Remote restart of wind and solar power generators

Abstract

A method for controlling a remote restart function of a plurality of solar and / or wind power generators, hereinafter solar / wind power generator, the method comprising: (i) receiving, from each of a plurality of power systems, each power system comprising at least one of the plurality of solar / wind power generators and each power system being separately connected to an electrical grid, a value of one or more variables measured on the site of each of the plurality of solar / wind power generators; (ii) performing a first function to determine whether an abnormal solar / wind power generator is ready to have a normal operation remotely restarted based on the received value of the one or more variables; (iii) performing a second function to determine a restart order according to which a group of abnormal solar / wind power generators should have a normal operation remotely restarted, among the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted according to the first function; and (iv) outputting an indication of the determined restart order. A computer program, device and system are also provided.

Description

[0001] REMOTE RESTART OF WIND AND SOLAR POWER GENERATORS

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This Patent Application claims priority from Italian Patent Application No. 102023000015774 filed on July 26, 2023, the entire disclosure of which is incorporated herein by reference.

[0004] FIELD OF THE INVENTION

[0005] Example aspects herein relate to the wind and solar electrical power generation, and more specifically to a method for controlling a remote restart function of solar / wind power generators, a computer program, a device and a system.

[0006] BACKGROUND

[0007] Electrical energy demand is met increasingly using renewable energy sources such as wind or solar energy, for which wind turbines and solar panels are installed at locations optimizing the transformation of wind power or solar power into electrical power. Wind or solar power generators that are located in the vicinity of each other are grouped into power systems (also called wind or photovoltaic power station, or even "wind farm" and "solar farm"). The electrical power generated in each power system is transferred to the electrical grid.

[0008] Wind and / or solar power generators (hereinafter wind / solar power generators) may be put out-of-service or may be controlled to operate at a reduced capacity, either due to scheduled maintenance, a failure or breakdown of a component, etc. It is crucial to bring the wind / solar power generator back in service or to improve its operational capacity as soon as possible, thus reducing the so-called "down-time", to improve the efficiency of the power system as a whole.

[0009] While the presence of a technician may be required to repair some of the wind / solar power generator or replace components, in some situations the wind / solar power generator may be ready to be put back in service without requiring a technician to be in place. A remote restart function may be provided to transmit a command to the wind / solar power generator to restart or increase capacity, which avoids unnecessary time and resource allocations that would otherwise be required bring technicians to each wind / solar power generator.

[0010] However, the determination that a wind / solar power generator is ready to be put back in service or to increase electrical power production may be based on a vast number of pieces of data related to the wind / solar power generator, making this a difficult decision.

[0011] Additionally, when controlling multiple power systems, each of which typically comprises tens to hundreds of wind / solar power generator, there may be a large number of wind / solar power generators which can be remotely restarted. Whilst the restart command may be transmitted remotely (e.g. automatically from a computer), the power generator may need to be monitored to ensure the restart is successful, thus restarting power generators can require a large amount of time and processing resources.

[0012] For example, a typical operator may be faced with more than a hundred of out-of- service wind / solar power generators, and a comparable number of generators that operate at reduced capacity, requiring the analysis of hundreds of variables to identify which wind / solar power generator should be restarted.

[0013] There is therefore a need for an improved process for returning wind / solar power generators to a normal operation.

[0014] Summary

[0015] According to a first aspect of the present invention, there is provided a method for controlling a remote restart function of a plurality of solar and / or wind power generators, hereinafter solar / wind power generator, the method comprising:

[0016] - receiving, from each of a plurality of power systems, each power system comprising at least one of the plurality of solar / wind power generators and each power system being separately connected to an electrical grid, a value of one or more variables measured on the site of each of the plurality of solar / wind power generators;

[0017] - performing a first function to determine whether an abnormal solar / wind power generator is ready to have a normal operation remotely restarted based on the received value of the one or more variables,

[0018] - performing a second function to determine a restart order according to which a group of abnormal solar / wind power generators should have a normal operation remotely restarted, among the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted according to the first function; and

[0019] - outputting an indication of the determined restart order.

[0020] Accordingly, abnormal wind / solar power generators may be returned to normal operation whilst reducing required human resources.

[0021] Preferably, the one or more variables measured on site includes at least one of: a temperature of one or more components of the solar / wind power generator, an electrical quantity measured in one or more electrical components of the solar / wind power generator, a mechanical quantity measured for the power generator, a meteorological quantity measured in an environment of the solar / wind power generator.

[0022] Preferably, the meteorological quantity comprises at least one of a wind speed, a wind direction, a wind distribution, and an amount and / or angle of received sunlight.

[0023] Preferably, the electrical quantity comprises a value of a current present in an electrical component of the solar / wind power generator.

[0024] Preferably, the one or more variables measured on site includes at least one of: the temperature of one or more components of the solar / wind power generator, the wind speed, the wind direction, and the current in electrical components of the solar / wind power generator.

[0025] Preferably, the one or more components comprise at least one of a mechanical component, a hydraulic component, an electrical component and an electronic component.

[0026] Accordingly, the return of abnormal wind / solar power generators to normal operation may be optimised by taking into account the state or environment of each solar / wind power generator.

[0027] Preferably, the first function is performed using a deterministic algorithm and / or an Al model receiving, as inputs, the plurality of variables and a history of remote restarts performed in a predetermined period of time for solar / wind power generators.

[0028] Preferably, the restart order is determined based on at least one of a forecast of energy production by means of the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted, and one or more optimization criteria related to the grid and / or the solar-wind power generator. Preferably, the second function is performed using a deterministic algorithm and / or a trained Al model which:

[0029] - elaborate as inputs (i) the forecast of energy production, and (ii) the one or more optimization criteria related to the grid and / or the solar-wind power generator, and

[0030] - provide as output the restart order.

[0031] Preferably, a value of a plurality of types of variables measured on the site of the solar / wind power generators is received, wherein, optionally, at least one type of variable indicates a respective abnormality of a solar / wind power generator, wherein the first function is performed using a first Al model, wherein the first Al model receives, as input, the value of each type of variables, and wherein the first Al model is trained to identify a first subset of the types of variables to be prioritised for determining whether an abnormal solar / wind power generator is ready to have a normal operation remotely restarted.

[0032] Preferably, the types of variables measured on the site of the solar / wind power generator comprise one or more type of internal variables related to an internal operation of the solar / wind power generator and one or more type of external variables related to an environment of the solar / wind power generator, wherein the method further comprises receiving one or more triggers, each trigger being for causing an alert to be raised, the alert indicating a potential abnormality related to the power generator or a safety risk related to an operation of the power generator, wherein the one or more triggers comprise at least one internal trigger generated based on an internal variable, the first Al model receives, as input, a value of each type of variables, and the first function is performed using a first deterministic algorithm, the first deterministic algorithm receiving, as input, the at least one internal trigger only.

