the communication system 60 may comprise RS-485, CAN, etherCAT, Ethernet, modbus, IEEE 61850 wired connections and / or wireless communication links with a carrier bandwidth in, e.g., the sub-1 GHz range, without being limited thereto. The communication system 60 may comprise a communication switch device 61 or edge device.

[0148] The electric power system 10 comprises the predictive system 70. The predictive system 70 is operative to predict aging, degradation, and / or other conditions related to when maintenance is required, such as maintenance of primary system equipment of the electric power system. This is also referred to as predictive asset health assessment. As compared to assessing the current state only, the predictive asset health assessment involves a prediction over a predictive time horizon.

[0149] The predictive system 70 may be operative to perform the predictive asset health assessment by communicatively interfacing with the first and second control and protection subsystems of the SCADA system 40. The predictive system 70 may be operative to use data obtained from the first and second control and protection subsystems of the SCADA system 40 in various ways, such as by using configuration data from the first and second control and protection subsystems to detect configuration changes and discriminate configuration changes from degradation-caused changes, by using online measurement data (which may comprise timeseries data or event-based data) to infer drifts or other trends, by storing at least a sub-set of the obtained data for later use in association with other offline data, or using other techniques described herein, without being limited thereto.

[0150] Figure 2 is a schematic representation diagram of a SCADA system 40 comprising a first control and protection subsystem 62 and a second control and protection subsystem 65. The first control and protection subsystem 62 comprises a first set of control and protection devices 63. The control and protection devices of the first set 63 may be operative to perform control and protection functions for a first electric power subsystem 30 of, e.g., an HVDC transmission system. The first control and protection subsystem 62 comprises a first set of sensors 64. The control and protection devices of the first set 63 may be operative to perform the control and protection functions for the first electric power subsystem of the HVDC transmission system based on sensor output of sensors included in the first set 64 and independently of sensor output of sensors included in a second set 67 that will be described below.

[0151] The second control and protection subsystem 65 comprises a second set of control and protection devices 66. The control and protection devices of the second set 66 may be operative to perform control and protection functions for a second electric power subsystem 32 of, e.g., an HVDC transmission system. The second control and protection subsystem 65 comprises a second set of sensors 67. The control and protection devices of the second set 66 may be operative to perform the control and protection functions for the second electric power subsystem of the HVDC transmission system based on sensor output of sensors included in the second set 67 and independently of sensor output of sensors included in the first set 64.

[0152] The first electric power subsystem 30 may comprise primary system equipment 31 controllable by the first control and protection subsystem 62. The second electric power subsystem 32 may comprise primary system equipment 33 controllable by the first control and protection subsystem 62. The primary system equipment 31, 33 of each of the first and second electric power subsystems 30, 32 may respectively comprise one or several transformers, one or several DC grounding switches, one or several AC / DC CBs, converter / inverter cooling systems, without being limited thereto.

[0153] The SCADA system 40 comprises a change-over logic 69 operative to cause operation to switch between the first control and protection subsystem 62 and its associated first electric power subsystem 30 of the HVDC system and the second control and protection subsystem 66 and its associated second electric power subsystem 32 of the HVDC system.

[0154] The SCADA system 40 comprises the predictive system 70. The predictive system 70 is operative to communicatively interface with the first control and protection subsystem 62 and the second control and protection subsystem 65 to obtain electric power system information therefrom and to use the obtained electric power system information to perform the predictive asset health assessment. The obtained electric power system information comprises both information on the first electric power subsystem 30 and the first control and protection subsystem 62 operatively coupled thereto, and on the second electric power subsystem 32 and the second control and protection subsystem 65 operatively coupled thereto.

[0155] The predictive system 70 may be operative to obtain at least configuration data from the first and second control and protection subsystems 62, 65. The configuration data may define a configuration of the first electric power subsystem 30 (such as information included in a single line diagram thereof and / or hardware configurations of the primary system equipment and / or nameplate data of the primary system equipment of the first electric power subsystem 30), a configuration of the first control and protection subsystem 62 (such as information on one, several, or all of device and sensor hardware configurations, device and sensor names, names of function instances, binding of function instance inputs to incoming signals, names of function instance outputs of the first control and protection subsystem 62), a configuration of the second electric power subsystem 32 (such as information included in a single line diagram thereof and / or hardware configurations of the primary system equipment and / or nameplate data of the primary system equipment of the second electric power subsystem 32), and a configuration of the second control and protection subsystem 65 (such as information on one, several, or all of device and sensor hardware configurations, device and sensor names, names of function instances, binding of function instance inputs to incoming signals, names of function instance outputs of the second control and protection subsystem 65). The predictive system 70 may be operative to use this configuration data to, e.g., determine which changes in electric power subsystem behavior of the first and second electric power subsystems 30, 32 may be caused by different hardware and which may be caused by maintenance related aspects (such as aging, degradation, or other maintenance related aspects).

[0156] The predictive system 70 may optionally be operative to obtain operation data from the electric power system information from the first control and protection subsystem 62 and the second control and protection subsystem 65. The operation data may comprise sensor outputs of at least some sensors of the first set 64 and of at least some sensors of the second set 65. The operation data may comprise inputs and / or outputs of function instances of at least some control or protection functions performed by the first control and protection subsystem 62 and inputs and / or outputs of function instances of at least some control or protection functions performed by the second control and protection subsystem 65. This information may be obtained as online data as it becomes available. The predictive system 70 may be operative to use offline data that may be based on the operation data to, e.g., determine drifts or other trends based on the operation data or data derived therefrom, as aggregated over a longer period. The predictive system 70 is operative to use the obtained power system information for performing the predictive asset health assessment. As the predictive system 70 is communicatively interfaced with both the first control and protection subsystem 62 and the second control and protection subsystem 65, the predictive system 70 can use the data obtained from both subsystems in combination to determine trends and forecast, based on the observed trends, how asset health will continue to evolve over a predictive time horizon. The availability of data from both the first control and protection subsystem 62 and the second control and protection subsystem 65 can be used in such a manner that the predictive system 70 is operative to use data from one of the the first control and protection subsystems 62, 65 as a baseline for a comparison of data from the other one of the the first and second control and protection subsystems 62, 65 therewith. Thereby, trends can be established more reliably than by considering the time evolution of the data from each of the control and protection subsystems 62, 65 in isolation from each other.

