Coordinated control of an aggregate, to provide a primary frequency adjustment

Abstract

The invention relates to a method for controlling a virtual power plant that is composed of a battery and hydroelectric power plants and is connected to a power grid. The virtual power plant is subject to an overall power setpoint exceeding the operational limit of the battery. A first setpoint is sent to a local controller of a hydroelectric power plant, the first setpoint not taking into account hydroelectric power measurements but depending on an order of priority and on limits of the battery. Next, a second setpoint is provided, based on measurement of the overall power of the hydroelectric power plants. Simultaneously, the battery receives a third setpoint adjusted depending on the overall setpoint and the hydroelectric power measurements. The setpoints aim to maintain the stability over time in the overall power supplied by the virtual power plant, in compliance with the requested power setpoint.

Description

Description Title: Coordinated control of an aggregate to provide primary frequency control technical field

[0001] The present invention relates to the field of hydroelectric power plants, in particular of the "run-of-river" type, and battery storage systems interfaced by power electronic converters (inverters), connected on the same synchronous electrical network.

[0002] Aggregations of decentralized sources, also called "virtual power plants" or "VPP" ("Virtual Power Plant" in English), participate in the primary frequency regulation of the network.

[0003] The present invention relates more particularly to a method of controlling such a virtual power plant, as well as to a centralized controller implementing such a method and to a computer program intended to be executed by such a centralized controller.

[0004] This control method allows for the coordinated control of a battery and a set of run-of-river hydroelectric power plants, having different dynamic characteristics, to provide an aggregated primary frequency control service, particularly when the network is in an "Alert State" or an "Emergency State".

[0005] This service can in particular be integrated into the "FCR" ("Frequency Containment Reserve" in English), which is the primary frequency control service of the European electricity system, and more specifically the expected operation of so-called "limited reservoir" (LER or "Limited Energy Reservoir" in English) aggregates when the network is in "Alert State" or "Emergency State", triggered by criteria on the network frequency. Previous technique

[0006] Today, more and more new technologies, such as battery storage systems, are being installed in electrical grids to provide services to the grid, such as primary frequency regulation. Frequency regulation has historically been provided by conventional generation methods (hydroelectric, thermal, nuclear, etc.), and this comes at a cost to the producer, because to maintain the reserve associated with frequency regulation, the generating units must keep a reserve and therefore operate below their nominal power output.

[0007] Using batteries to provide this frequency regulation service can optimize conventional power generation equipment. Indeed, if the entire reserve is provided by batteries, this relieves a constraint on conventional generators, which can then operate at their nominal power.

[0008] Numerous battery-powered frequency control projects have already been built in France and the rest of the world.

[0009] Today, some existing means of production, such as run-of-river hydropower, have existing active power flexibility capacities which are not exploited in practice, but which can, technically, be activated on an ad hoc basis to provide primary frequency reserve in the event of a major incident on the network causing a significant variation in frequency.

[0010] The aggregation of a battery with one or more run-of-river hydroelectric power plants can optimize both the sizing of the battery and the use of the hydroelectric power plants, from a technical and economic point of view.

[0011] Under normal operating conditions (frequency close to the reference value, i.e. 50 Hz in France), hydroelectric power plants can be placed at their optimal operating point, which depends in particular on the incoming hydraulic flow, while the battery ensures the entire frequency adjustment (it discharges when the frequency is low, and recharges when the frequency is high).

[0012] In exceptional circumstances (large frequency variations), the hydroelectric power plants connected to the battery can temporarily increase or decrease their output to provide additional primary reserve, supplementing the power supplied or absorbed by the battery. In this way, the suboptimal operation of the hydroelectric power plants is limited to these exceptional events, and it is possible to supply the grid with a power reserve exceeding the battery's maximum capacity. This allows for optimized battery sizing, which can improve the profitability of a project.

[0013] The patent application FR3133714A1 describes a known method of control, implemented by a centralized controller, of a virtual power plant composed of a battery and a set of hydroelectric power plants, in order to perform a primary adjustment of the frequency of the electrical network.

[0014] This document presents a mechanism by which a battery power adjustment setpoint, calculated on the basis of a total power produced by The power output of all hydroelectric power plants at a given time is provided to a local battery controller. The battery power adjustment setpoint is intended to modify the battery control by the local controller in order to compensate for a variation in total power relative to an overall setpoint power for all hydroelectric power plants.

[0015] According to this mechanism, a power adjustment command for a hydroelectric power plant, calculated based on a grid frequency measurement, is also provided to a corresponding local controller. The power adjustment command for the hydroelectric power plant is intended to cause a variation in the power produced by said hydroelectric power plant in accordance with a battery management strategy.

[0016] However, a difficulty arises in the event of exceptional frequency incidents, during which it is desirable for the virtual power plant to provide, for a prolonged period, a power reserve greater than that which can be provided by the battery alone.

[0017] The patent application FR3133714A1 does not provide sufficient information to enable a person skilled in the art to control a virtual power plant in such a way as to not only activate but also maintain a sufficient power response from the virtual power plant in such scenarios.

[0018] Therefore, there is a need for a control mechanism for a virtual power plant capable of responding effectively to exceptional frequency incidents by providing an adequate power reserve over time. Summary

[0019] This disclosure improves the situation.

[0020] According to one aspect, a control method is proposed, implemented by a control entity, for a virtual power plant connected to an electrical network, the virtual power plant comprising at least one battery and a set of hydroelectric power plants, the virtual power plant being subject to an overall power setpoint exceeding an operational power limit of the battery, the method comprising: a provision of a first power adjustment setpoint to a local controller of a hydroelectric power plant within the group of hydroelectric power plants, the first adjustment setpoint taking into account a priority order of the hydroelectric power plants and the operational power limit of the battery, and the first setpoint adjustment that does not take into account any power measurement; and After the first power adjustment command is provided, a second power adjustment command is provided to the local controller of the hydroelectric plant, the second power adjustment command taking into account the overall power measurement of all the hydroelectric plants, in which, when the second power adjustment setpoint is provided, the battery is subjected to a third power adjustment setpoint which takes into account the overall power setpoint and the overall power measurement of the entire hydroelectric power plant complex, and in which the supplies of the adjustment setpoints contribute to maintaining a temporal stability of an overall power of the virtual plant in accordance with the overall power setpoint.

[0021] According to one aspect, a computer program is proposed containing instructions for the implementation of the process described herein when this program is executed by a control entity of a virtual power plant connected to an electrical network, the virtual power plant comprising at least one battery and a set of hydroelectric power plants.

[0022] According to one aspect, a data storage medium is also proposed that stores the instructions for the computer program described herein.

[0023] According to one aspect, a control entity for a virtual power plant connected to an electrical grid is also proposed, the virtual power plant comprising at least one battery and a set of hydroelectric power plants, the virtual power plant being subject to an overall power setpoint exceeding an operational battery power limit, the control entity being configured to implement: the provision of an initial power adjustment command to a local controller of a hydroelectric power station within the group of hydroelectric power stations, the initial adjustment command taking into account a priority order of the hydroelectric power stations and the operational power limit of the battery, and the initial adjustment command not taking into account any power measurement; and After the first power adjustment command is provided, a second power adjustment command is provided to the local controller of the hydroelectric plant, the second power adjustment command taking into account an overall power measurement of all the hydroelectric plants, in which, when providing the second power adjustment command, the battery is subject to a third power adjustment setpoint which takes into account the overall power setpoint and the overall power measurement of all hydroelectric power plants, and in which the supplies of the adjustment setpoints contribute to maintaining a temporal stability of an overall power of the virtual plant in accordance with the overall power setpoint.

[0024] Providing the initial adjustment command allows for the management of an overall power demand that exceeds the battery's operational power limit by intelligently distributing the excess power among the hydroelectric plants, taking into account their priority order. This approach ensures better utilization of available resources and prevents overloading the battery.

[0025] Providing the second adjustment setpoint allows for dynamic adjustment of the power distribution. This enables a response to fluctuations and adaptation of the setpoints based on the actual performance of the hydroelectric power plants, thereby improving the responsiveness and reliability of the virtual power plant for primary frequency control.

[0026] The temporal stability of the overall power supplied by the virtual power plant makes it possible to avoid rapid fluctuations that could compromise the stability of the electrical grid.

[0027] The proposed process, computer program, control entity and data storage medium may optionally include certain additional functions as defined below.

[0028] In one example, the first power adjustment instruction takes into account at least one parameter of the hydroelectric power plant from among: availability, an available capacity, a response time, a power limitation.