[0033] Preferably, the first deterministic algorithm is based on one or more first predetermined correlations, each first predetermined correlation being between one or more internal trigger and a determination that an abnormal solar / wind power generator is ready to have a normal operation remotely restarted, and the first Al model is trained to identify a new correlation between one or more type of variables and a determination that an abnormal solar / wind power generator is ready to have a normal operation remotely restarted.

[0034] Preferably, a value of a plurality of types of variables measured on the site of the solar / wind power generator is received, wherein, optionally, at least one type of variable indicates a respective abnormality of a solar / wind power generator, wherein the second function is performed using a second Al model, wherein the second Al model receives, as input, the value of each type of variables, and wherein the second Al model is trained to identify a subset of the types of variables to be prioritised for determining the restart order

[0035] Preferably, the types of variables measured on the site of the solar / wind power generator comprise one or more type of internal variables related to an internal operation of the solar / wind power generator and one or more type of external variables related to an environment of the solar / wind power generator, wherein the method further comprises receiving one or more triggers, each trigger being for causing an alert to be raised, the alert indicating a potential abnormality related to the power generator or a safety risk related to an operation of the power generator, wherein the one or more triggers comprise at least one internal trigger generated based on an internal variable, the second Al model receives, as input, a value of each type of variables, and the second function is performed using a second deterministic algorithm, the second deterministic algorithm receiving, as input, the one or more triggers only.

[0036] Preferably, the second deterministic algorithm is based on one or more second predetermined correlations, each second predetermined correlation being between one or more triggers and a prioritisation of one abnormal solar / wind power generator ready to have a normal operation remotely restarted over another, and the second Al model is trained to identify a new correlation between one or more type of variables and a prioritisation of one abnormal solar / wind power generator ready to have a normal operation remotely restarted over another.

[0037] Preferably, outputting the indication of the determined restart order comprises at least one of notifying an operator of the determined restart order, and automatically transmitting a restart command to the solar / wind power generators in accordance with the determined restart order.

[0038] Preferably, the first function is performed upon detection of a predetermined event, the predetermined event including at least one of a predefined duration elapsing and a solar / wind power generator interrupting an operation due to an alert. Preferably, the second function is performed every time an abnormal solar / wind power generator is determined to be ready to have a normal operation remotely restarted by means of the first function.

[0039] Preferably, each abnormal solar / wind power generator is out-of-service or operates at a reduced capacity.

[0040] Preferably, the method comprises determining whether at least one solar / wind power generator is out-of-service or operates at reduced capacity.

[0041] Preferably, the determining of whether the at least one solar / wind power generator is out-of-service or operates at reduced capacity may be based on the received value of one or more variables measured on the site of the solar / wind power generator.

[0042] Preferably, the determining of whether the at least one solar / wind power generator is out-of-service or operates at reduced capacity may be performed periodically.

[0043] Preferably, the first function is performed when it is determined that at least one solar / wind power generator is out-of-service or operates at reduced capacity.

[0044] Preferably, the method comprises determining whether a group of solar / wind power generators is ready to be remotely restarted (i.e. to have a normal operation restarted remotely).

[0045] Preferably, the group comprise at least a predetermined number. For example, the predetermined number may be 5, more preferably at least 10, at least 25, at least 50 power generators.

[0046] Accordingly the second function would be performed only when more than a predetermined number of power generators need to be remotely restarted. Thus, a human operator may determine the restart order when few power generators are to be restarted (i.e. when no assistance is required).

[0047] Preferably, the determining whether the group of solar / wind power generators is ready to be remotely restarted is performed each time the first function determines that a solar / wind power generator is ready to be remotely restarted.

[0048] Preferably, the value of one or more variables is received repeatedly.

[0049] Preferably, the value of one or more variables is received periodically, wherein each value may have a corresponding predefined period. According to a second aspect of the present invention, there is provided a computer program comprising instructions which, when executed by one or more processors, cause the one or more processors to execute the method according to the first aspect.

[0050] Preferably, the computer program is stored in a computer-readable storage medium.

[0051] According to a third aspect of the present invention, there is provided a device comprising means for performing the method according to the first aspect.

[0052] Preferably, the device comprises an interface module and at least one processor, the interface module configured to receive the value of the one or more variables, and to output the indication of the determined order, and the processor being configured to process the received value of the one or more variables to perform the first function and the second function.

[0053] According to a fourth aspect of the present invention, there is provided a system comprising the device according to the third aspect, and a plurality of power systems, each power system comprising at least one of the plurality of solar / wind power generators and each power system being separately connected to an electrical grid.

[0054] LIST OF FIGURES

[0055] Embodiments of the present invention, which are presented for better understanding the inventive concepts, but which are not to be seen as limiting the invention, will now be described with reference to the figures in which:

[0056] Figure 1 shows a schematic diagram of a system connected to an electrical grid, in accordance with example embodiments;

[0057] Figure 2 shows a schematic diagram of a device in accordance with example embodiments;

[0058] Figure 3 shows a process performed by the device in example embodiments;

[0059] Figure 4 shows a schematic diagram of the first function performed by the device, in accordance with example embodiments;

[0060] Figure 5 shows a schematic diagram of the second function performed by the device, in accordance with example embodiments;

[0061] Figure 6 shows a process for training an Al model, in example embodiments; and

[0062] Figure 7 shows a process performed for controlling a remote restart function of a plurality of solar and / or wind power generators, in example embodiments. DETAILED DESCRIPTION

[0063] Although example embodiments will be described below, it will be evident that various modifications may be made to these example embodiments without departing from the broader spirit and scope of the invention. Accordingly, the following description and the accompanying drawings are to be regarded as illustrative rather than restrictive.

[0064] Components described herein such as the device, the power system controllers, the electrical grid controller, etc. may use any suitable communication link to exchange data, such as wireless communication link (for example a Wi-Fi, cellular telephone data link such as LTE / 5G, Bluetooth or Bluetooth Low Energy (BLE) ), a wired communication link (e.g. DSL, fibreoptic cable, Ethernet etc.). Each communication link may not be permanent.

[0065] In the following description and in the accompanying figures, numerous details are set forth in order to provide an understanding of various example embodiments. However, it will be evident to those skilled in the art that embodiments may be practiced without these details.

[0066] For example, whilst the figures and following description may omit certain connections between components, the person skilled in the art would understand that such connection may be present, even if not represented or described explicitly.

[0067] The operation of a wind / solar power generator may be affected by various issues, such as damage due e.g. to wear, or the failure of one or more components. This may cause the wind / solar power generator to operate at reduced capacity (i.e. that the electrical power generated by the power generator is less than a normal rated power), due for example to reducing a speed at which a turbine rotates, or to using only a subset of photovoltaic panels to generate electrical energy. In the following, such wind / solar power generator that is out- of-service or that operates at a reduced capacity will be referred to as an abnormal power generator, for brevity.