[0157] The predictive system 70 may be operative to use various techniques, in isolation or in combination, to process the data obtained from the first and second control and protection subsystems 62, 65 to perform the predictive asset health assessment. The processing system 70 may be operative to compare the data obtained from one of the first and second control and protection subsystems 62, 65 to data obtained from the other one of the first and second control and protection subsystems 62, 65 and / or may process the data using a drift detection logic that may be based, at least in part, on a data-driven processing technique and / or may process data obtained using sensors 68 not operatively associated with any control or protection function of the first and second control and protection subsystems 62, 65.

[0158] The predictive system 70 may be operative to control at least one HMI to provide output that is based on the predictive asset health assessment. Alternatively or additionally, the predictive system 70 may be operative to generate output that allows the change-over logic 69 to place at least one device of one of the first and second control and protection systems into maintenance mode.

[0159] Figure 3 is a block diagram of the predictive system 70. The predictive system 70 comprises at least one interface 71, a storage system 72, and at least one processing circuit 80. The at least one processing circuit 80 may be operative to perform the operations disclosed below in several processing layers.

[0160] The at least one interface 71 is operative to receive electric power system information 74, 74', 75, 75'.

[0161] The electric power system information may comprise at least configuration data 75, 75' from the first and second control and protection subsystems 62, 65. The configuration data 75, 75' may comprise a standardized configuration description (SCD), a proprietary configuration description, and / or a standardized configuration language (SCL)-based description of configurations. The configuration data 75, 75' may be machine-readable configuration files or other machine-readable configuration descriptions.

[0162] The configuration data 75, 75' may define a configuration of the first electric power subsystem 30 (such as information included in a single line diagram thereof and / or hardware configurations of the primary system equipment and / or nameplate data of the primary system equipment of the first electric power subsystem 30), a configuration of the first control and protection subsystem 62 (such as information on one, several, or all of device and sensor hardware configurations, device and sensor names, names of function instances, binding of function instance inputs to incoming signals, names of function instance outputs of the first control and protection subsystem 62), a configuration of the second electric power subsystem 32 (such as information included in a single line diagram thereof and / or hardware configurations of the primary system equipment and / or nameplate data of the primary system equipment of the second electric power subsystem 32), and a configuration of the second control and protection subsystem 65 (such as information on one, several, or all of device and sensor hardware configurations, device and sensor names, names of function instances, binding of function instance inputs to incoming signals, names of function instance outputs of the second control and protection subsystem 65). The predictive system 70 may be operative to use this configuration data to, e.g., determine which changes in electric power subsystem behavior of the first and second electric power subsystems 30, 32 may be caused by different hardware and which may be caused by maintenance related aspects (such as aging, degradation, or other maintenance related aspects).

[0163] The predictive system 70 may be operative such that receipt of the configuration data 75, 75' is repeated continually (i.e., in an ongoing manner) during operation of the SCADA system with the predictive system 70 comprised in the SCADA system. Thereby, the predictive system 70 becomes aware of configuration changes that influence its predictive asset health assessment.

[0164] The electric power system information may comprise operation data from the electric power system information from the first control and protection subsystem 62 and the second control and protection subsystem 65. The operation data may comprise sensor outputs of at least some sensors of the first set 64 and of at least some sensors of the second set 65. The operation data may comprise inputs and / or outputs of function instances of at least some control or protection functions performed by the first control and protection subsystem 62 and inputs and / or outputs of function instances of at least some control or protection functions performed by the second control and protection subsystem 65. This information may be obtained as online data as it becomes available. The predictive system 70 may be operative to use offline data that may be based on the operation data to, e.g., determine drifts or other trends based on the operation data or data derived therefrom, as aggregated over a longer period. The at least one processing circuit 80 may be operative to implement a retrieval 81 of the electric power system information. The retrieval 81 may comprise selectively determining which received data includes information suitable for performing a predictive asset health assessment for at least one asset. For illustration, data unrelated to the asset(s) for which the predictive asset health assessment is to be performed may be discarded. The retrieval 81 may comprise selective storage of configuration data 73 and operation data 74 received via the at least one interface 71 for subsequent use.

[0165] The at least one processing circuit 80 is operative to perform a predictive asset health assessment 82 based on the electric power system information. The predictive asset health assessment 82 may comprise using both data obtained from the first control and protection subsystem 62 (such as configuration and / or operation data) and data obtained from the second control and protection subsystem 65 (such as configuration and / or operation data) to perform a predictive asset health assessment for an asset (and optionally for several or all assets) of the first electric power subsystem 30. The predictive asset health assessment 82 may comprise using both data obtained from the first control and protection subsystem 62 (such as configuration and / or operation data) and data obtained from the second control and protection subsystem 65 (such as configuration and / or operation data) to perform a predictive asset health assessment for an asset (and optionally for several or all assets) of the second electric power subsystem 32. In this way, combining the information from both control and protection subsystems 62, 65 allows predictive asset health assessment to be performed more reliably. In particular, the data obtained from one of the control and protection subsystems can act as baseline for a comparison with the data obtained for the other of the control and protection subsystems. The comparison need not be explicit but may also be encoded in, e.g., a logic that receives both data obtained from the first control and protection subsystem 62 (such as configuration and / or operation data) and data obtained from the second control and protection subsystem 65 (such as configuration and / or operation data) as input and that outputs the predictive asset health assessment for an asset (and optionally for several or all assets) of, e.g., the first electric power subsystem 30.

[0166] Performing the predictive asset health assessment may comprise performing a drift detection 84 or other trend detection to detect time-dependent changes that allow the future evolution of an asset health (such as an asset state that may be selected from "normal", "incipient", "degraded") to be predicted over a predictive time horizon. This may comprise determining the RUL, the end of life, or a time evolution of probabilities indicating that the respective asset will be "normal", "incipient", "degraded" at a time in the future and preferably at several times in the future over the predictive time horizon.

[0167] The predictive asset health assessment 82 may use data-driven techniques, such as the use of a support vector machine (SVM) or other data-driven processing techniques that process configuration data and operation data to determine the RUL, end of life, or other indicator for asset health over a predictive time horizon. Features of such techniques will be described in more detail with reference to Figure 5 to Figure 9 below.

[0168] The at least one processing circuit 80 may be operative to perform an interface control 83 to cause output 79 to be generated and provided, which output 79 is dependent on a result of the predictive asset health assessment. The output may comprise control commands to control at least one HMI and / or control commands usable by, e.g., the change-over logic 69 or another device of the SCADA system 40 to perform a control operation.