[0029] These parameters make it possible to improve coordination between the different means of production and their efficient use.

[0030] In one example, the first power adjustment instruction takes into account an estimated power regulation gain from the hydroelectric plant following a first delay, and the provision of the second power adjustment instruction is implemented after a second delay that is longer than the first delay.

[0031] This can help avoid overcompensation or excessive response delays, thus optimizing frequency regulation.

[0032] In one example, the second power adjustment setpoint is periodically updated based on a fluctuation in an overall power measurement of the hydroelectric plants and / or a fluctuation in an environmental parameter of at least one hydroelectric plant in the set of hydroelectric plants.

[0033] This can allow for a more precise response to network needs, finer adjustment of available resources and faster compensation of potential discrepancies between a setpoint power and an associated power measurement.

[0034] In one example, the battery's operational power limit takes into account the battery's state of charge and / or thermal condition.

[0035] This improves battery usage under safe conditions.

[0036] In one example, the provision of the first adjustment setpoint is carried out during a rapid variation in a frequency of the electrical network.

[0037] For example, the frequency variation may be greater than 50 MHz, or greater than 100 MHz, or greater than 150 MHz, or greater than 200 MHz in a period of less than five minutes, or less than two minutes, or less than one minute, or less than 40 seconds, or less than 20 seconds, or less than 10 seconds.

[0038] In one example, The overall power setpoint is adjusted, if necessary, after the first adjustment setpoint is provided and before the second adjustment setpoint is provided, and The second adjustment instruction takes into account the overall adjusted power instruction.

[0039] Adjusting the overall setpoint allows the plant to react to changes in production and demand conditions before applying the second setpoint. This can enable more precise regulation and improved responsiveness.

[0040] In one example, the second adjustment instruction includes compensation for at least one unforeseen power variation from at least one hydroelectric plant in the set of hydroelectric plants.

[0041] This can improve the stability of the plant's response to the overall power setpoint.

[0042] In one example, a first decision tree is used to determine the first adjustment instruction.

[0043] In one example, a second decision tree is used to determine the second adjustment instruction.

[0044] Using decision trees to determine adjustment guidelines allows for a structured and predictable approach to decision-making.

[0045] In one example, the process includes providing a set of first power adjustment instructions to local controllers of a plurality of hydroelectric generating stations in the hydroelectric generating station set, the set of first adjustment instructions taking into account an order of priority and not taking into account any power measurement, and the provision of the set of first adjustment instructions including the provision of the first adjustment instruction.

[0046] By providing a set of simultaneous instructions to multiple power plants, the proposed method improves load distribution and reduces the response time of the virtual power plant. This allows for faster and better-coordinated activation of hydroelectric power plants.

[0047] In one example, the process includes a verification of the hydroelectric plant's capacity to respond to the first adjustment setpoint or the second adjustment setpoint.

[0048] Verifying the ability of power plants to follow instructions prevents resource overloads or underutilization. This allows instructions to be adapted in real time to the actual capacities of the power plants, thus improving power distribution.

[0049] In one example, the second adjustment setpoint takes into account a comparison between a power measurement of the hydroelectric plant after the provision of the first adjustment setpoint and a target value.

[0050] This comparison allows verification that the control unit is correctly following the initial setpoint and, in case of a discrepancy, adjusting the second setpoint accordingly. This can lead to greater precision in adjusting the power delivered by the control unit.

[0051] In one example, the decision tree adjusts the adjustment instructions for hydroelectric power plants based on an iterative analysis, until the overall power reaches a stable value within a defined tolerance range.

[0052] This improves the stability of the virtual power plant's response to network requirements. Brief description of the drawings

[0053] Other features, details, and advantages will become apparent upon reading the detailed description below and analyzing the attached drawings, on which: Fig. 1

[0054] [Fig. 1] illustrates, in the form of a decision tree, the provision of at least one initial power adjustment instruction, in an example implementation. Fig. 2

[0055] [Fig. 2] illustrates, in the form of a decision tree, the provision of at least one initial power adjustment instruction, in an example implementation. Fig. 3 Fig. 4 Fig. 5 Fig. 6

[0056] [Fig. 3], [Fig. 4], [Fig. 5] and [Fig. 6] illustrate results of a simulation of the implementation of the provision of initial power adjustment instructions according to the decision tree in Figure 1. Fig. 7

[0057] [Fig. 7] illustrates, in the form of a decision tree, the provision of at least one second power adjustment instruction, in an example implementation. Fig. 8

[0058] [Fig. 8] illustrates, in the form of a decision tree, the provision of at least one second power adjustment instruction, in an example implementation. Fig. 9

[0059] [Fig. 9] illustrates, in the form of a decision tree, the provision of at least one second power adjustment instruction, in an example implementation. Fig. 10

[0060] [Fig. 10] illustrates, in the form of a decision tree, the provision of at least one second power adjustment instruction, in an example implementation. Fig. 11 Fig. 12 Fig. 13 Fig. 14

[0061] [Fig. 11], [Fig. 12], [Fig. 13] and [Fig. 14] illustrate results of a simulation of the implementation of the provision of second power adjustment instructions according to the decision trees in Figures 7 and 9. Description of the implementation methods

[0062] The drawings and description may not only help to better understand this disclosure, but may also contribute to its definition, if necessary.

[0063] The proposed technique concerns the control of a virtual power plant comprising a battery and a set of hydroelectric power plants, for example of the "run-of-river" type.

[0064] The virtual power plant can also be referred to by the acronym "VPP" for "virtual power plant" in English, or by the term aggregate.

[0065] Reference is now being made to a known method of controlling a virtual power plant.

[0066] This process is implemented by a two-level control architecture, in which the battery and the hydroelectric power plants are each controlled by a local controller, and a centralized controller communicates with the local controllers via communication links.

[0067] The centralized controller, or control entity, is a device or system responsible for managing, monitoring, and adjusting the operation of a virtual power plant. The control entity coordinates all the production resources of the virtual power plant. The control entity can be, for example, a server or software installed on such a server. The control entity receives frequency measurements from the grid and power measurements from the virtual power plant's production resources and then generates and transmits commands. intended for local controllers to increase or decrease the power delivered by each means of production.

[0068] In this control architecture, the local battery controller is configured to perform primary frequency regulation of the electrical grid, i.e., a power response proportional to the frequency variation, based on a control gain that can be transmitted by the central controller, a measured electrical grid frequency, and the nominal electrical grid frequency. The control gain can also be a fixed or variable parameter directly implemented in the local battery controller. It is also known to add a battery power adjustment setpoint provided by the central controller to the frequency regulation term.

[0069] The overall battery setpoint, denoted P re f bat can thus be expressed as follows: with: K is the primary reserve control gain (typically expressed in MW / Hz), f the frequency measurement of the electrical network, f0, the nominal frequency of the electrical network, P Obat the battery power adjustment setting.

[0070] In the control architecture, the local controllers of hydroelectric power plants are configured to perform level regulation, i.e., regulation of the power or flow of a hydroelectric power plant by the corresponding local controller based on the measurement of the level of its reservoir or the measurement of the incoming hydraulic flow.

[0071] Such regulation can, for example, increase the power (or flow rate) setpoint when the tank level is too high in order to lower the tank level, and conversely, reduce the power (or flow rate) setpoint when the level is too low in order to raise the tank level. One or more setpoint levels can be defined for this purpose. Such regulation can be implemented using a Proportional-Integral ("PI") controller within the local controller.

[0072] The level control function, or dynamic level control (known as "RGZ"), of a hydroelectric power plant regulates the level of the hydraulic reservoir supplying the plant with water to a setpoint level. This function is This is important for the safety of hydraulic installations and therefore must not be deactivated when a hydroelectric plant is integrated into a virtual power plant. Level control must therefore remain the priority over other instructions.

[0073] However, a power adjustment setpoint provided by the centralized controller can be added to the level regulation setting.

[0074] The overall setpoint of a hydroelectric power plant, denoted P r ef hydro , can thus be expressed in the following way: P ref_hydro = P rgz + P flex with: P rgz the power setpoint determined by the local controller for level regulation purposes, and Pfi ex the power adjustment setpoint provided by the centralized controller.

[0075] Battery control relies on measurements from power sensors connected to the local controllers of the hydroelectric power plants. These power measurements are transmitted from the local controllers to the central controller.

[0076] This allows the centralized controller to monitor in real time the overall power output from the hydroelectric plants of the virtual power plant and to compensate for variations in overall power in real time by adjusting the setpoint P Obat transmitted to the local battery controller.