[0068] In the case of an out-of-service power generator, a restart command may be a message causing the power generator to restart and begin a normal operation, and in the case of a power generator operating at reduced capacity, a restart command may be a message causing the power generator to increase capacity to a normal level (i.e. restart normal operation).

[0069] An abnormal wind / solar power generator may restart a normal operation following an automatic restart (i.e. the power generator initiates the restart without receiving any external command), a remote restart (i.e. upon reception of a restart command), or a local restart, requiring the presence of a human operator to restart. The automatic restart is the preferred, as it requires the least amount of resources or processing, followed by the remote restart, and then the local restart.

[0070] Figure 1 shows a schematic diagram of a system connected to an electrical grid, in accordance with example embodiments.

[0071] The electrical grid 10 connects the power system 30-1, 30-2, 30-3 and 30-4 to intended destinations of electrical power, i.e. loads, such as end users or electric substations connecting different electric networks (not shown).

[0072] Each power system 30-1, 30-2, 30-3 and 30-4 is connected to the electrical grid 10 via a separate connection, 12-1, 12-2, 12-3 and 12-4, respectively. Although four power systems are shown on Figure 1, it would be understood that any number of power systems may be connected to the electrical grid 10 (e.g. two, three, five or more).

[0073] Each power system has one or more power generator 32. Each power generator may be a solar power generator (e.g. converting solar energy into electrical energy via one or more photovoltaic panels or surface) or a wind power generator (e.g. converting wind energy into electrical energy via one or more wind turbine). For brevity, these will be referred to as power generator 32 in the following. The power generators 32 in a power system are connected to a power system controller 34, which transmits electrical power from the power generators 32 to the electrical grid 10. The power generators 32 may have series connections to the power system controller 34, as shown in power systems 30-1 and 30-2, parallel connections to the power system controller 34, as shown in power systems 30-3 and 30-4, or any combinations thereof. Each power system 30 may have any number of power generator, as the number shown on Figure 1 is purely illustrative and not limiting.

[0074] The electrical grid 10 comprises a centralised controller 11, which performs monitoring functions such as a supervisory control and data acquisition system (SCADA) function for the power systems, for which it may collect variables measured on the site of each power generator 32, such as one or more of: a temperature of one or more components of the power generator 32, and in particular one or more of: o a temperature of mechanical component(s) such as a nacelle, gearbox, bearing, frame of a photovoltaic panel, etc. o a temperature of hydraulic component(s) such as a fluid, pump, a cooling system, etc. o a temperature of electrical or electronic component(s) such as a generator, a power converter, a transformer, etc. an electrical quantity measured in one or more electrical components of the power generator, such as a current, voltage, complex impedance, resistance, capacitance, impedance, active power, reactive power, load factor, etc. a mechanical quantity measured for the power generator, such as a pressure, a torque, etc. a meteorological quantity measured in an environment of the power generator 32, and that influences the amount of electrical energy that can be produced by the power generator 32, and in particular one or more of: o a distribution, speed, and / or direction of wind at a location near the power generator 32, a turbulence intensity, o a quantity indicating an amount of sunlight received by a photovoltaic panel, such as a solar irradiance, or an angle of received sunlight, o an average air density in the vicinity of the power generator, and o an ambient temperature in the environment of the power generator 32.

[0075] The value of the one or more variables for each power generator 32 in the power system may be measured, for example, using sensors placed in and around each power generator 32. In each power system, the power system controller 34 may obtain the value from each power generator 23 and transmit the collected values to the centralised controller 11 using a communication link (not shown). The centralised controller 11 then forward the values for each power generator to the device 20. The centralised controller 11 may comprise or be coupled to memory for storing information, such as the value of the one or more variables for each power generator 32.

[0076] As such, it would be understood that the measured variables may be of various types, and may be categorised into internal variables (if they relate to an internal operation of the power generator 32, such as the temperature of component(s), the electrical quantity, or the mechanical quantity) or an external variable (if they relate to an environment of the power generator, such as the meteorological quantity).

[0077] In addition, the centralised controller 11 generates, as part of the SCADA function, triggers (e.g. trigger signals, or flags), based on the variables measured on the site of each power generator. Each trigger is used to indicate a state of each power generator, and may be activated for causing an alert to be raised, indicating a potential abnormality related to the power generator. For example, a set of triggers may be generated for the power generator (or a number of components of the power generator) based on a measured temperature, electrical quantity, mechanical quantity, meteorological quantity, or any combination thereof.

[0078] For each trigger, a predetermined range may be defined, and if a measured value of the corresponding variable is outside the predetermined range, the trigger signal may be activated. For example a temperature trigger signal may be activated if a temperature of a component or of the power generator is higher than a predetermined threshold.

[0079] For example, the predetermined range may be defined using a threshold value (e.g. a range that is less than the threshold value, equal to greater than the threshold value, etc) or two threshold values (e.g. greater than a lower threshold value and less than or equal to a higher threshold value, etc.).

[0080] Triggers that are based on internal variables (e.g. temperature of components, electrical quantity and / or mechanical quantity) may be defined as internal triggers, and triggers based on variables relating to the environment of the power generator may be defined as external triggers.

[0081] The centralised controller 11 also operates a safety system and may, as part of the safety system, generate one or more safety triggers.

[0082] For example, the safety system may monitor the location of human operators (e.g. technicians or other field personnel) in the vicinity of the power generator 32, e.g. when performing maintenance or repair for the power generator 32. As part of the monitoring system, the centralised controller 11 may store, in memory, information indicating the presence of human operator(s) near a power generator 32. If human operator(s) are present in the vicinity of a power generator, such that an operation of the power generator 32 may be unsafe for the human operator(s), a safety trigger indicating this risk may be activated for that power generator 32. The centralised controller 11 also store in memory, for each power generator: information identifying each power generator 32, such as an identifier, a type, model, manufacturer of the power generator 32, information used to schedule preventative or corrective maintenance, such as an age of the power generator (e.g. since installation), a predetermined amount of resource(s) (e.g. time and / or cost) expected to be needed to perform different tasks of maintenance (e.g. inspection, lubrication, cleaning or adjusting mechanical components, replacement of hydraulic, mechanical, electrical or electronic components, etc.). historical data (e.g. a log) such as o data regarding previous operational state of the power generator (e.g. a timeline indicating periods where the power generator had a normal operational state, periods where the power generator operated with reduced capacity and period where the power generator was out-of-service or offline), o data regarding past restart commands transmitted for that power generator, at least for a predetermined period of time (e.g. the last 3 months). o data regarding past maintenance and / or abnormalities detected in that power generator, data regarding future scheduled maintenance or downtime, forecast information for energy production, based for example on measured and / or predicted meteorological information in the environment of each power generator, and a state of each trigger (internal triggers, external triggers and safety triggers).