[0169] The at least one processing circuit 80 may comprise any one or any combination of integrated circuits, integrated semiconductor circuits, processors, controllers, application specific integrated circuits (ASICs), circuit(s) including quantum bits (qubits) and / or quantum gates, without being limited thereto, to perform the mentioned functions.

[0170] Figure 4 is a flow chart of a method 100. The method 100 may be performed automatically by or using the processing system 70.

[0171] At process block 101, the processing system 70 obtains configuration data. The configuration data may define a configuration of the electric power subsystem 30, the first control and protection subsystem 62, the second electric power subsystem 32, and the second control and protection subsystem 65. Retrieval or receipt of the configuration data may be repeated continually by the processing system 70 to detect configuration changes.

[0172] At process block 102, the processing system 70 receives operation data. The operation data may comprise sensor output (such as one, several, or all of measurements relating to electric characteristics, temperature measurements, measurements relating to insulation fluid (e.g., DGA results or cooling fluid temperature measurements)).

[0173] At process block 103, the processing system 70 uses the obtained data to perform a predictive asset health assessment. The predictive asset health assessment may be performed in various ways, which may include, for example: using a legacy technique (such as any of the techniques disclosed in EP 3 923 214 Al, EP 3 923 213 Al, EP 3 923 101 Al, which are applicable to electric power system assets) to determine a first predictive asset health assessment for an asset of the first electric power subsystem 30 from the data obtained from the first control and protection subsystem 64, determine a second predictive asset health assessment for an asset of the second electric power subsystem 32 from the data obtained from the second control and protection subsystem 64, and optionally use the first and second predictive asset health assessment in combination to detect, e.g., discrepancies therebetween that warrant at least one of the first and second predictive asset health assessment to be changed (e.g., by further reducing the RUL of an asset if the comparison shows that the asset may have degraded more than originally anticipated); using other data-driven techniques such as a SVM or trained artificial intelligence (Al) model to determine the predictive asset health assessment; features of such techniques will be described in association with Figure 5 to Figure 9; using an historical evolution of an asset health for an asset of one of the first and second electric power subsystems for predictive asset health assessment for an asset of the other one of the first and second electric power subsystems (e.g., when there is no available data for the asset of the other one of the first and second electric power subsystems); without being limited thereto.

[0174] Changes in statistical properties over a period of time may also be tracked.

[0175] At process block 104, the processing system 70 generated and provides output based on a result of the predictive asset health assessment. Generating the output may comprise controlling an HMI to provide information on the predictive asset health assessment and / or maintenance scheduled based on the predictive asset health assessment, and / or providing control commands to other SCADA system devices (such as the change-over logic 69) to automatically perform control operations, optionally pending operator approval. A change in statistical properties over a period of time may also be tracked and used to generate the output.

[0176] At process block 105, the processing system 70 may determine, based on the configuration data receives, whether there has been a change in configuration. As previously noted, the processing system 70 may continually receive configuration data from both the first control and protection subsystem and the second control and protection subsystem to detect configuration changes. Unless there is a change in configuration, process blocks 102 to 105 may be repeated.

[0177] At process block 106, responsive to detecting a change in configuration, the processing performed for the predictive asset health assessment may be modified. For illustration, maintenance on an asset of the second electric power subsystem 32 may cause discrepancies between the observations for the first and second electric power subsystems, but this is not indicative of a degradation of the asset of the first electric power subsystem 30. This may be taken into account in the following iterations.

[0178] Figure 5 is a block diagram of the predictive system 70 to explain additional features that may be implemented.

[0179] The predictive system 70 is operative to maintain a knowledge base 76. The at least one processing circuit 80 may be operative to perform a storage control 85 to store at least part of the received data 74, 74', 75, 75' and / or part of the output 79. By storing part of the output 79 or data derived therefrom, a knowledge base 76 of the predictive system 70 may be augmented over time. This knowledge, which may include previously determined results of the predictive asset health assessment, may be used by the at least one processing circuit 80 to perform the predictive asset health assessment. For illustration, degradation as previously predicted may be compared to more recent observations to update asset health assessments. Alternatively or additionally, the at least one processing circuit 80 may use the degradation as previously predicted as a starting point to generate a refined predictive asset health assessment that takes into account more recent observations.

[0180] The at least one processing circuit 80 may be operative to perform a data-driven processing 90 to implement the predictive asset health assessment. The data-driven processing may be based on an SVM, a trained Al model, or other processing techniques having parameters set in a data-driven manner. Parameters of the data-driven processing 90 as determined during training may be stored in data 75 that may comprise model parameters and / or hyperparameters (such as a number of layers of an Al model) and may be used by the at least one processing circuit 80 to perform the predictive asset health assessment. The at least one processing circuit 80 may be operative to retrain the data-driven processing 90 as more observations become available, e.g., by comparing predicted changes in asset health (e.g., a predicted degradation) to observations over the predictive time horizon, and by then adjusting the parameters of the data-driven processing 90 to account for any differences between previous predictions and actual observations.

[0181] The processing 90 of the predictive asset health assessment may be based on historical data for asset health changes over time. Thus, the processing 90 of the predictive asset health assessment may be performed in a data-driven manner, which makes the processing 90 of the predictive asset health assessment more objective than, e.g., human expert based logic designs.

[0182] To train the processing 90, historical data (e.g., data for degradation of at least 100, at least 200, at least 300, or at least 500 assets) may be used. The historical data may be partitioned into a training set, a test set, and a validation set. For illustration rather than limitation, the training set may be, e.g., 70% of the historical data, the test set may be, e.g., 15% of the historical data, and the validation set may be, e.g., 15% of the historical data. The split of available data for training, testing and validations can be performed based on the volume of data available, computing power of the device doing the processing. There are various possible implementations, such as using small batches for training that can also accommodate online batches to update the models or in offline mode. When there is static data, a 70 to 30% split considering the types of methods (supervised, unsupervised and semisupervised methods) can be applied, without being limited thereto. Data augmentation may be performed if desired to, e.g., have a more balanced set of cases with and without abnormality.