[0077] This compensation is carried out so that the total power, denoted P vpp The power of the virtual power plant, that is to say the sum of the power of the battery and the hydroelectric power plants, follows the following equation: P Ovppa desired overall power setpoint at the global level of the virtual power plant, which according to this equation is the power produced by the virtual power plant when the current frequency of the electrical grid is equal to the nominal frequency, K is a constant representing the regulating energy of the virtual power plant, f is the current frequency of the electrical grid, and f0la nominal frequency of the electrical network, which is for example set at 50 Hz in Europe.

[0078] In other words, the instruction P Obat The battery power adjustment is adjusted to compensate for measured power fluctuations from all the hydroelectric plants in the virtual power plant, denoted P tot_hydro The instruction P Obat is calculated using the following equation:

[0079] The known method for controlling a virtual power plant, as described, has certain limitations when it comes to managing exceptional situations on the electrical grid, such as sudden and significant frequency variations. In particular, the method does not include a mechanism for effectively responding to the constraints imposed by such exceptional situations, where it becomes necessary to activate and maintain a substantial power reserve over an extended period.

[0080] In the European Union, electricity grid regulation imposes strict requirements for responding to frequency deviations, particularly in "alert" and "emergency" situations. These states, defined by European regulations, impose specific constraints on limited energy storage systems and, by extension, on aggregates composed of these storage systems—that is, virtual power plants like the one described.

[0081] The alert state is declared when the absolute deviation of the frequency from the nominal frequency f0 (e.g., 50 Hz in Europe) exceeds 50 MHz for more than 15 minutes, or exceeds 100 MHz for more than 5 minutes, but does not exceed 200 MHz. In this case, regulations may require the aggregator to freeze its power setpoint P. Ovpp to stabilize the situation.

[0082] The state of emergency, or "emergency state" in English, is declared when the absolute deviation of the frequency from the nominal frequency f0 exceeds 200 mHz.

[0083] Furthermore, according to European regulations, the power reserve must be fully activated in less than 30 seconds to meet network security requirements.

[0084] In both situations, any power variation within the aggregate independent of the frequency variation, such as that caused by the activation of a level control system at a hydroelectric plant, can lead to an unwanted power variation at the aggregate level if it is not properly compensated by another unit within the aggregate, thus rendering the power response of the VPP non-compliant. This situation can pose a risk to the stability of the electrical grid, and the operator of the VPP may be penalized by the network manager for non-compliance of the service provided.

[0085] During exceptional frequency incidents for which the overall instruction P Ovpp must be fixed, and when the aggregate needs to provide a significant power variation for a certain duration, the activation sequence of the hydroelectric power plant reserves must be executed very precisely, because in these situations, the battery can become saturated at its maximum (or minimum) power, thus limiting its ability to compensate for all power variations from the hydroelectric power plants to maintain the overall VPP power at its setpoint level. In these situations, the adjustment setpoints P fiex Hydroelectric power plants must be calculated dynamically using a process specific to the management of these adjustment instructions.

[0086] However, the known method does not include any algorithm for calculating the Pfi setpoints. ex hydroelectric power plants that are effective in the event of exceptionally frequent occurrences, such as a grid alert or emergency.

[0087] Therefore, there is a need for a control process capable, specifically, of managing the power commands of hydroelectric power plants in these situations.

[0088] To this end, one objective of the proposed technique is to provide primary frequency reserve in the event of a significant frequency deviation of the network requiring a significant activation of the reserve at the VPP level, particularly in scenarios where the network is in a state of alert or emergency.

[0089] The general principle of the proposed technique is based on controlling hydroelectric power plants by modifying their power adjustment instructions in order to optimize the activation of their reserve, taking into account their physical characteristics such as their response times and their level regulation dynamics, and supplementing the reserve with the power of the battery which, in this situation, may approach, or reach, its operational power limit.

[0090] The battery's operational power limit refers to the maximum or minimum power that the battery can safely deliver or absorb for a given period without risk of damage or malfunction. This limit is determined by the battery's technical characteristics (e.g., its nominal capacity, state of charge, temperature) and can vary depending on operating conditions. For example, if a battery can deliver a maximum power of 12 MW, its operational power limit can be set at 10 MW to avoid reaching extreme physical limits and thus preserve battery life.

[0091] With this control strategy, battery sizing can be optimized, and hydroelectric power plants are used to provide additional frequency reserve to that of the battery in case of large frequency deviations for which the power of the battery alone is insufficient.

[0092] An example of implementing the proposed technique is now described. Unless otherwise specified, all detailed actions can be implemented by the centralized controller.

[0093] A global instruction P vpp The virtual power plant is calculated based on a basic setpoint P Ovpp and the frequency f of the electrical network measured at a given instant t.

[0094] The overall instruction P vpp can, for example, be calculated in the same way as in the known method, namely according to the following formula:

[0095] This presentation focuses on situations where the basic instruction P Ovpp is fixed, that is to say, constant over time. Such situations correspond to scenarios where the state of the network switches either to a state of alert or to a state of emergency, in accordance with European regulations.

[0096] Of course, disregarding these regulatory considerations, the proposed method is also applicable to situations where the basic instruction P Ovpp is variable over time. In such situations, the virtual power plant's power target can be updated according to variations in the basic setpoint.

[0097] To better illustrate the scenarios where the basic instruction P Ovpp is frozen, let us consider two distinct moments.

[0098] At a first instant t1, before the transition to a state of alert or emergency, the total power P to t hydro ( fi) produced by hydroelectric power plants is measured and transmitted to the central controller. Initially, the battery adjustment setpoint is assumed to be zero: P 0_bat (t1) = 0. Similarly, at the first instant, the contribution of hydroelectric power plants to the power reserve: P flex (t1) = 0. Thus, the instruction of base at the first instant is equal to the sum of the hydroelectric powers measured at the first instant

[0099] At a second instant t2, where the measured frequency f(t2) is such that the network enters an alert or emergency state, the basic setpoint is fixed; that is, it is kept constant and can no longer be modified to account for subsequent frequency variations. Thus, P 0_vpp (t2) = = P tot_hydro (t1).

[0100] Thus, in the case of piloting according to the known process, with Pref_bat = -K(f - f0) + P 0_bat , and in an initial situation where hydroelectric power plants do not contribute to the P reserve flex = 0) and the battery adjustment setpoint is initially zero (P Obat = 0), then P Ovpp is equal to the sum of the hydroelectric powers measured at the moment preceding the transition to a state of alert or state of emergency of the network which leads to the freezing of this setpoint.

[0101] At the second instant t2, a target of power P cible is also calculated. This power target corresponds to the power that the battery should provide to ensure the primary frequency regulation on its own, i.e. if the battery was not associated with the set of hydroelectric power plants, and if the battery had no power limit.

[0102] For example, the power target can be calculated by applying the following formula: P cible= -K(f - f0).

[0103] The power target is compared to an operational power limit of the battery in order to calculate a reserve setpoint that the hydroelectric power plants are responsible for activating.

[0104] In a situation where the power target P cibie is positive and greater than the maximum power P max_bat , of the battery in injection (discharge) then the hydroelectric power adjustment setpoint P ref_hydro is calculated as the difference between the target power P cible and P max_bat .

[0105] In a situation where the power target P cibie is negative and less than the maximum power of the battery during charging (draw-through) P min_bat , then the hydroelectric power setpoint P ref_hydro is calculated as the difference between the minimum battery power and the target power P cible .

[0106] In other situations, the hydroelectric power setpoint can be set to zero, the primary frequency control can be ensured exclusively by the battery, and the control of the virtual power plant can be implemented in the same way as the known process.

[0107] These different situations can be translated into the following instructions: SI P cible > P max_bat SO P ref_hydro = P cible - P max_bat , If target < minimum bat SO P r ef hydro ~ Pmin bat ~ Target, e t

[0108] In situations where a hydroelectric setpoint P ref_hydro If a non-zero value is determined, it can be distributed among the different hydroelectric power plants, for example in the form of adjustment instructions P fiex .

[0109] Depending on a possible distribution, a two-phase coordination is implemented.

[0110] In a first phase, initial adjustment instructions are transmitted to the local controllers of a sufficient number of hydroelectric power plants, following a predefined order of priority, to ensure that the total target power is reached in 30 seconds, taking into account the dynamic characteristics of each power plant.

[0111] In a second phase, the adjustment setpoints are regulated in a closed loop, in order to follow as closely as possible the overall hydroelectric setpoint P ref_hydro and to compensate for power variations in individual hydroelectric power plants, for example, related to their level regulation dynamics.

