Joint programming of production and consumption flexibilities

Abstract

The invention relates to a method for the coordinated management, over a period of time, of production and consumption units. The method comprises, during a programming phase: - receiving (S1) consumption simulations of each consumption unit covering the time period and parameterized by a nominal transition probability characterizing the probability of moving from one state to another, - determining (S2), based on the simulations, a target average consumption of the consumption units and a control signal, - transmitting (S3), to each consumption unit, the control signal to update (S4) the nominal transition probability based on the received control signal and a local function characterizing the operating specificities of the consumption unit, and - transmitting (S3), to each production unit, a production program consistent with the target average consumption. Abstract Figure: Figure 2.

Description

Title of the invention: Joint programming of production and consumption flexibilities technical field

[0001] The field of the invention relates to the coordinated programming and management of production and consumption flexibilities, in particular the control of consumption units, supplied with electrical energy by production units, capable of modulating their consumption at the request of a flexibility operator. Previous technique

[0002] Consumption flexibility is the ability of a consumption unit, for example in the industrial sector, to modulate its electricity consumption at the request of an operator, called a flexibility operator. The operator can thus remotely control the operation of industrial or domestic equipment at the level of one or more consumption units or delay such operation.

[0003] A demand from the flexibility operator to a consumption unit may be a request to reduce its electricity consumption during periods of high demand on the electrical system. The consumption unit may then respond by shutting down industrial equipment for a few minutes or even several hours. This is known as demand response. Conversely, the flexibility operator may also request an increase in electricity consumption to absorb excess electricity production.

[0004] The programming and management of consumption flexibility thus aim to reduce or take advantage of an imbalance between the supply and demand of electrical energy within a system comprising consumption units supplied by production units. The challenge of consumption flexibility is therefore to achieve savings in electrical energy, optimize energy performance, and more generally, enable such a system to adapt to fluctuations in electricity production and consumption.

[0005] Today, consumption flexibilities are sometimes ignored in favor of other issues related to the scheduling of production units, such as sizing or taking into account the technical constraints of the means of production. When consumption flexibilities are taken into account, they are modeled by a virtual storage system that aggregates all the flexibilities without distinguishing between the individual constraints of the different pieces of equipment in the units. consumption. Automatic control of consumption units can also be implemented, for example, on a day-to-day basis, by determining a forecast program provided in advance to each consumption unit, which is then controlled in real time according to the setpoint indicated in this forecast program.

[0006] Generally speaking, a distinction is usually made between a local approach, i.e., an approach from the point of view of a consumption unit managing the various industrial equipment in real time to reduce their electricity consumption, and a global approach, in which the flexibility operator aggregates the flexibilities offered by all the consumption units to optimize overall electricity consumption.

[0007] However, such methods do not guarantee that a target consumption determined by the flexibility operator is achievable by all consumption units. Furthermore, predictive models that propose real-time updates to a forecast schedule and recalculation of the consumption setpoint at each event are very resource-intensive and computationally time-intensive. Summary

[0008] The present invention improves the situation.

[0009] In this respect, the present invention relates to a method, implemented by computer means, for the coordinated management, over a predetermined period of time sampled into a succession of instants, of a set of one or more production units and a set of one or more consumption units supplied with electrical energy by the set of one or more production units. The method comprises, during a programming phase: - receive at least one consumption simulation for each consumption unit, each consumption simulation of a consumption unit covering the time period and being parameterized by a nominal transition probability characterizing the probability of the consumption unit switching, at a given instant of the time period, from one operating state to another, - determine, based on the received consumption simulation(s), a target average consumption for one or more consumption units and a control signal sampled over the time period, - transmit, to each consumption unit, the control signal to update, at the level of each consumption unit, the nominal transition probability based on the received control signal and a local function characterizing the operating specifics of the consumption unit considered, and - transmit, to each production unit, a production program over the period of time consistent with the target average consumption.

[0010] In one or more embodiments, the production program includes at least, for each production unit, a production profile determined jointly with the target average consumption and the control signal.