[0183] The generation of the processing 90 of the predictive asset health assessment may comprise setting decision thresholds in a data-driven manner. More complex techniques, such as techniques using at least one Al model to perform processing, may be used. The at least one Al model may comprise at least one Al model operative to process operation data, which may comprise timeseries 1 data of measurements and / or timeseries of control and protection function outputs of the first and second control and protection subsystems.

[0184] Features of the data-driven processing that may be used by the predictive system 70 will be described in more detail below with reference to Figure 6, Figure 7 , Figure 8, and Figure 9.

[0185] The processing 90 of the predictive asset health assessment may comprise at least one neural network (NN), which may comprise a convolutional NN (CNN) or recurrent NN (RNN).

[0186] Figure 6 is a schematic representation of processing 90 that comprises a NN model. The NN model has an input layer 91 operative to receive at least part of the electric power system information, e.g., operation data. The NN model has an output layer 92 operative to provide the predictive asset health assessment, e.g., a RUL, an end of life, or other quantities related to maintenance, e.g., an asset state (e.g., "normal", "incipient", "degraded") at at least one time in the future or probabilities for the asset having a certain asset health state (e.g., probabilities for "normal", "incipient", "degraded" asset states at at least one time in the future).

[0187] The processing 90 of the predictive asset health assessment may comprise a long short term memory (LSTM) cell.

[0188] Figure 7 shows the LSTM cell 130. The processing system 70 may be operative such that the model data 75 comprises parameters of the LSTM cell 130. The processing system 70 may be operative to perform the predictive asset health assessment using an Al model that comprises at least one LSTM cell 130, optionally a plurality of LSTM cells.

[0189] In Figure 7, C designates a cell state. The parameter h designates a hidden state. The parameter x designates an input. The subscript respectively designates the time. The subscript t designates the time of the input processed by the cell illustrated in Figure 7. The subscript t - 1 designates the preceding time.

[0190] During training, parameters of the LSTM cell or of several LSTM cell arranged in a stacked structure may be trained. The training may comprise training of a forget gate (designated by ft), an input gate (designated by it) and an output gate (designated by ot).

[0191] Training the various gates may comprise training the weight parameters and bias parameters of each of the gates of the LSTM cell.

[0192] As previously explained, the LSTM cell(s) may be trained using historical data of asset health changes.

[0193] During inference, the LSTM cell(s) or an Al model comprising the LSTM cell(s) and optional pre- and / or post-processing (such as a normalization layer, a feed forward neural network, a CNN, a RNN, and / or another deep learning (DL) or shallow learning model used in combination with the LSTM cell (s)) may be used to process at least part of the electric power system information to obtain the predictive asset health assessment. The processing 90 of the predictive asset health assessment may comprise an Al model having at least one attention mechanisms, e.g., a self-attention mechanisms, a cross-attention mechanism, a multi-head self-attention mechanisms, and / or a multi-head cross-attention mechanism. The Al model may comprise an Al-transformer comprising a stack of Al-transformer model blocks (also referred to as attention blocks), each comprising at least one attention mechanisms. In the art, such a model is also known as a "transformer." To clearly distinguish the Al model from the physical entity of a power system transformer, the term "Al-transformer" is used to refer to the Al model comprising one or several attention mechanisms.

[0194] Figure 8 shows the Al-transformer which may comprise an Al-transformer encoder 140 and / or an Al-transformer decoder. The processing system 70 may be operative such that the model data 75 comprises parameters of the Al-model transformer. The processing system 70 may be operative to perform the predictive asset health assessment using an Al model that comprises at least one Al- transformer, optionally an Al-transformer encoder and / or decoder having a plurality 142 of Al- transformer blocks 143. The plurality of Al-transformer blocks may be stacked so that the output of one Al-transformer block is input to the subsequent Al-transformer block.

[0195] Each Al-transformer block 143 comprises at least one attention mechanisms 145, e.g., a selfattention mechanism, a cross-attention mechanism, a multi-head self-attention mechanisms, and / or a multi-head cross-attention mechanism. Each Al-transformer block 143 may comprise pre- and / or post-processing (such as a normalization layer 144, a feed forward neural network 146, a convolutional neural network, a recurrent neural network, and / or another deep learning (DL) or shallow learning model).

[0196] The Al-transformer encoder and / or decoder may comprise a token generation 141 that processes the Al-transformer input to a token. The token processing 141 may map a sequence of input values (such as datapoints of a timeseries) to query (Q), key (K) and value (V) of a QKV transformer model.

[0197] During training, at least parameters of the processing 141 (such as parameters that define how the Al model input is mapped to the Q, K, and V values of the Al-transformer model) may be trained.

[0198] During inference, the Al-transformer model(s) or an Al model comprising the Al-transformer model(s) and optional pre- and / or post-processing (such as a normalization layer, a CNN, a RNN, and / or another deep learning (DL) or shallow learning model used in combination with the LSTM cell(s)) is used to process the electric power system information to perform the predictive asset health assessment.

[0199] The Al model may comprise an Al model input operative to receive at least part of the electric power system data or Al model input derived from the electric power system data by preprocessing (e.g., obtained by filtering, normalization, or other preprocessing) and an Al model output, with the at least one processing circuit 80 being operative to generate the output indicative of an asset health assessment over a predictive time horizon in the future. For illustration, the Al-transformer encoder 140 and / or an Al-transformer decoder may have an input operative to receive at least part of the electric power system data or Al model input derived from the electric power system data by preprocessing (e.g., obtained by filtering, normalization, or other preprocessing) and an Al-transformer output, with the at least one processing circuit 80 being operative to generate the output indicative of the predictive asset health assessment based on the Al-transformer output.

[0200] Various pre-processing techniques may be applied to monitoring data before the monitoring data are input to a trained Al model or other abnormality detection logic. For illustration, any one or any combination of filtering, zero padding, normalization, or other techniques may be used to bring the monitoring data to a format suitable for processing by the trained Al model or other drift detection logic.

[0201] Figure 9 shows a schematic representation of a predictive asset health assessment that may comprise pre- and post-processing used in combination with the drift detection 84 and / or the data- driven processing 90.

[0202] Preprocessing 96 may be performed on received electric power system data (e.g., on operation data). The preprocessing may comprise filtering, zero padding, normalization, or other techniques may be used to bring the monitoring data to a format suitable for the next processing steps.