[0112] In parallel with the calculation and sending of power adjustment instructions to hydroelectric power plants, an adjustment instruction P 0_batis calculated and sent to the battery in order to finely regulate the overall power at the VPP level, based on the actual power produced and measured at the level of all the hydroelectric power plants.

[0113] Reference is now made to figures 1 and 2, which each detail a flowchart of a possible algorithm suitable for implementing the initial phase of activation of a hydroelectric power reserve, or "first phase" in examples of implementation of the proposed technique.

[0114] As a reminder, in the event of a significant frequency deviation triggering the network's alert or emergency state, the power setpoint of the virtual power plant (i.e., the power outside the setpoint, or in other words, the power that the power plant virtual electrical supply would provide if the current network frequency was equal to the nominal network frequency) must be fixed, according to the European rules in force defining the behavior of limited energy assets participating in primary frequency regulation (FCR).

[0115] In the context of this discussion, this amounts to fixing the value of P. Ovpp -

[0116] Figure 1 presents, in the form of a decision tree, a possible set of instructions that can be executed by the centralized controller, in an example implementation, to calculate the power adjustment setpoints of each hydroelectric power plant in the case where P cible > P max_bat and P ref_hydro > 0.

[0117] In this case, one objective is to increase the power produced by hydroelectric plants to provide upside reserve. This situation corresponds to a grid alert or emergency state that requires a significant call for reserve from the virtual power plant, exceeding the maximum operational power of the battery alone.

[0118] During the initialization of the algorithm at block 11, a total power to be distributed among the hydroelectric power plants is determined from the quantity P ref_hydro calculated previously, and adding a margin P mar This margin can be set to anticipate and compensate for the potential non-response of one or more power plants, for example (i.e., coverage in case of an unforeseen event). Thus, the total power P restante to be distributed across hydroelectric power plants can be defined as P restante = P ref_hydro + P marge = P cible - P max_bat + Pmarge .

[0119] There are N hydroelectric power plants, where N > 1, and they are ranked according to a priority order, from rank 1 to rank N.

[0120] During the initialization of the algorithm at block 11, a counter i is initialized: i = 0.

[0121] In block 12, the value of counter i is compared to the number N of hydroelectric power plants.

[0122] If i = N, then the algorithm has already considered all the hydroelectric power plants of the virtual power plant in turn; this is a stopping criterion for the algorithm. The calculation of the power adjustment setpoints P ftex is stopped at block 14 and is considered finished.

[0123] If i < TV, then the algorithm has not considered all the hydroelectric plants in the virtual power plant. The value of counter i is incremented by 1 in block 13, and then the hydroelectric plant with rank i is considered in block 15.

[0124] If the rank i hydroelectric plant is not available to participate in the primary frequency adjustment, a zero setpoint is assigned to the rank i hydroelectric plant in block 17 and the algorithm continues its operation in block 12.

[0125] If the first-order hydroelectric power station is available to participate in primary frequency regulation, then the value of the total power P res The amount to be distributed is checked in block 16.

[0126] If the value of Postante is equal to zero, that is, if there is no remaining need for power reserve, then a zero setpoint is assigned to the hydroelectric power plant of rank i in block 17 and the algorithm continues its operation in block 12.

[0127] If the value of Postante is greater than zero, that is, if there is still a need for power reserve, then a positive setpoint P flex (i), which corresponds to all or part of P restante , is allocated to the rank i hydroelectric power station in block 18. The value of P flex (i) complies with an initial maximum setpoint P flex_init (i) predefined for the i-rank central unit.

[0128] In other words,

[0129] Pfi ex . nit(i) is the maximum permissible power variation setpoint for the response of the rank i hydroelectric power plant to the demand of the virtual power plant in the initial phase of activation of the hydroelectric power reserve.

[0130] When a non-zero adjustment setpoint P fiex (i is determined for a rank i power plant in block 18, the power P 30s (j) which will actually be released by this power plant 30 seconds after the supply of Pf tex i is determined or estimated in block 19.

[0131] Indeed, due to slow response times and / or gains that may exist on the local controllers of hydroelectric power plants, the adjustment setpoint Pfiex(i P e must not be respected within 30 seconds. A parameter denoted K 30sis then defined for each power plant in order to calculate its expected contribution as a function of the adjustment setpoint according to the following formula: P 30s (0 = K 30s (i) x Pfi ex (i.

[0132] For example, for a k-rank hydroelectric power plant that responds linearly to a setpoint with a response time of 60 seconds, it can be determined that half the power of the adjustment setpoint will be supplied by this plant after 30 seconds. In this case, the value of the parameter K 30s perhaps fixed at K 30s (k) = 0.5.

[0133] Finally, in block 20, the value of the remaining power to be distributed (Posting) is updated by subtracting the expected contribution from the i-th power plant: P restante = P restante - P 30s (i), then the algorithm continues its operation at block 12.

[0134] In this way, the overall instruction P ref_hydrois fully distributed across all or part of the available power plants (within the limit of the sum of all the Pflexjnit limits) ■

[0135] Each hydroelectric power plant is considered iteratively, one by one following a pre-established priority order, to allocate the total power to be distributed. The decision tree in Figure 1 does not take into account any power measurement of any individual hydroelectric power plant, nor any total power measurement of all hydroelectric power plants.

[0136] Similarly, Figure 2 presents a possible set of instructions that can be executed by the centralized controller, in an example embodiment, to calculate the power adjustment setpoints for each hydroelectric power plant in the case

[0137] In this case, one objective is to reduce the power produced by hydroelectric plants to provide reserve downside, in the event of a significant rise in network frequency placing the network in a state of alert or emergency, and requiring a significant call for reserve from the virtual power plant, exceeding the minimum operational power of the battery alone.

[0138] The instructions shown in Figure 2 are identical to those in Figure 1, with the following exceptions.

[0139] These exceptions are due to the fact that, in Figure 2, it is a matter of activating a power reserve downwards by distributing negative power (i.e. by reducing hydroelectric production) to meet the demand on the network.

[0140] In block 11, Prestante is calculated according to the following formula: Prestante Pref hydro Pmarge, where Prestante and Prefhydro have negative values, and Pmarge has a positive value.

[0141] In block 16, the outcomes of the verification of the total power output to be distributed are as follows: either Prestante is equal to 0, in which case a zero setpoint is assigned to the rank i hydroelectric power plant in block 17, either Prestante is less than 0, in which case in block 18, the setpoint P fiex (i) for a central unit of rank i is negative. It is determined according to the following formula: where Pfiex init (i) est a positive value defining the maximum power reduction allowed for the rank i power plant during the activation phase of the hydroelectric power reserve.

[0142] In block 19, the power P 30s (j) released after 30 seconds is calculated according to formula P 30s (i) = -K 30s (i) × P flex (i).

[0143] A practical example is provided to illustrate the operation of the algorithm in Figure 1 for activating an upstream hydroelectric power reserve. In this example, it is assumed that five hydroelectric power plants are aggregated with a battery in a virtual power plant.

[0144] The individual characteristics of the hydroelectric power plants are provided in Table 1. [Table 1]

[0145] The individual characteristics of the hydroelectric power plants considered in Table 1 include the availability of power plants, their available capacity, their response time, and their power limitation.

[0146] The order of priority is established here based on the individual characteristics provided.

[0147] It should be noted that the order of priority can be set by taking into account more or less individual characteristics of the power plants, including for example, the efficiency of energy conversion, local environmental conditions, the maintenance history of the power plant, its operational flexibility, its ability to provide auxiliary services, a need for regulation of the level of the power plant's tanks, regulatory constraints, a production cost, etc.

[0148] The order of priority can be set by taking into account a combination of individual characteristics of the power plants.

[0149] For example, available power plants can be ranked according to their estimated power regulation gain after a given delay (e.g., 30 seconds), this gain being a function of their response time and available capacity.

[0150] The order of priority can also be set arbitrarily without taking into account any individual characteristics of the power plants.

[0151] The parameter It is not used in the initial activation phase of the hydroelectric reserve. Its definition and terms of use are detailed later in this document.

[0152] The battery specifications are P maXh f = 10 MW, and P m r ji n rib.at = -10 MW.

[0153] The virtual control unit parameters for frequency adjustment are: K = 000W / Hz, and f0 = 50 Hz.

[0154] The power margin is P marge = 1 MW.