[0011] Advantageously, the sampled control signal is determined during the programming phase according to an iterative algorithm defined as follows: [Math.l] V hL ■ ,P x U + L , L AC UA p\ K ij - p .y pjy i ' l - ] q = ] p \ K / / \ .' J kJ with : [Math.2] { T \ 1 / 0( v) + E Æ=04(^)^)} [Math.3] = argminj (x J ) - £} \ JJ \ 7 \ fi [Math.4] . exp ( - LL 0U 00+ YW "(A) " E ..exp ( - E * = o 0 p M+ Y^U'' '"('t)) [Math.5] AND v / VM i, q ] \ To q / \ , _ , 4 = U H Or : - Xp(tk) is the value at time tk of the control signal at iteration p, the control signal being initialized to a predetermined initial control signal Xo; - pp is a positive real number; -fo is a function that penalizes a difference between a production profile of one or more production units and a consumption profile of one or more consumption units; - T is the number of moments in the succession of moments in the predetermined time period; ​- VP is a failure profile at iteration p, Vp(tk) is the value at time tk of the failure profile; - V is a real number characterizing a zero energy deviation constraint of the set of one or more consumption units with respect to a nominal consumption; - n is the number of production units; - Ai is a set of technical constraints constituting the set of states techniques in which the j-th production unit can operate; - Pj(xj) is the production profile of the j-th production unit when it operates in a technical state xi, Pj(xj)(tk) being the value at time tk of the production profile Pj(xJ); -fj(xj) is a production cost of the j-th production unit when it is operating in a technical state x1; - N is the number of consumption units; - M is the number of consumption simulations for each consumption unit; - k is a predetermined constant; - exp(-) is the exponential function; and - U^tk) is the value at time tk of the consumption of the i-th consumption unit according to the q-th consumption simulation, the iterative algorithm being interrupted when a predefined stopping criterion is met.

[0012] The stopping criterion is defined, for example, as follows: [Math.6] It v - E" (?) + E* || < e 1 PJ - L j \ P / 1 = ! Q = 1 P Or : -()-(1 is a norm; and - e is a real positive.

[0013] Typically, the initial control signal used to initialize the control signal prior to the implementation of the iterative algorithm is the control signal obtained at the end of a period of time preceding the period of time considered.

[0014] Advantageously, the control signal determined at the end of the iterative algorithm is defined as follows: [Math.7] = A * p where p* is the rank of the iteration, among all the iterations up to the last iteration satisfying the stopping criterion, for which a duality jump is minimal, i.e.: [Math. 8] p* = argmin { Gap{p)} p with : [Math.9] Gap(p) = f Q [ E " = - E* =) E - E .p (x-jit k ) + E ,E (b)iy E t n E >E -4j}u iti J = H y \ K / p \ KJ J ! l=\lj=\N)\K]

[0015] Advantageously, the target average consumption is defined as follows: [Math. 10] rW = L N \ fV Z mW [ ~ £ MW — | p\s Or : - r(tk) is the value at time tk of the average target consumption of the set of one or more consumption units; and - p* is the rank of the iteration, among all the iterations up to the last iteration satisfying the stopping criterion, for which a duality jump is minimal, i.e.: [Math. 11] p * = argmin { Gap ( p )} p with : [Math. 12] E " _ (O - £ '.A - / »(%) - £ LoV'"* 0 ',!' 1 ) - E "^(v» - E ",E ",«.j; V'Or / LL *, E "

[0016] In one or more embodiments, the nominal transition probability is Update as follows during the programming phase: [Math. 13] ^^dY^^Y) with : [Math. 14] V(l k . X) = J ^,(¾. X, dYW(h^. Y)exp (-# / U +l k(r)j [Math. 15] V'(t k .X) = ^P0(uX,dY)v\tk ^Y)exp (-^'(t kkl ):c,(Y^ Or : - Xj, X2 and Y are possible states of a consumption unit, dXh, dX2 and dY are the differentials of such states, - PodtbXbdXd is the nominal transition probability characterizing the probability of the i-th consumption unit to switch, at time tk, from one operating state X] to another operating state X2; - QoKthXiydXP is the nominal transition probability updated during the initial programming phase, characterizing the probability of the i-th unit of consumption of switching, at time tk, from one operating state X2 to another operating state X2; - V^X) is the value at time tk of the local function characterizing the operating specificities of the i-th consumption unit and technical characteristics of said equipment, when the i-th consumption unit operates in state X; - exp(-) is the exponential function; - N is the number of consumption units; - k is a predetermined constant; - )C(tk) is the value at time tk of the determined control signal; and - Ci being a function which, to a state of the i-th consumption unit, associates a consumption profile, Ci(Y) being the value of the consumption of the i-th consumption unit operating in state Y.