[0083] For each power generator 32, the device 20 obtains from the centralised controller 11 the information identifying the power generator in association with: the value of the one or more variables, the set of one or more activated triggers (internal triggers, external triggers and safety triggers), the information used to schedule preventative or corrective maintenance, the historical data, the forecast information for energy production, and optimization information, namely information used to derive one or more optimization criteria related to the electrical grid 10 and / or the power generator 32, which may be based on constraints set for the electrical grid or overhauls planned for the power generator 32.

[0084] The information used to derive the one or more optimization criteria include, for example, information on a node connecting the power generator to the electrical grid 11 (e.g. a node on or connecting the power system to the electrical grid), and / or information on a power generation capacity of the power generator (e.g. a value in MWh). For example, the information on the connecting node may be used to determine whether there are grid constraints set for the node, such as an amount of electrical power that may flow through one or more branches connected to the connecting node. The restart order may be optimised by prioritising the restart of power generators connected via a node (i.e. a connecting node) that is not subjected to grid constraints. As another example, the restart order may be optimised by prioritising power generators having a higher power generation capacity, thus allowing a higher increase in the generated power with fewer remote restarts.

[0085] In addition, one or more optimization criteria may be derived using the information related to preventative maintenance. For example, a power generator having a scheduled maintenance soon may be given a lower priority on the restart order, thus allowing the scheduled maintenance to take place.

[0086] For example, the centralised controller 11 may transmits the data to the device 20 using a communication link, or the device 20 may retrieve the data directly from the memory in which the centralised controller 11 stores this data.

[0087] For the triggers, the device 20 may obtain only the activated triggers or may obtain a state of each trigger and select only the activated triggers as the set of one or more activated triggers.

[0088] As will be explained in more detail below, the device 20 determines, by performing a first function, whether an abnormal power generator is ready to have a normal operation remotely restarted, and determines, by performing a second function, an order according to which a group of abnormal power generators should have a normal operation remotely restarted (i.e. a restart order). The device 20 then notifies an operator of the determined restart order. The notification may be, for example, transmitted to the centralised controller 11 which causes a display to present the restart order to an operator. Instead of, or in addition to the display, an audio signal indicating the restart order may be provided to the operator (i.e. any audio-visual output may be used). The centralised controller may also store log the restart order in the historical data regarding past restart command transmitted to the power generators.

[0089] Additionally or alternatively, the device 20 may transmit a restart command to the power generators 32 in accordance with the determined order. A restart command may be transmitted via the centralised controller 11 and the corresponding power system controller 34, or directly to the power generator 32. The restart command may also be transmitted by the device 20 directly to the power generator 32 and to the centralised controller 11, for logging. In any case, if a restart command is transmitted, it is done so automatically, i.e. without an operator triggering its transmission.

[0090] Although the device 20 is shown to be separate from the centralised controller 11, this is merely for illustrative purposes, as the device 20 may be in a same location (e.g. a room) as the centralised controller 11. Furthermore, it would be understood to the skilled person that the centralised controller 11 need not be a single device and may comprises processing device(s) (e.g. computers) that process data related to the electrical grid 10, network device(s) controlling the transfer of data between devices, memory device(s), display(s), etc.

[0091] Figure 2 shows a schematic diagram illustrating an exemplary implementation of the device 20.

[0092] In an embodiment, the device 20 includes at least processor 22, a memory 23 and a I / O interface 21. The processor 22 is configured to execute instructions comprised in a computer program stored on the memory 23 to execute any of the functions of the device 20, as described herein.

[0093] The memory 23 is configured to store the computer program. In addition, the memory 23 may also store additional information, which may comprise at least one of:

[0094] 1. information used to perform the first function and / or the second function, such as information defining deterministic algorithm(s) and / or Al model(s) used for these functions, past outputs of the first function and / or the second function, past outputs of the deterministic algorithm(s) and / or the Al model(s), etc. 2. information related to each power generator 32 that is obtained from the centralised controller 11, and

[0095] 3. forecast information for energy production, based for example on measured and / or predicted meteorological information in the environment of each power generator.

[0096] When executing the instructions of the computer program, the processor 22 receives information from other devices by means the I / O interface 21, and output information to the other devices by means of the I / O interface 21.

[0097] It would be understood that the various elements of the device 20 need not be located in a single housing, as the box is purely illustrative. For example, the processor 22 may correspond to a server on located on a rack, the I / O 21 may correspond to a router and / or a switch that is connected to the server, and the memory 23 may correspond to a cluster storage system comprising multiple storage units.

[0098] Figure 3 shows a process performed by the device in example embodiments. For example, the process may be performed by the processor 22.

[0099] At step 302, the device 20 receives a value of one or more variables measured on the site of each of the power generators 32. For example, each power system controller 34 may collect the data related to each power generator 32 in the power system (i.e. the value of one or more variables measured on the site of the power generator), and transmit the collected data to the I / O 21 in the device 20 via any communication link, and the processor 22 then obtains the data from the I / O 21.

[0100] In addition, the device 20 may also receive, at step 302, the information used to schedule preventative or corrective maintenance, the historical data, the forecast information for energy production, and.

[0101] The processor 22 may then cause the memory 23 to store the received data.

[0102] At step 304, the device 20 determines whether one or more power generator 32 is abnormal, i.e. it is out-of-service or operates at a reduced capacity.

[0103] For example, this may be based on a previous decision taken at the centralised controller 11 or at the power system controller 34 to take the power generator 32 out of service or to reduce operational capacity. Alternatively, the determination may be based on an indication transmitted by the power system controller 34 indicating one or more power generators 32 has an abnormal operation (including being out-of-service), or based a value of an electrical quantity measured at the power generator 32 received by the device 20 (for example if the received value is different than an expected value for a normal operation of the power generator).

[0104] The device 20 may perform this determination in sequence for each power generator 32, or the device 20 may perform the determination for any number of power generators in parallel (for example by simultaneously making a determination for each power generator of a power system).

[0105] In addition, the device 20 may store in memory 23, for each power generator 32 determined to be abnormal, information identifying the power generator 32 in association with an indication of an abnormality.

[0106] If no power generator 32 is determined to be abnormal (NO at step 304), the process returns to step 302 to receive new values of the one or more variables.

[0107] If instead, at least one power generator 32 is determined to be abnormal (YES at step 304), the processing proceeds to step 306.

[0108] At step 306, the device 20 performs a first function to determine whether an abnormal power generator 32 is ready to have a normal operation remotely restarted based on the received value of the one or more variables.