[0203] Missing data replacement 97 may be performed to ensure that the drift detection 84 and / or the data-driven processing 90 can receive all input signals for which it has been set up (e.g., during training) even when the respective sensor signals are missing. For example, if the drift detection 84 and / or the data-driven processing 90 requires a sensor output (e.g., a temperature measurement) that is not available, the respective input to the drift detection 84 and / or the data-driven processing 90 may be synthetically generated. This is also referred to as missing data replacement. The generation of the missing input may comprise any one or any combination of techniques known to the skilled person, such as hard imputation (i.e., replacement by a fixed value), regression, multivariate regression, or other more complex techniques such as decision-tree techniques. Synthetically generated inputs are illustrated by filled arrows in Figure 9, while unfilled arrows represent inputs obtained from sensor readings.

[0204] The drift detection 84 and / or the data-driven processing 90 can then be performed, acting on the inputs for which it has been set up.

[0205] Cleaning 98 of the output of the drift detection 84 and / or the data-driven processing 90 may be performed, e.g., to remove unwanted or implausible outliers.

[0206] As previously explained, the predictive system 70 is operative to generate and provide output based on the predictive asset health assessment. The output may comprise control commands that, e.g., can cause devices to transition to maintenance mode. The output may alternatively or additionally comprise control commands that cause asset health related information and / or maintenance timing information derived therefrom to be output via an HMI.

[0207] Figure 10 shows a schematic representation of a system 120 comprising the SCADA system 40 that includes the first and second control and protection subsystems 62, 65 and the predictive system 70. The predictive system 70 may be operative to control a communication interface circuit to transmit cause information based on the predictive asset health assessment to be output. For illustration, the predictive system 70 may cause transmission, via a push mechanisms, of information for displaying on a dedicated maintenance portion 122 of an HMI device 121. The HMI device 121 may be a portable device, in particular a handheld device. The predictive system 70 may be operative to cause the information to be transmitted via a communication network 124 that may comprise a wide area network and / or cellular communication links for provision to a maintenance engineer 123 via the HMI device 121.

[0208] A communication network 124 is shown in Figure 10. Both a system 120 having a cloud platform and a system 120 having no cloud platform can be implemented. Presence of a cloud platform may be dependent on power system operator need and approval. Both implementations are possible, with no change in the techniques for performing predictive health monitoring. Some changes in the configuration of the deployment may be required to accommodate cloud-based implementations, with such configuration changes being known to the skilled person in the pertinent field.

[0209] Figure 11 is a flow chart of a method 130. The method 130 may be performed automatically by or using the predictive system 70. Process blocks 101, 102, 103, and 104 may be implemented as generally described with reference to Figure 4. In the method 130, process blocks 101 or 102 may also comprise retrieval of failure modes, effects and criticality analysis (FMECA) parameters. The FMECA parameters may be retrieved from a knowledge base (as described in detail herein), and may be used as one of the inputs for performing the predictive asset health assessment. Process block 104 may comprise generating FMECA parameters updated based on the analysis performed using the operational parameters.

[0210] At process block 131, the predictive system 70 may store at least part of the output provided at process block 104 data derived therefrom in a knowledge base. By storing part of the output or data derived therefrom, a knowledge base of the predictive system may be augmented over time. Storing in the knowledge base may comprise storing the FMECA parameters as updated based on the analysis performed using the operational parameters.

[0211] At process block 132, the predictive system 70 may use the knowledge base. This may comprise comparing degradation as previously predicted to more recent observations to update asset health assessments. Alternatively or additionally, process block 132 may comprise using using the degradation as previously predicted as a starting point to generate a refined predictive asset health assessment that takes into account more recent observations. Using the knowledge base may also comprise retrieving, in a later iteration through the method 130, the FMECA parameters (as updated) from the knowledge base for use in the later iteration of method 130.

[0212] Exemplary SCADA system components with which the predictive system 70 can be used in a technically beneficial manner are described in association with Figure 12. For clarity, only the control or protection devices of one of the first and second control and protection subsystems 62, 65 are shown.

[0213] Figure 12 shows a converter / inverter asset 140 comprising power semiconductor devices having a gate electrode. For illustration, and as shown in Figure 12, the converter / inverter asset 140 may comprise thyristor valves 141, 142, 143, 144 that each may comprise a plurality of thyristors. The converter / inverter asset 140 may alternatively or additionally comprise IGBT valves.

[0214] More complex configurations may be used, such as 12 pulse configurations or even more complex configurations. The converter / inverter asset 140 may be operative to convert AC to DC and / or vice versa. The converter / inverter asset 140 may be operative to provide bidirectional AC / DC conversion.

[0215] A cooling system 160 may be associated with the converter / inverter asset. The cooling system 160 may have an inlet to receive an inlet cooling fluid flow 161, and outlet to output an outlet cooling fluid flow 162, and fluidic circuit components such as a one-way valve 165 and a pump 164 to convey the cooling fluid. Various sensors 157, 158, 159 may be provided to sense an inlet temperature (inlet temperature sensor 158) and an outlet temperature (outlet temperature sensor 157) of the cooling fluid. Optional additional sensors 159 may be provided.

[0216] The SCADA system 40 comprises a control device 151. The control device 151 comprises a valve control function 152 (optionally a valve control function for each thyristor valve). In field operation, the control device 151 may be operative to perform the valve control function 152 responsive to sensor outputs of sensors 154, 155.

[0217] The SCADA system 40 comprises a further control device 155. The further control device 155 comprises a cooling system control function 156. In field operation, the further control device 155 may be operative to perform the cooling system control function 156 responsive to sensor outputs of sensors the sensors 157, 158, 159.

[0218] The predictive system 70 may be operative to monitor a degradation of, e.g., thyristor valves and / or the cooling system 160. Using data acquired from the control and protection subsystem that comprises the control devices 151, 155, and using data acquires from the other control and protection subsystem that comprises redundant implementations of the control devices 151, 155 that perform other instances of the same valve control function 152 as the control device 151 (for converter / inverter control) and another instance of the same cooling system control 156 as the further control device 155 (for cooling system control), the predictive system 70 can detect a relative drift between the various redundant subsystems, which is indicative of a change in asset health and which can be used to perform predictive asset health assessment.

[0219] While Figure 12 shows a converter / inverter having thyristor valves, the techniques disclosed herein may also be applied to a converter / inverter 180 that comprises insulated-gate bipolar transistor (IGBT) valves and / or its cooling system or to converter / inverter assets comprising other power semiconductor devices and / or their cooling systems, without being limited thereto.