[0155] Consider an incident in which the electrical grid frequency suddenly drops from 50 Hz to 49.8 Hz. The target power is then calculated using the following formula: Pcibie = -K f ~ = -60 X (49.8 - 50) = 12 MW.

[0156] The virtual power plant must then provide 12 MW of reserve power for increased demand. Since the battery's operational power limit is 10 MW, the contribution of hydroelectric power plants is necessary, and at least 2 MW must be activated within 30 seconds to reach a total of 12 MW at the virtual power plant within 30 seconds. The battery is considered fast and capable of supplying the full 10 MW in less than 30 seconds.

[0157] There are 5 hydroelectric power plants, so N = 5.

[0158] Following the logical scheme presented in the previous section, the following steps are implemented successively.

[0159] Condition P cib[e > P maXbat is indeed verified, and we have: ref hydro ~ target ~ Pmax bat ~ 12 — 10 — 2 MW. We add the margin: We initialize the counter i = 0.

[0160] 1 ère iteration: i < N so we increment the counter —> i = 1. Is the Tier 1 power plant (Power Plant A) available? Yes, it is available. Is the remaining power to be distributed > 0? ^remaining 3 SWEET YES. We calculate the adjustment setpoint for the A control unit. 1.6 MW. We calculate the expected power in 30 seconds with this instruction. 0.5 x 1.6 = 0.8 MW. We update the remaining power to be distributed. ^remaining = ^remaining ~ ^30s(l) = 3 — 0.8 = 2.2 MW.

[0161] 2 è iteration: i < N so we increment the counter —> i = 2. Is the second-tier power plant (B power plant) available? No, it is not available. Therefore, she cannot participate and is assigned a zero instruction. P flex (2) = 0 MW. The remaining power to be distributed remains unchanged. ^remaining = 2.2 MW.

[0162] 3 è iteration: i < N so we increment the counter —> i = 3. Is the Tier 3 power plant (C power plant) available? No, it is not available. Therefore, she cannot participate and is assigned a zero instruction. ex (3) = 0 MW. The remaining power to be distributed remains unchanged: Postante = 2.2 MW.

[0163] 4 è iteration: i < N so we increment the counter —> i = 4. Is the Tier 4 power plant (Power Plant D) available? Yes, it is available. Is the remaining power to be distributed > 0? Prestante = MW dOHC OUL We calculate the adjustment setpoint for the D control unit. 1.2 MW. We calculate the expected power in 30 seconds with this instruction. 0.75 x 1.2 = 0.9 MW. We update the remaining power to be distributed. Present = Prestant ~ ^30s(4) = 2.2 — 0.9 = 1.3 MW.

[0164] 5 è iteration: i < N so we increment the counter —> i = 5. Is the Tier 5 power plant (Power Plant E) available? Yes, it is available. Is the remaining power to be distributed > 0? Prestante = 1.3 MW dOHC OUÏ. We calculate the adjustment setpoint for the E control unit. 1.3 MW. We calculate the expected power in 30 seconds with this instruction. 1.3 = 1.3 MW. We update the remaining power to be distributed. Present = Prestante ~ ^30s(3) = 1-3 — 1.3 = 0 MW.

[0165] 6 è iteration: i = N — > end of loop.

[0166] The adjustment commands calculated in this example to distribute the initial hydroelectric power command are summarized in Table 2. [Table 2]

[0167] These instructions are transmitted to the hydroelectric power plants so that they can supply the power required to ensure the activation of the reserve within 30 seconds.

[0168] In parallel with sending these instructions to the hydroelectric power plants, the battery reacts to the frequency variation by delivering its maximum power almost instantaneously. Indeed, according to the battery regulation law P re f bat = ~ K f ~ fo) + P Obat When the frequency f drops significantly, the term P re f batincreases sharply in response to this frequency variation.

[0169] The centralized controller constantly monitors the evolution of the overall hydroelectric power output in response to the commands sent. The goal is to reach the required overall power output in less than 30 seconds, but without exceeding the overall setpoint at the virtual power plant level. This is achieved thanks to the power margin P. mar Given that it is taken into account in the hydroelectric setpoint allocation algorithm, it is likely that the aggregate response of the hydroelectric power plants exceeds even the overall setpoint P re f hydro at a certain point. If the variation in total hydroelectric power exceeds the overall setpoint P r ef hydro, then the centralized controller can send adjustment instructions to the battery so that it reduces its power according to the overshoot on the hydroelectric side, so that the overall power of the virtual power plant is properly regulated.

[0170] The reserve contribution preserves hydro Hydroelectric power plants are monitored by calculating at each instant the difference between the total measured hydroelectric power, that is, the sum of the powers P hy dro ( ) of each hydroelectric power plant and the overall setpoint outside of setting P Ovpp , according to the following formula:

[0171] Controlling the battery adjustment setpoint P Obat can be adjusted to compensate for power fluctuations from all the power plants hydroelectric aggregate. An example of a method applicable to real-time control of the battery adjustment setpoint is described in patent application FR3133714A1.

[0172] According to this example of a process, the instruction P Obat can thus be calculated using the following formula:

[0173] Assuming that the power P bat The battery complies with the instruction P re f bat That is, assuming the battery is not fully charged, then

[0174] On the other hand, when the battery is saturated with power and cannot comply with the setpoint P re f bat , Pbat * P re f bat and therefore P vpp * P Ovpp - K f - f0).

[0175] Since P r ef bat is a function of P Obat, which in turn is a function of the hydroelectric power dPhydro( ), the activation of a hydroelectric power reserve according to the proposed technique can make it possible to obtain a setpoint for adjusting the battery P Obat such that the battery is not saturated with power. Thus, calculating the adjustment setpoints of hydroelectric power plants according to the proposed technique can make it possible to provide the additional primary reserve to avoid, or at least limit, the occurrence of situations where the battery is saturated with power.

[0176] A possible simulation of the implementation of the algorithm in Figure 1, in an example of implementation, was carried out with the same parameters as those presented in the example in Tables 1 and 2 where a hydroelectric power reserve is activated upwards.

[0177] The result of the simulation is shown in figures 3 to 6.

[0178] Figure 3 shows, from the instant following the frequency drop (33), the result (32) of the calculation of the power target P cibie = 12 MW and the results (31) of the calculation of the adjustment setpoints for hydroelectric power plants: P / to (l) = l.6MW, ex (2) = 0 MW, ex (3) = 0 MW, PflexW = 1.2 MW, and ex (5) = 1.3 MW.

[0179] Figure 4 shows the evolution of their power outputs after the instructions were sent to the requested hydroelectric power plants.

[0180] Power plants B and C remain (42, 43) at constant power because their adjustment setpoint is zero since they were indicated as unavailable to participate in the reserve.

[0181] Power plant A provides (41) about 0.8 MW of power 30 seconds after the triggering event (i.e., at time t = 10 + 30 = 40 s), which is consistent with the estimate made in block 19 of the algorithm in Figure 1.

[0182] Plant D provides (44) about 0.9 MW of power 30 seconds after the triggering event (i.e., at time t = 10 + 30 = 40 s), which is consistent with the estimate made in block 19 of the algorithm in Figure 1.

[0183] Power plant E provides (45) approximately 1.3 MW of power 30 seconds after the triggering event (i.e., at time t = 10 + 30 = 40 s), which is consistent with the estimate made in block 19 of the algorithm in Figure 1.

[0184] Figure 5 shows the evolution (51) of the total power of the hydroelectric power plants, which reaches and then exceeds the desired reserve complement (2 MW) in a time of less than 30 seconds, in particular thanks to the margin parameter.

[0185] Figure 5 also shows the initial saturation (53, 54) of the battery power (power limited to 10 MW while the setpoint rises to 12 MW) as well as the variation (52) of the adjustment term P Obat This value, which fluctuates according to the total power output of the hydroelectric plants, reaches a minimum of -4 MW. From this point onward, the battery's capacity drops below its maximum capacity by approximately 2 MW. This 2 MW corresponds to the 2 MW overshoot of the total power output of the hydroelectric plants, ultimately resulting in a variation of +4 MW for a demand of +2 MW.

[0186] Finally, Figure 6 shows the total power of the virtual power plant, that is, the sum (61) of the powers of the hydroelectric plants and the battery, and the overall setpoint Pref^ (62) of the virtual power plant. It can be seen that the overall setpoint is not reached immediately. This phase corresponds to the moment when the battery saturates at its full power and the hydroelectric plants respond more slowly. The overall response time is well under 30 seconds thanks to the proposed control strategy.