[0109] For example, the device 20 may perform the first function for all power generators determined to be abnormal in sequence (e.g. based on an order of the identifiers of each power generator), or the device 20 may perform the first function for multiple (or all) abnormal power generator in parallel. The device 20 may identify the abnormal power generators by retrieving from the memory 23 the identifiers associated with an indication of an abnormality, if stored as part of the processing of step 304.

[0110] For each abnormal power generator that is determined to be ready to have a normal operation remotely restarted, the device 20 may also store in memory 23, information identifying the power generator 32 in association with an indication that the power generator may be remotely restarted. In addition, the device 20 may transmit the indication to the centralised controller 11, for each power generator.

[0111] At step 308, the device 20 determines whether a group of abnormal power generators 32 is ready to have a normal operation remotely restarted. Any number of two or more power generators 32 may define a group, for which a restart order would have to be determined. The power generators 32 in the group may be located in the same power system or in different power systems.

[0112] For example, the device 20 may determine whether the identifiers of two or more power generators is associated with an indication that the power generator may be remotely restarted, if this is stored as part of the processing of step 306.

[0113] However, it would be understood that in some embodiments, the group may have a minimum number of power generators, for example 3, 5, 10, 15, 20, 25, or more.

[0114] If no group of power generators is ready to be remotely restarted (NO at step 308), the process returns to step 302.

[0115] On the other hand, if a group of power generators is ready to be remotely restarted (YES at step 308), processing proceeds to step 310.

[0116] At step 310, the device 20 performs a second function to determine a restart order according to which the group of abnormal power generators should have a normal operation remotely restarted, among the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted according to the first function.

[0117] Then, at step 312, the device 20 outputs an indication of the restart order by notifying an operator of the centralised controller 11. For example, the notification may cause a display screen to list the power generators in the determined restart order, so the operator can proceed to remotely restart the power generators in the indicated order.

[0118] Alternatively or additionally, the device 20 may transmit to the power generators 32 a restart command, following the determined restart order, causing each power generator to begin operation (if it is out of service) or to increase capacity to return to a normal level.

[0119] If the monitoring of each restart by the human operator is required, the device 20 may delay the second and each subsequent restart commands, for example by delaying the transmission of each restart command for a predetermined time duration (e.g. 20-30 minutes) during which the human operator may monitor the power generator that is restarting normal operation. Alternatively, each restart command may indicate a time duration that should elapse between the reception of the restart command and the power generator's initiation to return to normal operation. By way of non-limiting example, and referring to Figure 1, the device 20 may determine that first a power generator in the power system 30-1 should be restarted (i.e. caused to return to normal operation), second a power generator in the power system 30-2, then third another power generator in the power system 30-2, etc.

[0120] However, it should be understood that the restart order may indicate more than one (e.g. two or three) power generators should be remotely restarted simultaneously.

[0121] The processing then returns to step 302.

[0122] Figure 4 shows a schematic diagram of the first function performed by the device, in accordance with example embodiments.

[0123] As shown on Figure 4, the first function 40 comprises one or more first deterministic algorithms 41, one or more first Al models 42, and a combining module 43 that combines an output of the first deterministic algorithm(s) 41 and an output of the one or more first Al model(s) 42. The output of the combining module 43, which is the output of the first function 40, is a determination 65 of whether an abnormal power generator is ready to have a normal operation remotely restarted.

[0124] The variables 61, the triggers 62, the historical data 63 and data regarding the scheduled maintenance and downtime may be provided to the first function 40 for each power generator that is known to be abnormal.

[0125] The first deterministic algorithm(s) 41 receives, as inputs, the safety trigger 62a, the internal triggers 62b, and the historical data regarding past restart commands transmitted for that power generator, to generate a binary decision of whether the power generator can be transmitted a command to have a normal operation remotely restarted.

[0126] For example, a first deterministic algorithm 41 may be based on a correlation between any of the following and a determination that the power generator 32 is not ready to have a normal operation restarted: a safety trigger indicates that a remote restart may cause a safety risk to human operators in the vicinity of the power generator, historical data regarding past restart commands indicates that at least a predetermined number of restart commands have been transmitted to the power generator in a past period (e.g. at least N times in the past 24 hours, where N is a positive integer number), an internal trigger indicates that a temperature of a component of the power generator is higher than the predetermined range.

[0127] The first deterministic algorithm 41 may be configured, in the absence of any of the above, to determine that the power generator 32 is ready to have a normal operation restarted.

[0128] It would be understood that the above examples causing the deterministic algorithm 41 to consider a power generator to not be ready for a remote restart are purely illustrative, and may be based on any other, or any additional correlation between one or more inputs and an outcome of the determination.

[0129] The first Al model(s) 42 receive, as an input, all variables 61 measured on site of the power generator 32 (namely the internal variables 61a and the external variables 62a), the triggers 62 (i.e. the safety triggers 62a, the internal triggers 62b and the external triggers 62c), the historical data 63, and the data regarding future scheduled maintenance or downtime 64.

[0130] The first Al model(s) 42 may comprise one or more models that can be trained (e.g. using machine learning or deep learning), such as neural networks, to determine whether a power generator can have a normal operation remotely restarted. As an example, the first Al model(s) 42 may be classifier(s) (as an example, of the "extra trees" type), which are trained in a supervised manner, based on a plurality of sets of training values relating to the variables 61, to the triggers 62 and to the historical data 63, each set of training values being associated to a respective label indicating whether the power generator can have a normal operation remotely restarted or not.

[0131] Accordingly, both the first deterministic algorithm(s) 41 and the first Al model(s) 42 may make a separate determination on whether a power generator can have a normal operation remotely restarted and output these respective determinations. As the determination are binary, the outputs may be a one bit value, e.g. 1 if the power generator can have a normal operation remotely restarted and 0 otherwise.

[0132] The combining module 43 is configured to combine the determination from both, and output a combined determination 65.

[0133] In the present example, the combining module 43 may determine that a power generator can have a normal operation remotely restarted only if both the first deterministic algorithm(s) 41 and the first Al model(s) 42 determine so, in which case the determination is made safer. That is, the output of the combining module 43 corresponds to an AND operation being performed on the outputs of the first deterministic algorithm(s) 41 and the first Al model(s) 42.

[0134] Figure 5 shows a schematic diagram of the second function performed by the device, in accordance with example embodiments.

[0135] The second function 50 comprises one or more second deterministic algorithms 51, one or more second Al models 52 and a combining module 53.

[0136] The combining module 53 combines the output of the second deterministic algorithm(s) 51 and the output of the second Al model(s) 52, to generate the restart order 67.

[0137] The variables 61, the triggers 62, the historical data 63, the data regarding the scheduled maintenance and downtime 64, and the forecast of energy production 66 may be provided to the second function 50 for each power generator 32 in the group that is known to be abnormal and can receive a remote command to have a normal operation restarted (e.g. the group determined at step 308).