[0220] An asset is typically composed of several components. For illustration, the cooling system 160 may include various fluidic circuit components 164, 165 that may degrade over time. Separate sensors may be provided for each of these components 164, 165. The predictive system 70 may be operative to determine the RUL or end of life of the asset based on the RUL and / or end of life of its components. For illustration, the predictive system 70 may be operative to determine the RUL of each of the components of an asset and may determine the RUL of the asset as shortest of the RUL values. Alternatively or additionally, the predictive system 70 may be operative to determine the end of life of each of the components of an asset and may determine the end of life of the asset as earliest of the end of life values.

[0221] Figure 13 is a flow chart of a method 160 according to an embodiment. The method 160 may be performed automatically by a system that comprises the SCADA system 40 that comprises the first control and protection subsystem, the second control and protection subsystem, and the predictive system.

[0222] At process block 161, the first control and protection subsystem performs control and protection functions for the first electric power subsystem, and the second control and protection subsystem performs control and protection functions for the second electric power subsystem. The control and protection functions may comprise control of cooling systems for converter / inverter assets, and / or control of switches or CBs to perform protection functions.

[0223] At process block 162, the predictive system 162 perform the predictive asset health assessment based on data obtained from the first electric power subsystem and the second electric power subsystem.

[0224] At process block 163, a control action is performed based on a result of the predictive asset health assessment. The control action may comprise HMI control and / or control actions that cause a switchover between the first and second electric power subsystems and / or control actions that cause a device of at least one of the first and second control and protection subsystems to enter maintenance mode.

[0225] As explained with reference to Figure 1 to Figure 13, the invention provides systems and methods operative to perform predictive asset health assessment, taking advantage of the redundancy provided by first and second control and protection subsystems and of the first and second electric power subsystems with which they are associated. Thereby, improved techniques are provided that facilitate detecting asset issues that require maintenance in a predictive manner and pro-actively taking appropriate corrective action.

[0226] The techniques disclosed herein, in which the predictive system 70 is added to redundant SCADA subsystems, may be used in association with, e.g., Hitachi Energy's Modular Advanced Control for HVDC system, which enables a high degree of integration and management for all control and protection functions with advanced fault detection and remote-control functions. The techniques disclosed herein, in which the predictive system 70 is added to conventional control and protection subsystems, can generally be used in association with SCADA systems that provide redundancy to improve the reliability and availability. An advantage of this approach is that control equipment maintenance does not require any shutdown of main circuit equipment. The redundancy may in particular involve a hardware redundancy, with distinct devices being provided to perform instances of the same protection functions that are redundant implementations in the sense that one instance performs the same control or protection function for the first electric power subsystem as the other instance does for the second electric power subsystem.

[0227] Conventional SCADA systems that provide redundancy focus on continuous operation monitoring and control. Provision of the additional predictive system 70, which may be an integral part of the SCADA system, provides a channel for continuous monitoring for predictive maintenance. With an infrastructure of 25 to 30 years of service, much of the HVDC equipment needs attention. Traditional maintenance practices have various shortcomings, and these are mitigated by the techniques disclosed herein.

[0228] The systems and methods disclosed herein may take advantage of an increasing availability of measurement data and data-driven techniques. The systems and methods may be operative to use available measurement data to add further layer of information redundancy with a view to predictive aspects and with the ultimate aim of avoiding trips of, e.g., a grounding switch. The predictive system 90 may be used to determine predictive actions supported by data-driven techniques (which may be supervised or unsupervised or semi-supervised based on the use-case of monitoring).

[0229] The methods and systems disclosed herein provide the predictive system 90 for predictive maintenance purposes, which may use data-driven processing techniques. The provision of the predictive system 90 in addition to the first and second control and protection subsystems also prevents the predictive system 90 from interfering with the control and protection subsystem operation, as the control and protection subsystems may continue to operate based on their legacy logic.

[0230] Aspects that may be used alone or in any combination in the methods and systems disclosed herein include the following: • Retrieval of configuration data: HVDC stations of varying types and configurations exist. Thus, the predictive system 90 fetches the system configuration from the existing control and protection subsystems that may be engineered to meet power transmissions needs.

[0231] • Data selection: Data useful for predictive asset health assessment may be channeled into a data aggregation store in the storage system 72. The useful data may be selected based on whether it allows predictive asset health assessment to be performed.

[0232] • Processing the electric power system information: Data driven techniques may employ Al techniques, such as trained machine learning (ML) models to perform predictive asset health assessment. Examples of such techniques comprise any one or any combination of the following: o RNN-based methods like LSTM or LSTM with attention mechanism or LSTM with transformer or LSTM with gated-recurrent unit (GRU) etc. to determine the RUL of an asset or its components (with the latter then being processed to the RUL of the asset). Exemplary techniques include any one, any combination, or all of: i. Performance degradation prediction using at least one LSTM cell. ii. Predicting a degradation trend using at least one LSTM cell. ill. Dual LSTM for change detections iv. Predicting the RUL using LSTM. o SVM, regression techniques, k-nearest neighbor, cluster techniques, and / or other techniques that allow the operation data associated with an asset to be processed into a RUL, end of life, or other quantity indicative of a predictive asset health assessment. o Deep learning techniques can be used to determine performance issues, which in turn can be mapped to a RUL, end of life, or other quantity indicative of a predictive asset health assessment. o Other ensemble techniques can be used that include two or more of the above- mentioned techniques or other data-driven approaches.

[0233] • The models for data-driven processing may operative according to different strategies like online models (that process live data feed) or offline models (that work on batch data) or a combination thereof.

[0234] • Operating states may be derived from raw data that is collected and processed by the models for data-driven processing.

[0235] • The output of the models used to perform the predictive asset health assessment may be stored into an output database for history which is in turn used to create a knowledge base (KB) that may include knowledge for each component / asset like a deterioration rate, failure rate, etc.

[0236] • Such a knowledge base can serve as an input for a next cycle of the failure modes, effects and criticality analysis (FMECA). The predictive system 70 may be operative to access the knowledge base to perform such a FMECA analysis based thereon. The predictive system 70 may be operative to update the FMECA parameters. To this end, the predictive system 70 may retrieve FMECA parameters from a knowledge base, update the FMECA parameters based on the processing performed by the predictive system 70, and store the updated FMECA parameters in the knowledge base.