[0187] Figures 3 to 6 illustrate a battery saturation sequence following a frequency drop, an initial distribution of power plant adjustment commands Tl hydroelectric to provide a reserve supplement in less than 30 seconds, an overshoot of the hydroelectric contribution after a certain time, then a decrease in battery power to compensate for the overshoot of the total hydroelectric power in order to correctly follow the overall setpoint at the virtual power plant level.

[0188] This initial phase therefore meets the dynamic criterion of reserve release within 30 seconds. However, over time, this distribution is not optimal because the hydroelectric power plants find themselves in a situation of over-stress.

[0189] It is desirable to minimize the demand placed on hydroelectric power plants while ensuring the service provided meets the required standards. Limiting the demand on hydroelectric power is particularly necessary to minimize the feedback from level control functions, which can counteract the adjustment setpoint and thus cancel, at least partially, the reserve provided. The feedback from a level control function of a hydroelectric power plant is all the more significant as the demand P flex The size of this hydroelectric power plant is important.

[0190] Reference is now made to Figures 7 to 10, each of which details a flowchart of a possible algorithm suitable for implementing the regulation phase of the hydroelectric power reserve, or "second phase," in examples of the proposed technique. This second phase follows the first phase, or initial phase of activation of the hydroelectric power reserve, as described in Figures 1 to 6.

[0191] During the second phase, the power adjustment instructions for the hydroelectric plants are controlled in a fine and dynamic manner, in order to activate just enough reserve to provide the necessary hydroelectric demand to complement the demand on the battery, while limiting the over-demand on the hydroelectric plants.

[0192] This regulatory phase comprises two parts.

[0193] During the first part of the regulation phase, the adjustment instructions P fiex Hydroelectric power plants are reduced in size when there is over-activation of the reserve relative to demand. The first part of the regulation phase begins once the power outputs of the hydroelectric power plants have stabilized following the initial sending of commands at the beginning of the initial activation phase of the hydroelectric power reserve. For example, the first part of the phase of regulation can begin approximately one to two minutes after the frequency event that triggers the alert or emergency state, depending on the dynamic characteristics of the hydroelectric plants.

[0194] During the second part of the regulation phase, the adjustment instructions P fiexHydroelectric power plants are adjusted to compensate for the effects of level regulation functions which, after a certain time, begin to counteract the reserve provided by hydroelectric power plants.

[0195] As a reminder, the reserve contribution Pr ré cseerrv t / ee h*y „dro of all hydroelectric power plants can be obtained by applying the following formula: where Phydro ) is a measure of the sum of the powers of hydroelectric power plants.

[0196] Alternatively, each power measurement P hy dro (P) can be deduced from a measurement of an environmental parameter of plant i, such as a measurement of water flow at plant i.

[0197] Figure 7 presents a possible set of instructions that can be executed by the central controller during the first part of the regulation phase. The scenario considered is that of an initial frequency drop that has led to an upward activation of reserve capacity, and where the reserve supplied by the hydroelectric power plants exceeds, or is about to exceed, the initial demand. Executing the instructions in Figure 7 reduces all or part of the adjustment commands for the hydroelectric power plants in order to avoid, or at least limit, an overactivation of the hydroelectric power reserve.

[0198] During this first part of the regulation phase, it is desirable that the power adjustment setpoints P ftex hydroelectric power plants should be gradually reduced to both limit over-activation of the reserve and ensure grid stability.

[0199] A power margin logic is applied here, so as not to activate the reduction of the P setpoints fiex that if the overactivation exceeds a power margin Pmarge, in order to maintain a stabilized power level. Thus, the hydroelectric power reserve is maintained at a level corresponding to the sum of the overall setpoint P r ef hydro and the power margin P mar e-

[0200] This margin allows for consideration of the risk that one or more hydroelectric power plants may not be able to precisely follow the adjustment instructions P fiex after their reduction. Without this margin, there is a risk that the battery will become saturated again if the total power supplied by hydroelectric plants is insufficient to meet demand. The margin P marge is fixed at a value such that, even in the event of variation in the response of the power plants, the hydroelectric power reserve remains high enough to stabilize the network and avoid overloading the battery.

[0201] At block 71, a triggering condition is met. In the scenario of Figure 7, the triggering condition is the presence of an upward reserve surplus. This condition is met when the overall setpoint P r ef hydro is positive and Zi Preserves hydro (i) ~ P ref hydro > Pmarge - As long as the condition is not met, no subsequent instruction is implemented. The check provided in block 71 can be repeated over time.

[0202] In the following description of the actions implemented in blocks 72 to 80, we are in a scenario where, at a current moment, the triggering condition is verified in block 71 and is fulfilled.

[0203] The algorithm initialization is implemented in block 72. A counter i is initialized: i = N + 1. A variable P resta nte is initialized according to the following formula:

[0204] In block 73, the value of counter i is compared to a value of a stopping criterion, in this case 1.

[0205] If i = 1, then the algorithm has already considered, in turn, all the hydroelectric power plants of the virtual power plant at the current time, and the algorithm's stopping criterion is met. The calculation of the power adjustment setpoints P fiex is stopped at block 75 and is considered finished for the current time.

[0206] If i > 1, then the algorithm has not considered all the hydroelectric plants in the virtual power plant at the current time. The value of counter i is incremented by one in the opposite direction at block 74, and then the hydroelectric plant with rank i is considered at block 76.

[0207] In this way, the Nth-rank hydroelectric plant, the last in the priority order, is considered first in the first iteration where i = N, and the plant hydroelectric of rank 1, the first in the order of priority, is considered last in the last iteration, where i = 1.

[0208] The value of the total Posting power to be distributed is checked at block 76.

[0209] If the value of Postante is equal to zero, that is, if there is no current need for modification of the hydroelectric power reserve, then block 77 plans to maintain the setpoint P fiex (i) previously allocated to the hydroelectric power station of rank i at an unchanged value. The algorithm continues its operation at block 73.

[0210] If the value of Prestante is greater than zero, that is, if there is still a need at the current time for modifying the hydroelectric power reserve, then the value of the setpoint Pfi ex (i) previously allocated to the rank i hydroelectric power station is verified in block 78.

[0211] If the instruction Pfi ex (i) is already zero, so block 77 is designed to maintain it at zero. The algorithm continues its operation in block 73.

[0212] If the instruction ex(i) is positive, so it is planned in block 79 to reduce P ex (i) by applying a corrective term AP to it. AP corresponds to all or part of Postante, without exceeding P ex (i). In other words, AP = min (P fiex (i), Prestante )■

[0213] The Postante variable is updated in block 80 by subtracting the AP correction term, then the algorithm continues its operation in block 73.

[0214] Similarly, Figure 8 presents a possible set of instructions that can be executed by the central controller, in an example embodiment, to adjust the power adjustment setpoints of each hydroelectric power plant during the first part of the regulation phase. The scenario considered is that of an initial frequency increase that has led to a downward reserve activation, and where the downward reserve provided by the hydroelectric power plants exceeds, or is about to exceed, the initial demand. Executing the instructions in Figure 8 has the effect of increasing all or part of the hydroelectric power plant adjustment setpoints in order to avoid, or at least limit, an overactivation of the downward hydroelectric power reserve.

[0215] The instructions shown in Figure 8 are identical to those in Figure 7, with the following exceptions.

[0216] These exceptions are due to the fact that, in Figure 8, the aim is to increase hydroelectric power by distributing positive power to compensate for a decreasing reserve surplus.

[0217] In block 71, the verified triggering condition is the presence of a surplus reserve on the downward side. This condition is met when the overall setpoint P r ef hydro is negative P r ef hydro have negative values, and P marge has a positive value.

[0218] In block 72, the variable P res aunt is initialized according to the formula:

[0219] In block 76, the outcomes of the verification of the total power value Prestante to be distributed are as follows: either Prestante is equal to 0, in which case at block 77 there is no change in the adjustment setpoint Pf tex( ) is not implemented, either Prestante is less than 0, in which case at block 78 the value of the adjustment setpoint Pfiex ) is verified.

[0220] In block 79, the value of the AP correction term is AP = max P fi ex ( ), Poster)-

[0221] Reference is now made to figures 9 and 10, which describe the second part of the regulation phase.

[0222] As a reminder, the power setpoint P r ef hydro of a hydroelectric power plant participating in a primary frequency regulation service is composed of two terms: r pref fhydro = r prgz + T r pf f the ex-

[0223] The term P rgz corresponds to the power plant's level regulation. The term P ftex, non-zero, is the power adjustment setpoint corresponding to the activation of a power reserve. This reserve is activated upwards or downwards depending on the sign of Pfi ex -

[0224] The initial application of an adjustment instruction P ex (i) non-zero to a hydroelectric power station shifts the operating point of the power station, increasing or decreasing the outflow from the reservoir, which results in a change in level.