[0138] The second deterministic algorithm(s) 51 receive, as an input, the triggers 62 (i.e. the safety trigger 62a, the internal triggers 62b and the external trigger 62c). Based on these triggers 62, the second deterministic algorithm (51) orders each power generators 32 in the group to form a restart order.

[0139] For example, the second deterministic algorithm (51) may order the power generators 32 in the group such that the one with the fewest number of activated triggers 62 is restarted first, and order the remaining power generators 32 in the group by increasing number of activated triggers 62.

[0140] The second Al models(s) 52 receive, as inputs, the variables 61, the triggers 62, the historical data 63, the scheduled maintenance and downtime 64, and the forecast information for energy production 66.

[0141] The second Al model(s) 52 may comprise one or more models that can be trained (e.g. using machine learning or deep learning), such as neural networks, to optimise the order in which the power generators 32 should be ordered in the restart order.

[0142] The second deterministic algorithm(s) 51 and the second Al model(s) 52 may each generate a separate restart order, which can be combined by the combining module 43 to output a combined restart order 67. In each of the first function and the second function, the deterministic algorithm(s) allows for a relatively quicker configuration based on a relatively smaller number of variables that are identified to be relevant for the determination (e.g. following Pareto's principle) as these algorithms do not, require real data to perform. On the other hand, the Al model(s) may be trained using historical data and past outputs to improve the determination by identifying further variable over a training period.

[0143] In particular, because the first Al model(s) 42 receive the external triggers, internal variables, external variables, historical data and scheduled maintenance and / or downtime data, which are not used by the first deterministic algorithm(s) 41, the determination from the first Al model(s) 42 may be more accurate (at least once trained) as is takes into account all (or a much larger number of) variables underlying the internal triggers.

[0144] Similarly, by taking into account additional data, and in particular forecast data of energy production and / or meteorological information, the second Al model(s) 52 may define a restart order that can optimise the additional amount of energy obtained by remotely causing the group of power generators to restart a normal operation.

[0145] It would be understood that in each function a different deterministic algorithm and / or a different Al model may be provided for each type of power generator, each power system, etc.

[0146] Referring now to Figure 6, the training of an Al model will now be described.

[0147] The training data for the Al model may be obtained from various sources, depending on the purpose of the training. For example, the Al model may be trained to do one or both of the following: reduce the effect of "false positives" that affect a deterministic algorithm, and predict a time until a next maintenance visit will be required for the power generator.

[0148] In the first case, the training data may be obtained by a comparison of past outputs of the deterministic algorithm and the result of investigations by human operator on the power generator, which may be logged in the historical data regarding past maintenance and abnormalities 63c. The Al model may be trained to detect when the output of the deterministic algorithm does not correspond to the result of the human investigation, and to emphasise an output different from the deterministic algorithm in that case, to reduce the effect of false positives (i.e. outputs of the deterministic algorithm that have been identified to be false.

[0149] In the second case, the training data may be obtained based on the scheduled maintenance and downtime. The Al model may be trained to promote power generators having a longer time until the next scheduled maintenance over power generators having a shorter time until the next scheduled maintenance.

[0150] The first Al model 42 may use this information to determine the abnormal power generators that would likely require a maintenance visit soon (i.e. in less than 1 week from a potential remote restart) to not be ready to have a normal operation remotely restarted. Accordingly, a remote restart that would only lead to a normal operation of a short duration may be avoided.

[0151] The second Al model 52 may use this information to prioritise the remote restart of power generators estimated to require a maintenance visit the latest possible, thus increasing the chance that a new restart of the power generator will be required.

[0152] Referring now to Figure 6, a deterministic algorithm 71 (e.g. a first deterministic algorithm 41 or a second deterministic algorithm 51) can operate in parallel to an Al model 72 (e.g. a first Al model 42 or a second Al model 52), such that they both generate an output (e.g. a determination of whether an abnormal power generator can have a normal operation restarted remotely, or a restart order for a group of power generators), which is then combined by a combining module 73.

[0153] During a training phase, the combining module 73 may store in memory (e.g. memory 23) the outputs of the determination algorithm 71 and outputs of the Al model 72, as a database of past outputs 74.

[0154] A training module 75 obtains the inputs 76 provided to the deterministic algorithm 71, the inputs 77 provided to the Al model 72, and the resulting outputs from the database 74. The training module 75 also obtain the data on past maintenance and abnormalities 63c, and the data on scheduled maintenance and downtime 64.

[0155] The training module 75 trains the Al model using any suitable method, such as a back- propagation method, and a data indicating a correct answer that is derived from the data on past maintenance and abnormalities 63c and / or the data on scheduled maintenance and downtime 64. The training module 75 also calculates an accuracy of the Al model 72, to determine whether the Al model is sufficiently trained.

[0156] The training phase may begin when the device 20 is in operation, and may last for a predetermined duration, or may continue until the training module 75 determines that the accuracy of the Al model 72 is sufficiently accurate (i.e. is above a threshold value). It would be understood that for each Al model, a training phase may have a different duration.

[0157] The training module 75 may also provide, for example after each update to the Al model 72, a value indicating the accuracy of the Al model 72 to the combining module 73. This allows the combining module to assign a weight corresponding to the accuracy of the Al model 72 when combing the outputs of the deterministic algorithm 71 and the Al model 72.

[0158] In summary, it will be appreciated from the description above that certain example embodiments perform processing operations to effect a method as shown in Figure 7 that is performed by a device for controlling a remote restart function of a plurality of solar and / or wind power generators, as described above.

[0159] At step 702, the device receives from each of a plurality of power systems, each power system comprising at least one of the plurality of solar / wind power generators and each power system being separately connected to an electrical grid, a value of one or more variables measured on the site of each of the plurality of solar / wind power generators.

[0160] At step 704, the device performs a first function to determine whether an abnormal solar / wind power generator is ready to have a normal operation remotely restarted based on the received value of the one or more variables.

[0161] At step 706, the device performs a second function to determine a restart order according to which a group of abnormal solar / wind power generators should have a normal operation remotely restarted, among the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted according to the first function.

[0162] At step 708, the device outputs an indication of the determined restart order.

[0163] Modifications and variations

[0164] A trigger, as those described in exemplary embodiments above, need not have a corresponding predetermined range. Instead, the mere reception of a value for a given variable may cause a corresponding trigger to be activated. For example, a sensor located in the nacelle of a wind power generator may detect if a level of fumes is present in the nacelle, and transmit a value if any level of fume in the nacelle is detected. The centralised controller 11 may generate a trigger upon reception of the value, without comparing it to any predetermined range.

[0165] Although it has been explained that the centralised controller 11 holds environmental information (e.g. meteorological quantity) for each power generator 32, the embodiments are not limited in this aspect. Instead, a single representative value may be obtained for a set of one or more (or all) power generators in a power system, for example an average value, which may be used as the value for each power generators. Accordingly, the repeated transmission of values that are unlikely to fluctuate within a power system, or for which an accurate measurement is not crucial, may be avoided.