[0237] • The techniques disclosed herein may be used to notify an expert user. In this case, the predictive system 70 may be operative to control an HMI to enable the expert user to provide input for later use by the predictive system 70.

[0238] • The SCADA system 40, e.g., the predictive system 70 of the SCADA system 40, may be operative such that the result of the predictive asset health assessment is used for implementing maintenance activities with a timing dependent on the predictive asset health assessment. For illustration, the SCADA system 40, e.g., the predictive system 70 of the SCADA system 40, may be operative to generate a maintenance graph for each station.

[0239] • The techniques disclosed herein allow maintenance to be performed in accordance with the predictive asset health assessment, i.e., in a predictive need-based approach.

[0240] • The SCADA system 40, e.g., the predictive system 70 of the SCADA system 40, may be operative such that the result of the predictive asset health assessment is used for providing information related to the predictive asset health assessment or maintenance activity based thereon on a dedicated region (e.g., a dedicated maintenance page) of an HMI screen that the maintenance personal at site can see and track use.

[0241] • The SCADA system 40, e.g., the predictive system 70 of the SCADA system 40, may be operative to use the results of the predictive asset health assessment to verify alarms and events from the control and protection subsystem operation.

[0242] • The output of the the predictive system 70 may be used to perform architecture and engineering tasks based on the predictive asset health assessment. Thus, according to further aspects of the invention, there is provided the use of results of the predictive asset health assessment for performing engineering tasks for new electric power system installations.

[0243] Various effects and advantages are attained by the methods and systems disclosed herein. The operation of the predictive system does not adversely affect the operation of the control and protection subsystems, thus affording integration of the predictive system into the SCADA system in a safe manner. The predictive system has good monitoring capability as it can use the data available in the first and second control and protection systems. With the predictive system being provided as a dedicated system for maintenance-related aspects, it is possible to implement it as a highly available system.

[0244] As previously noted, integration of the predictive system into the SCADA system also reduces the complexity of processing off premises and addresses safety concerns of infrastructure operators. The predictive system can be operative to handle online data, offline data, or a combination of both. Thus, it can provide a data aggregation system to store the output and metric from data-driven processing. HMI interface control can be implemented to enable inputting of additional information for subsequent use by the predictive system. Notifications may be provided based on a result of the predictive asset health assessment via, e.g., a SCADA system HMI with a dedicated section for maintenance related activities.

[0245] By retrieving configuration data in an ongoing manner, the predictive system can be kept in synchronization with the configurations of the first and second control and protection systems.

[0246] The predictive system can also perform control operations that cause feedback to be provided to personnel in a station monitoring and / or maintenance team.

[0247] The systems and methods disclosed herein thus provide improved techniques useful in association with maintenance of electric power systems.

[0248] While embodiments have been described in detail with reference to the drawings, various modifications may be implemented in other embodiments. For illustration rather than limitation:

[0249] • While embodiments have been described in which the SCADA system is a SCADA system for an HVDC electric power system, the techniques can also be applied to a SCADA system for another electric power system, a legacy AC power generation, transmission, and / or distribution system, for a microgrid or for a distributed energy resource (DER), provided that it implements the redundancy leveraged by the predictive system as discussed herein.

[0250] • While embodiments have been described in which a control and protection system is operatively associated with primary system assets such as converter / inverter assets and / or other primary system equipment such as valves, CBs, etc., the techniques disclosed herein are applicable when the SCADA system is operative to perform control and protection functions to other equipment, such as, without limitation: o a heating, ventilation, air conditioning (HVAC) system; o another auxiliary system of a station; o motors, pumps, and / or other components of primary system equipment or auxiliary systems. • While embodiments have been described in which the SCADA system performs control and / or monitoring function for primary system assets or other primary system equipment, the SCADA system may alternatively or additionally be operative to perform monitoring and / or control functions for secondary system devices (e.g., for central processing unit (CPU) usage, communication bandwidth usage, etc. of devices comprised by the SCADA system).

[0251] • While embodiments have been described in which a converter / inverter comprises thyristors or IGBTs, the converter / inverter may have other configurations, in particular configurations comprising controllable power semiconductor devices.

[0252] • While embodiments have been described in which the predictive system 70 itself may be operative to update the data-driven processing logic based on observations, the determination of the updated parameters may also be performed by a separate computing system (e.g., by a server system that performs Al model training).

[0253] Embodiments may be used in association with a power grid comprising a HVDC power transmission system, without being limited thereto.

[0254] This description and the accompanying drawings that illustrate aspects and embodiments of the present invention should not be taken as limiting-the claims defining the protected invention. In other words, while the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative and not restrictive. Various mechanical, compositional, structural, electrical, and operational changes may be made without departing from the spirit and scope of this description and the claims. In some instances, well- known circuits, structures, and techniques have not been shown in detail in order not to obscure the invention. Thus, it will be understood that changes and modifications may be made by those of ordinary skill within the scope and spirit of the following claims. In particular, the present invention covers further embodiments with any combination of features from different embodiments described above and below.

[0255] The disclosure also covers all further features shown in the Figures individually although they may not have been described in the afore or following description. Also, single alternatives of the embodiments described in the Figures and the description and single alternatives of features thereof can be disclaimed from the subject matter of the invention or from disclosed subject matter. The disclosure comprises subject matter consisting of the features defined in the claims or the embodiments as well as subject matter comprising said features.

[0256] The term "comprising" does not exclude other elements or process blocks, and the indefinite article "a" or "an" does not exclude a plurality. A single unit or process block may fulfil the functions of several features recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Components described as coupled or connected may be electrically or mechanically directly coupled, or they may be indirectly coupled via one or more intermediate components. Any reference signs in the claims should not be construed as limiting the scope. A machine-readable instruction code may be stored / distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via a wide area network or other wired or wireless telecommunication systems. Furthermore, a machine-readable instruction code can also be a data structure product or a signal for embodying a specific method such as the method according to embodiments.

Claims

CLAIMS1. A method of operating an electric power system using a Supervisory Data Acquisition and Control, SCADA, system, the method comprising: monitoring, by the SCADA system, the electric power system, wherein the SCADA system comprises a first control and protection subsystem and a second control and protection subsystem, wherein the first control and protection subsystem and the second control and protection subsystem implement control and protection function redundancy; performing, by the first control and protection subsystem, control and protection functions for a first electric power subsystem of the electric power system and performing, by the second control and protection subsystem, control and protection functions for a second electric power subsystem of the electric power system, wherein the second electric power subsystem is a backup for the first electric power subsystem; communicatively interfacing, by a predictive system, with the first control and protection system and the second control and protection system to obtain electric power system information for performing an asset health assessment; using, by the predictive system, the obtained electric power system information to perform a predictive asset health assessment for at least one asset of the electric power system; and generating, by the predictive system, output based on the predictive asset health assessment.