[0225] From the perspective of the power plant's level regulation, the activation of the reserve, whether increasing or decreasing it, is seen as a disturbance. The term P rgz tends to vary in the inverse direction of P fiex in order to reduce or eliminate this disruption.

[0226] In the event of a frequency drop that triggers an upward activation of reserve capacity, after a certain period, the level control functions reduce the power output of the hydroelectric power plants. In this situation, the control system can compensate for this decrease by increasing the adjustment setpoints up to the limit of available power.

[0227] Figure 9 presents a possible set of instructions that can be executed by the central controller during the second part of the regulation phase. The scenario considered is that of an initial frequency drop that has led to an upward activation of reserve capacity, and where the level regulation dynamics of the hydroelectric power plants are causing, or are about to cause, a drop in hydroelectric power output that prevents compliance with the initial setpoints. Executing the instructions in Figure 9 has the effect of increasing all or part of the adjustment setpoints of the hydroelectric power plants to maintain their collective power response at the desired level, that is, at the level of expected contribution to the virtual power plant's reserve capacity.

[0228] These instructions can be repeated over time, for example at regular intervals of one or more seconds, in order to compensate for power drops related to level regulation functions, throughout changes in P rgz which continue as long as the values ​​of P fiex are positive.

[0229] At block 81, a triggering condition is met. In the scenario of Figure 9, the triggering condition is the presence of an insufficient reserve on the upside. This condition is met when the overall setpoint P r ef hydro is positive and iPreserves hydro (i) ~ P ref hydro < °- As long as the condition is not met, no further instructions are implemented. The check provided for in block 81 can be repeated over time.

[0230] In the following description of the actions implemented in blocks 82 to 92, we are in a scenario where, at a current moment, the triggering condition is verified in block 81 and is fulfilled.

[0231] The algorithm initialization is implemented in block 82. A counter i is initialized: i = 0. A Posting variable is initialized according to the following formula:

[0232] In block 83, the value of counter i is compared to a value of a stopping criterion, in this case N.

[0233] If i = N, then the algorithm has already considered, in turn, all the hydroelectric power plants of the virtual power plant at the current time, and the algorithm's stopping criterion is met. The calculation of the power adjustment setpoints P fiexThe process stops at block 85 and is considered complete for the current time. A delay is set at block 86. After the delay expires, the instructions are repeated, starting again at block 81.

[0234] If t < A, then the algorithm has not considered all the hydroelectric plants in the virtual power plant at the current time. The value of counter i is incremented by one in block 84, and then the availability of the hydroelectric plant with rank i is considered in block 87.

[0235] In this way, the first-rank hydroelectric plant, the first in the order of priority, is considered first in the first iteration where i = 1, and the N-rank hydroelectric plant, the last in the order of priority, is considered last in the last iteration, where i = N.

[0236] If the first-order control unit is unavailable, then block 89 is scheduled to maintain the setpoint Pfiex (i) previously allocated to this power plant at an unchanged value. The algorithm continues its operation at block 83.

[0237] If the rank i power plant is available, the value of the total Posting power to be distributed is checked at block 88.

[0238] If the value of Postante is equal to zero, that is, if there is no current need to modify the hydroelectric power reserve, then block 89 stipulates that the Pfi setpoint should be maintained. ex (i) previously allocated to this power plant at an unchanged value. The algorithm continues its operation at block 83.

[0239] If the value of Postante is greater than zero, that is, if there is still a need at the current time to modify the hydroelectric power reserve, then the value of the setpoint Pfi ex(i) previously allocated to the rank i hydroelectric power station is verified in block 90.

[0240] For the purposes of this verification, maximum values ​​of the adjustment setpoints of hydroelectric power plants, denoted Pfi ex are used. A value positive, defines the maximum power increase allowed for the rank i power plant during the hydroelectric power reserve regulation phase.

[0241] The Pfiex values ​​can be chosen to be equal to the Pfi values. eX init - Conversely, the Pfiex values ​​can be chosen to be different from the Pfi values eX init -

[0242] If the instruction Pfi ex (P) is not less than the value P fiex (P), therefore, it is planned in block 89 to maintain its value unchanged. The algorithm continues its operation in block 83.

[0243] If the instruction Pfi ex(P) is less than the value P fiex (P), so it is planned in block 91 to increase P ex (i) by applying a corrective term AP to it. AP corresponds to all or part of Postante' without exceeding the difference between P fiex (P) and Pfiex(P)-

[0244] In other words,

[0245] The Postante variable is updated in block 92 by subtracting the AP correction term, then the algorithm continues its operation in block 83.

[0246] Similarly, Figure 10 presents a possible set of instructions that can be executed by the centralized controller, in an example embodiment, to adjust the power adjustment setpoints of each hydroelectric power plant during the second part of the regulation phase. The scenario considered is that of an initial frequency increase that has led to a downward activation of reserve capacity, and where the level regulation dynamics of the hydroelectric power plants are causing, or are about to cause, a rise in hydroelectric power that prevents compliance with the initial setpoints. Executing the instructions in Figure 10 has the effect of lowering all or part of the adjustment setpoints of the hydroelectric power plants to maintain their collective power response at the desired level, that is, at the level of expected contribution to the virtual power plant's reserve capacity.

[0247] The instructions shown in Figure 10 are identical to those in Figure 9, with the following exceptions.

[0248] These exceptions are due to the fact that, in Figure 10, it is a matter of reducing hydroelectric power by distributing negative power to compensate for a lack of reserve on the downside.

[0249] In block 81, the verified triggering condition is the presence of a lack of reserve on the downward side. This condition is met when the overall setpoint P r ef hydro is negative

[0250] In block 82, the variable P res aunt is initialized according to the formula:

[0251] In block 88, the outcomes of the verification of the total Posting power value to be distributed are as follows: either Prestante is equal to 0, in which case at block 89 there is no change in the adjustment setpoint P fiex(p) is not implemented, either Prestante is less than 0, in which case at block 90 the value of the adjustment setpoint Pfiex( ) is verified.

[0252] In block 90, the outcomes of the verification of the adjustment setpoint value Pfiex are as follows: either the instruction ex (i) is not greater than a value Pf tex (i), in which case it is provided in block 89 that its value remains unchanged and the algorithm continues its operation in block 83, either the Pfi instruction ex ( ) is greater than the value Pf tex (p), in which case it is planned in block 91 to reduce P ex (i) by applying a corrective term AP = max (P flex (p) - Pflex (p)> Prestante^ ■

[0253] Minimum values ​​for the adjustment setpoints of hydroelectric power plants, denoted Pfiex min are used here. A value P fiex(i), negative, defines the maximum power reduction allowed for the rank i power plant during the hydroelectric power reserve regulation phase.

[0254] The negative AP correction term is subtracted from the Posting variable in block 92.

[0255] A possible simulation of the implementation of the algorithms in Figures 7 and 9, in an example of implementation, was carried out with the same parameters as those presented in the example in Tables 1 and 2 where a hydroelectric power reserve is activated upwards.

[0256] The result of the simulation is shown in figures 11 to 14.

[0257] This simulation shows the continuation of the virtual power plant's behavior during and after the reserve activation phase as illustrated in figures 3 to 6.

[0258] The frequency drop of the network from the nominal frequency of 50 Hz to a frequency of 49.8 Hz takes place at time t = 10 s.

[0259] In this illustrative example, only control units 1, D, and E are available to participate in primary frequency tuning. Control units B and C are unavailable.

[0260] Figure 11 shows the time evolution (111) of the power setpoint Pfiex ) transmitted to the local controller of power plant A and the time evolution (112) of the power P hy dro( ) produced by power plant A.

[0261] Figure 12 shows the time evolution (121) of the power setpoint Pfiex P ) transmitted to the local controller of the power plant D and the time evolution (122) of the power P hy dro(P') produced by plant D.

[0262] Figure 13 shows the time evolution (131) of the power setpoint Pfiex E) transmitted to the local controller of the power plant E and the time evolution (132) of the power P liy d ro (E) produced by power plant E.