[0166] In the process described in connection with Figure 3, the process is described to return to step 302 if no power generator 32 is determined to be abnormal (NO at step 304), if no group of power generators is determined to be ready to be remotely restarted (NO at step 308) or when the indication of the determined restart order is output at step 312.

[0167] However, the embodiments are not limited in this aspect. For example, the processing may end after any of these step (i.e. after NO at step 304, after NO at step 308 and / or after step 312), and the device 20 may be triggered to reinitiate the process again, for example periodically.

[0168] In addition, although step 302 is described as a single reception step, the value of the variables may be received repeatedly, even whilst the device 20 performs other steps of the process shown on Figure 3. In that case, the device 20 may use the latest received value of the variables, or any number of last received values corresponding to a predetermined number of values received last (e.g. the last 3 or 5 received values) or values received over a predetermined period (e.g. the values received for the last hour or day)).

[0169] Although in the processing described on Figure 3, the process of steps 308 and 310 are performed after steps 306, the embodiments are not limited in this aspect. It would be understood that the process of Figure 3 may be repeatedly performed. Accordingly, the processes of steps 308, and 310 may be performed in parallel to steps 304 and 306, or before steps 304 and 306, for example by taking into account whether a previous iteration of the process determined a number of abnormal power generators are ready to have a normal operation remotely restarted. In yet other examples, the processes of steps 308 and 310 may be performed independently from steps 304 and 306.

[0170] In some embodiments, the processing of step 304 may be omitted. For example, the centralised controller 11 and / or the power system controller 34 may provide an indication that a power generator is abnormal, such that the step of performing the first function may be performed without requiring the determination of step 304.

[0171] In some embodiments, the processing of step 306 may be omitted. For example, the device 20 may perform the process when it can be assumed that a group of abnormal power generators are ready to be remotely restarted, thus allowing the determination of step 306 to be avoided.

[0172] In some embodiments, the first function (i.e. at least the processing of step 306) may be performed upon detection of a predetermined event. In the exemplary embodiment described with reference to Figure 3, the first function is performed upon detecting that one or more solar / wind power generator is abnormal. Alternatively or additionally, the processing of step 306 may be performed upon detecting another event, such as that a predefined duration elapsed since a start of a timer (e.g. since the processing of step 306 were last performed, in other words, that the processing of step 306 is performed periodically), that a power generator interrupts an operation because an alert was received, the alert blocking the operation of the power generator. For example, the blocking alert may be an alarm that a temperature of a component of the power generator is too high, thus interrupting the generation of electrical power by causing a safe shutdown of the power generator.

[0173] It would be understood that the embodiments are not limited to first deterministic algorithm(s) 41 receiving the internal triggers 62b. Instead, the first deterministic algorithm(s) 41 may receive the internal variables 61a instead of the internal triggers 62a, and to determine the state of each internal triggers based on the value of the corresponding internal variables 61a.

[0174] In other embodiments, the first deterministic algorithm(s) 41 may receive the internal triggers 62b only. For example, the centralised controller 11 may generate no safety trigger 62a, and the historical data on past restart commands may not be available to the device 20.

[0175] It would be understood that the embodiments are not limited to first Al model(s) 42 receiving the triggers 62, in particular as the safety trigger 62a may be handled by the first deterministic algorithm(s) 41 and the first Al model(s) 42 already receives the internal and external variables 61a and 61b from which the other triggers are derived. Similarly, in some embodiments, the first Al model(s) 42 may only receive the internal variables 61a, and not receive the external variables 61b, the triggers 62, the historical data 63 and / or the scheduled maintenance and downtime 64.

[0176] As another example, the output of the first Al model(s) may have a smaller weight in the combination than the output of the first deterministic algorithm(s) 41, which may be determined based on an accuracy of the first Al model(s). For example, during a period where the first Al model(s) is being trained (i.e. a training phase), the combining module 43 may disregard the output of the first Al model(s), or may assign a weight to the output of the first Al model(s) that is proportional to a calculated accuracy of the Al model(s).

[0177] For example, a weight of 0.3 may be given to a first Al model having an accuracy less than 60%. If, for example, the first deterministic algorithm outputs a value '1' whereas the first Al model outputs a value 'O', the combined determination may be calculated as (1 + 0) / (1 + 0.3) ~ 0.76, which the combining module 43 may round up or down to the nearest one of 0 and 1, to maintain a binary decision (in the case of 0.76, this would be rounded up to 1).

[0178] Alternatively, the combining module 43 need not round up or down the combined determination value, which may be provided to the second function 50 as a value indicating which power generators may be safer to remotely restart (for example as a percentage, where a value of 0.76 indicates a 76% chance that the power generator is safe to have a normal operation remotely restarted). The second function 50 may base the restart order at least in part on this information, for example by causing power generators having a higher value to be restarted first.

[0179] In some embodiments, the second function 50 receives, for each abnormal power generator in the group determined to be ready to have a normal operation remotely restarted, the information on a forecast of energy production 66, and the optimization information stored by the centralised controller 11. Based on the optimization information, the second function determines one or more optimization criteria related to the grid and / or the solarwind power generator. The section function 50 then provides the forecast of energy production and the one or more optimization criteria to the deterministic algorithm(s) 51 and to the Al model(s) 52, as inputs. The deterministic algorithm(s) 51 and the Al model(s) 52 each determine a respective restart order, which is then combined by the combining module 53, as explained above. It would be understood that the deterministic algorithm(s) 51 and / or the Al model(s) 52 may receive additional inputs, such as the triggers 62, the variables 61, etc. in addition to the forecast of energy production 66 and the one or more optimization criteria.

[0180] It would be understood that the output of the deterministic algorithm 71 may be used to train the Al model 72 as well. However, the embodiments are not limited in this aspect and the deterministic algorithm, the reception of its inputs 77 by the training module 75, and the storing of past outputs of the deterministic algorithm in database 74 may be omitted.

[0181] It would be understood that the restart command transmitted to a power generator may, in some cases, be unsuccessful. For example, the power generator may fail to receive the restart command (e.g if there is for example an issue on the communication link between the device 20 and the power generator 32), the power generator may fail to process the restart command (e.g if there is a data corruption, or an error occurs in the data processing in the power generator), or the power generator may fail to successfully resume or restart the normal operation (e.g. due to a failure of a component).

[0182] The device 20 may, by receiving new values of the one or more variables measured at a site of a power generator that received a restart command, determine whether the restart command was successful (i.e. that the power generator successfully restarted having a normal operation) or unsuccessful.