2. The method of claim 1, wherein the obtained electric power system information comprises first configuration data obtained from the first control and protection subsystem and second configuration data obtained from the second control and protection subsystem, wherein the predictive system uses the first configuration data and the second configuration data to perform the predictive asset health assessment and generate the output.

3. The method of claim 2, wherein the predictive asset health assessment comprises discriminating, by the predictive system, an asset degradation from a hardware change, wherein the predictive system identifies the hardware change based on the first configuration data and the second configuration data.

4. The method of any one of the preceding claims, wherein the obtained electric power system information comprises operation data of the electric power system, wherein the predictive system uses the operation data to perform the predictive asset heath assessment and generate the output.

5. The method of claim 4, wherein the operation data comprises first sensor output of sensors of the first control and protection subsystem and second sensor output of sensors of the second control and protection system, wherein the predictive system is operative to perform the predictive asset health assessment based on an inconsistency between the first sensor output and the second sensor output.

6. The method of claim 4 or claim 5, wherein the predictive system is operative to store a subset of the operation data in a storage system of the predictive system or accessible to the predictive system, wherein the predictive asset health assessment is performed based on the stored subset of the operation data.

7. The method of claim 6, further comprising the following operations performed by the predictive system: selecting the subset of the operation data for storing in the storage system based on a relevance for performing predictive asset health assessment and storing the selected subset in the storage system.

8. The method of any one of the preceding claims, wherein the predictive system performs the predictive asset health assessment for an asset comprising by a plurality of components, wherein determining the predictive asset health assessment for the asset comprises establishing a time-dependent degradation of several components of the plurality of components and establishing a time-dependent degradation of the asset based on the timedependent degradation of the several components.

9. The method of any one of the preceding claims, wherein performing the predictive asset health assessment and / or generating the output comprises using an artificial intelligence, Al, model, wherein the Al model has an input operative to receive Al model input that is based on at least part of the obtained data and an Al model output, wherein the predictive system is operative to generate the output based on the Al model output.

10. The method of claim 9, wherein at least one of the following applies:the Al model is operative to process timeseries data; the Al model comprises long short term memory, LSTM, cells and / or gated recurrent units, GRUs, to process timeseries data; the Al model comprises at least one attention mechanism; the Al model comprises at least one self-attention mechanism; the Al model comprises a stack of self-attention mechanisms; the Al model comprises an Al-transformer model; the Al model comprises one or several autoencoders and / or a convolutional neural network, CNN.

11. The method of any one of the preceding claims, wherein generating the output comprises controlling, by the predictive system or by at least one of the first and second protection and control systems, at least one interface to provide the output.

12. The method of claim 11, wherein the at least one interface comprises a human machine interface, HMI.

13. The method of claim 11 or claim 12, wherein the output comprises control commands that cause a transition into a state for performing maintenance.

14. The method of claim 13, wherein the control commands comprise control commands that cause the first electric power subsystem to transition to a state in which maintenance can be performed on the first control and protection subsystem and that cause the second electric power subsystem to transition to an energy-carrying state.

15. The method of any one of the preceding claims, wherein the SCADA system is or comprises a high voltage direct current, HVDC, system SCADA system.

16. The method of claim 15, wherein the SCADA system comprises control and protection devices for HVDC power system assets, wherein the HVDC power system assets comprise one, several, or all of: an inverter / converter asset; a cooling system for an inverter / converter asset; a transformer comprising three single-phase transformers operated under control of the control and protection devices;a three-phase transformer; a heating, ventilation, air conditioning, HVAC, system; other auxiliary systems of stations; motors, pumps, and / or other components of primary system equipment or auxiliary systems.

17. The method of any one of the preceding claims, wherein the SCADA system is operative to monitor primary system devices, optionally wherein the SCADA system is operative to monitor hardware usage of secondary devices that perform control and protection functions.

18. The method of any one of the preceding claims, wherein the electric power system is or comprises a high voltage direct current, HVDC, system.

19. A predictive system for use with a first control and protection subsystem and a second control and protection subsystem of a Supervisory Control and Data Acquisition system, wherein the first control and protection subsystem and the second control and protection subsystem implement control and protection function redundancy, the first control and protection subsystem being operative to perform control and protection functions for a first electric power subsystem of an electric power system, the second control and protection subsystem being operative to perform control and protection functions for a second electric power subsystem of the electric power system, wherein the second electric power subsystem is a backup for the first electric power subsystem, wherein the predictive system comprises: at least one interface operative to communicatively interface the predictive system with the first control and protection system and the second control and protection system to obtain electric power system information for performing an asset health assessment, and at least one processing circuit operative to use the obtained electric power system information to perform a predictive asset health assessment for at least one asset of the electric power system; and generate output based on the predictive asset health assessment.

20. A Supervisory Control and Data Acquisition, SCADA, system for an electric power system, the SCADA system comprising: a first control and protection subsystem and a second control and protection subsystem, wherein the first control and protection subsystem and the second control and protection subsystem implement control and protection function redundancy, the first controland protection subsystem being operative to perform control and protection functions for a first electric power subsystem of an electric power system, the second control and protection subsystem being operative to perform control and protection functions for a second electric power subsystem of the electric power system, wherein the second electric power subsystem is a backup for the first electric power subsystem, and the predictive system of claim 19.

21. The SCADA system of claim 19, wherein the SCADA system is operative to perform the method of any one of claims 1 to 18.

22. An electric power system, comprising primary power system equipment; the SCADA system of claim 19 or claim 20, wherein the SCADA system is operative to perform control and protection functions for the primary power system equipment.

23. Machine-readable instruction code comprising machine-readable instructions which, when executed by at least one processing circuit, cause the at least one processing circuit to perform the method of any one of claims 1 to 18.

24. Non-transitory storage medium having stored thereon machine-readable instruction code comprising machine-readable instructions which, when executed by at least one processing circuit, cause the at least one processing circuit to perform the method of any one of claims 1 to 18.

Images