[0263] These temporal changes illustrate: the activation phase of the adjustment instructions in order to have a rapid activation of the reserve (to respect the dynamic criterion of 30 seconds), Then comes the phase of reducing the initial setpoint to reduce the excess reserve upwards (here, this refers to lowering the Pfi setpoint). ex E) at time t = 100 s), and the drop in hydroelectric power linked to the level control functions of the power plants: this can be seen here through the drop in P hy dro( ) and P hy dro(P') between t = 100 s and t = 500 s.

[0264] The increase in adjustment instructions Pf iex G4), P fiex(P) and Pfiex E) at approximately t = 500 s aims to compensate for the first power drop related to the level regulation functions.

[0265] Then, the level regulation functions continue to act, with drops in pH y right A ) and P liy d ro ( D visible between t = 500 s and t = 900 s, which again leads to an increase in Pfi ex E) just before t = 900 s to compensate for these new drops.

[0266] Finally, the overall operation of the virtual power plant in the simulation is illustrated in Figure 14, which represents, as a function of time, the frequency (141) of the network, the total hydroelectric power (142), the power (143) of the battery, and the overall power (144) of the virtual power plant.

[0267] This simulation confirms that the proposed technique allows, in a situation where the virtual power plant is subjected to an overall power setpoint exceeding an operational battery power limit: to reach this overall power target within a given timeframe, here less than thirty seconds, and to then maintain temporal stability of the overall power of the virtual power plant in accordance with the overall power setpoint.

[0268] This expression means that the control entity continuously and dynamically adjusts the energy output of all the power plants constituting the virtual power plant (battery and other energy sources) to ensure that the total power delivered consistently corresponds to the overall setpoint. This implies avoiding excessive fluctuations in the power delivered over time.

[0269] Compared to conventional control techniques for electrical production equipment, involving for example proportional-integral type controllers or any other control technique based on linear automation, the proposed technique is more robust for application cases where the electrical network is placed in a state of alert or emergency.

[0270] Indeed, the proposed technique is adapted to means of production having very different, and potentially non-linear, dynamic characteristics, for which the classic control techniques of linear automation do not work correctly and would risk causing network instabilities.

[0271] This disclosure is not limited to the examples described above, which are merely examples, but encompasses all the variations that a person in the trade may consider in the context of the protection sought.

[0272] For example, although the described implementation examples focus mainly on run-of-river hydroelectric power plants, the proposed technique can be applied to other types of power plants with a sufficiently fast response capability to cooperate with a battery in such a way as to provide a frequency regulation service.

[0273] For example, although the described implementation examples mention the use of a fixed priority order for allocating adjustment commands among hydroelectric power plants, a flexible allocation is also possible. For example, priorities can be dynamically adjusted based on instantaneous availability. power plants, their capacity to meet demand, or even depending on weather conditions impacting power plant production.

[0274] The provision of power adjustment instructions can also take into account data from external sources, such as grid demand forecasts or weather forecasts. This data can be integrated into the algorithms described to refine the adjustments of the hydroelectric power plants and the battery, in order to anticipate reserve or frequency regulation requirements.

[0275] In scenarios where the battery reaches its operational power limit, the algorithm can predict specific corrective measures to more quickly redistribute power to hydroelectric plants or other available units. Controlled overload mechanisms for hydroelectric plants can also be integrated to address battery saturation for short periods.

[0276] Mechanisms for optimizing response times for adjustment commands can be integrated to further improve the temporal performance of the virtual power plant. For example, the time required to provide power adjustment commands can be automatically adjusted based on the actual dynamics of each power plant at the time of the incident.

[0277] The proposed technique can also be adapted for hybrid networks that combine synchronous and asynchronous networks, taking into account the specific characteristics of frequency regulation in these contexts. Such an adaptation involves adjustments to the algorithm to calculate the adjustment setpoints in order to balance these different parts of the network.

[0278] Machine learning algorithms can be introduced to monitor and adjust the response of hydroelectric power plants and batteries in real time, based on past performance and anticipated trends. This type of predictive monitoring allows for better adaptation of adjustment instructions to the dynamics of the power plants.

[0279] It is also possible to provide for a dynamic adjustment of the distribution of adjustment instructions according to failures or sudden unavailability of certain hydroelectric power plants, using redundancy mechanisms within the aggregate to ensure the continuity of the frequency regulation service.

Claims

Demands

1. A method for controlling, implemented by a control entity, a virtual power plant connected to an electrical grid, the virtual power plant comprising at least one battery and a set of hydroelectric power plants, the virtual power plant being subject to an overall power setpoint exceeding an operational power limit of the battery, the method comprising: the provision of a first power adjustment command to a local controller of a hydroelectric power plant within the group of hydroelectric power plants, the first adjustment command taking into account a priority order of the hydroelectric power plants and the operational power limit of the battery, and the first adjustment command not taking into account any power measurement; and after the provision of the first power adjustment command, the provision of a second power adjustment command to the local controller of the hydroelectric power plant, the second power adjustment command taking into account the overall power measurement of the group of hydroelectric power plants, in which, when the second power adjustment setpoint is provided, the battery is subjected to a third power adjustment setpoint which takes into account the overall power setpoint and the overall power measurement of the entire hydroelectric power plant complex, and in which the supplies of the adjustment setpoints contribute to maintaining a temporal stability of an overall power of the virtual plant in accordance with the overall power setpoint.

2. A method according to the preceding claim, wherein the first power adjustment setpoint takes into account at least one parameter of the hydroelectric power plant from among: availability, an available capacity, a response time, a power limitation.

3. A method according to any one of the preceding claims, wherein the first power adjustment setpoint takes into account an estimated power regulation gain of the hydroelectric plant after a first delay, and the supply of the second power adjustment setpoint is implemented after a second delay that is longer than the first delay.

4. A method according to any one of the preceding claims, wherein the second power adjustment setpoint is periodically updated based on a fluctuation in an overall power measurement of the hydroelectric plants and / or a fluctuation in an environmental parameter of at least one hydroelectric plant in the set of hydroelectric plants.

5. A method according to any one of the preceding claims, wherein: The overall power setpoint is adjusted, if necessary, after the first adjustment setpoint is provided and before the second adjustment setpoint is provided, and The second adjustment instruction takes into account the overall adjusted power instruction.

6. A method according to any one of the preceding claims, wherein the second adjustment instruction includes compensation for at least one unforeseen power variation of at least one hydroelectric power plant in the set of hydroelectric power plants.

7. A method according to any one of the preceding claims, wherein a first decision tree is used to determine the first adjustment instruction.

8. A method according to any one of the preceding claims, wherein a second decision tree is used to determine the second adjustment instruction.

9. A method according to any one of the preceding claims, comprising providing a set of first power adjustment instructions to local controllers of a plurality of hydroelectric power plants in the hydroelectric power plant set, the set of first adjustment instructions taking into account an order of priority and not taking into account any power measurement, and providing the set of first adjustment instructions including the provision of the first adjustment instruction.

10. A method according to any one of the preceding claims, comprising a verification of the hydroelectric plant's ability to respond to the first adjustment setpoint or the second adjustment setpoint.

11. A method according to any one of the preceding claims, wherein the second adjustment instruction takes into account a comparison between a measurement of power of the hydroelectric plant after the provision of the first adjustment setpoint and a target value.

12. A method according to any one of the preceding claims, wherein the decision tree adjusts the adjustment instructions of the hydroelectric power plants according to an iterative analysis, until the overall power reaches a stable value within a defined tolerance range.

13. A computer program comprising instructions for implementing the method according to any one of the preceding claims when this program is executed by a control entity of a virtual power plant connected to an electrical network, the virtual power plant comprising at least one battery and a set of hydroelectric power plants.

14. Control entity of a virtual power plant connected to an electrical grid, the virtual power plant comprising at least one battery and a set of hydroelectric power plants, the virtual power plant being subject to an overall power setpoint exceeding an operational battery power limit, the control entity being configured to implement: the provision of a first power adjustment command to a local controller of a hydroelectric power plant within the group of hydroelectric power plants, the first adjustment command taking into account a priority order of the hydroelectric power plants and the operational power limit of the battery, and the first adjustment command not taking into account any power measurement; and after the provision of the first power adjustment command, the provision of a second power adjustment command to the local controller of the hydroelectric power plant, the second power adjustment command taking into account an overall power measurement of the group of hydroelectric power plants, in which, when the second power adjustment setpoint is provided, the battery is subjected to a third power adjustment setpoint which takes into account the overall power setpoint and the overall power measurement of the entire hydroelectric power plant complex, and in which the supplies of the adjustment setpoints contribute to maintaining a temporal stability of an overall power of the virtual plant in accordance with the overall power setpoint.

Images