[0183] Accordingly, in some embodiments, the historical data 63b on past restart commands may include, in association with each past restart command, an indication of whether the restart command was successful (e.g. in binary data field). This indication may be used in the training of the Al model(s) 75. For example, the Al model(s) 75 may use this indication to determine that a power generator is likely to be unsuccessful in restarting a normal operation, if the value of the variables are similar to a situation having led to an unsuccessful restart command, or if the power generator has had at least a predetermined number of unsuccessful restart command over a given period (e.g. 3 in the past week). As will be apparent to the person skilled in the art, any of the above-described example embodiments describing processing operations may be embodied by a computer program comprising computer- readable instructions, which, when executed by a computer or by one or more processor, causes the computer or the one or more processor to implement the method defined by processing operations. Said computer program may be stored in a transitory or non-transitory computer-readable storage medium, or carried by a signal.

Claims

CLAIMS1. A method for controlling a remote restart function of a plurality of solar and / or wind power generators, hereinafter solar / wind power generator, the method comprising:- receiving, from each of a plurality of power systems, each power system comprising at least one of the plurality of solar / wind power generators and each power system being separately connected to an electrical grid, a value of one or more variables measured on the site of each of the plurality of solar / wind power generators;- performing a first function to determine whether an abnormal solar / wind power generator is ready to have a normal operation remotely restarted based on the received value of the one or more variables,- performing a second function to determine a restart order according to which a group of abnormal solar / wind power generators should have a normal operation remotely restarted, among the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted according to the first function; and- outputting an indication of the determined restart order.

2. The method according to claim 1, wherein the one or more variables measured on site includes at least one of: a temperature of one or more components of the solar / wind power generator, an electrical quantity measured in one or more electrical components of the solar / wind power generator, a mechanical quantity measured for the power generator, a meteorological quantity measured in an environment of the solar / wind power generator, wherein, optionally, the meteorological quantity comprises at least one of a wind speed, a wind direction, a wind distribution, and an amount and / or angle of received sunlight, and, optionally, the electrical quantity comprises a value of a current present in an electrical component of the solar / wind power generator.

3. The method according to claims 1 or claim 2, wherein the first function is performed using a deterministic algorithm and / or an Al model receiving, as inputs, the plurality of variables and a history of remote restarts performed in a predetermined period of time for solar / wind power generators.

4. The method according to any of claims 1 to 3, wherein the restart order is determined based on at least one of a forecast of energy production by means of the abnormal solar / wind power generators that are determined to be ready to have a normal operation remotely restarted, and one or more optimization criteria related to the grid and / or the solar-wind power generator.

5. The method according to claim 4, wherein the second function is performed using a deterministic algorithm and / or a trained Al model which:- elaborate as inputs (i) the forecast of energy production, and (ii) the one or more optimization criteria related to the grid and / or the solar-wind power generator, and- provide as output the restart order.

6. The method according to any of claims 1 to 5, wherein a value of a plurality of types of variables measured on the site of the solar / wind power generators is received, wherein, optionally, at least one type of variable indicates a respective abnormality of a solar / wind power generator, wherein the first function is performed using a first Al model, wherein the first Al model receives, as input, the value of each type of variables, and wherein the first Al model is trained to identify a first subset of the types of variables to be prioritised for determining whether an abnormal solar / wind power generator is ready to have a normal operation remotely restarted.

7. The method according to claim 6, wherein the types of variables measured on the site of the solar / wind power generator comprise one or more type of internal variables related to an internal operation of the solar / wind power generator and one or more type of external variables related to an environment of the solar / wind power generator, wherein the method further comprises receiving one or more triggers, each trigger being for causing an alert to be raised, the alert indicating a potential abnormality related to the power generator or a safety risk related to an operation of the power generator,wherein the one or more triggers comprise at least one internal trigger generated based on an internal variable, the first Al model receives, as input, a value of each type of variables, and the first function is performed using a first deterministic algorithm, the first deterministic algorithm receiving, as input, the at least one internal trigger only.

8. The method according to claim 7, wherein the first deterministic algorithm is based on one or more first predetermined correlations, each first predetermined correlation being between one or more internal trigger and a determination that an abnormal solar / wind power generator is ready to have a normal operation remotely restarted, and the first Al model is trained to identify a new correlation between one or more type of variables and a determination that an abnormal solar / wind power generator is ready to have a normal operation remotely restarted.

9. The method according to any of claims 1 to 8, wherein a value of a plurality of types of variables measured on the site of the solar / wind power generator is received, wherein, optionally, at least one type of variable indicates a respective abnormality of a solar / wind power generator, wherein the second function is performed using a second Al model, wherein the second Al model receives, as input, the value of each type of variables, and wherein the second Al model is trained to identify a subset of the types of variables to be prioritised for determining the restart order.

10. The method according to claim 9, wherein the types of variables measured on the site of the solar / wind power generator comprise one or more type of internal variables related to an internal operation of the solar / wind power generator and one or more type of external variables related to an environment of the solar / wind power generator,wherein the method further comprises receiving one or more triggers, each trigger being for causing an alert to be raised, the alert indicating a potential abnormality related to the power generator or a safety risk related to an operation of the power generator, wherein the one or more triggers comprise at least one internal trigger generated based on an internal variable, the second Al model receives, as input, a value of each type of variables, and the second function is performed using a second deterministic algorithm, the second deterministic algorithm receiving, as input, the one or more triggers only.

11. The method according to claim 10, wherein the second deterministic algorithm is based on one or more second predetermined correlations, each second predetermined correlation being between one or more triggers and a prioritisation of one abnormal solar / wind power generator ready to have a normal operation remotely restarted over another, and the second Al model is trained to identify a new correlation between one or more type of variables and a prioritisation of one abnormal solar / wind power generator ready to have a normal operation remotely restarted over another.

12. The method according to any of claims 1 to 11, wherein outputting the indication of the determined restart order comprises at least one of notifying an operator of the determined restart order, and automatically transmitting a restart command to the solar / wind power generators in accordance with the determined restart order.

13. The method according to any of claims 1 to 12, wherein the first function is performed upon detection of a predetermined event, the predetermined event including at least one of a predefined duration elapsing and a solar / wind power generator interrupting an operation due to an alert.

14. The method according to any of claims 1 to 13, wherein the second function is performed every time an abnormal solar / wind power generator is determined to be ready to have a normal operation remotely restarted by means of the first function.

15. The method according to any of claims 1 to 14, wherein each abnormal solar / wind power generator is out-of-service or operates at a reduced capacity.

16. A computer program comprising instructions which, when executed by one or more processors, cause the one or more processors to execute the method according to any of claims 1 to 15.

17. A device comprising means for performing the method according to any of claims I to 14.

18. A system comprising the device according to claim 16 and a plurality of power systems, each power system comprising at least one solar / wind power generator and each power system being separately connected to an electrical grid.

Images