[0017]

[0018]

[0019]

[0020]

[0021] The local function depends, for example, on the programmed use of industrial equipment by the consumption unit, the technical characteristics of said industrial equipment, and local contingencies. In one or more embodiments, the method further comprises, following the programming phase, a management phase implemented over the time period including the real-time updating of the control signal according to the target average consumption; and the real-time updating of the nominal transition probability of each consumption unit based on the updated control signal. Advantageously, in such a case, the nominal transition probability is updated, for example, as follows during the management phase: [Math. 15] Or : 'X । dX2 ) is 'the nominal transition probability updated during the management phase characterizing the probability of the i-th consumption unit to switch, at time 4, from one operating state X] to another operating state - C( / k) is the value at time tk of the updated control signal. The management phase is implemented, for example, as follows: - to predict, at each instant of the time period, an average consumption at the next instant of the set of one or more consumption units as a function of the updated nominal transition probability of each consumption unit, - to update, at every instant, the control signal based on a comparison between the average consumption at the next predicted instant and the target average consumption, and - transmit, at every instant, the updated control signal to each consumption unit to update, at the level of each consumption unit, the nominal transition probability based on the updated control signal received.

[0022] Advantageously, the update of the control signal during the management phase is implemented as follows: - to construct a predictive consumption model that allows approximating, based on the control signal at a given moment, the average consumption predicted at the next moment, then, at each instant of the predetermined time period: - compare the actual average consumption with the average consumption forecast at that time for one or more consumption units, - adapt, through learning, at every moment, the predictive consumption model based on comparison, and - to update, at every instant, the control signal by reciprocal application of the predictive consumption model to the target average consumption aimed for at the next instant.

[0023] Advantageously, during the management phase implemented over the time period, each consumption unit operates according to the nominal transition probability updated in real time, while each production unit follows the received production schedule.

[0024] The invention further relates to a computer program, comprising instructions for implementing the process described above when the instructions are implemented by at least one processor.

[0025] Finally, the invention relates to a processing unit for the coordinated management, over a predetermined time period sampled into a succession of instants, of a set of one or more production units and a set of one or more consumption units supplied with electrical energy by the set of one or more production units. The processing unit is configured to, during a programming phase: - receive at least one consumption simulation of the consumption unit, each consumption simulation of a consumption unit covering the time period and being parameterized by a nominal transition probability characterizing the probability of the consumption unit switching, at a given instant of the time period, from one operating state to another, - determine, based on the consumption simulation(s) received, a target average consumption of one or more consumption units and a control signal sampled over the time period, - transmit, to each consumption unit, the control signal to update, at the level of each consumption unit, the nominal transition probability based on the received control signal and a local function characterizing the operating specifics of the consumption unit considered, and - transmit, to each production unit, a production program over the period of time consistent with the target average consumption. Brief description of the drawings

[0026] Other features, details and advantages will become apparent upon reading the detailed description below, and upon analysis of the accompanying drawings, on which:

[0027] [Fig-1] schematically illustrates a system comprising a set of one or several production units, a set of one or more consumption units and a processing unit according to the invention.

[0028] [Fig.2] illustrates a programming phase of the management process according to the invention; And

[0029] [Fig.3] illustrates a real-time management phase of the management process according to the invention. Description of the implementation methods

[0030] Fig. 1 schematically illustrates a SYS system.

[0031] The SYS system comprises a set of one or more production units UPi, UP2, UPn, a set of one or more consumption units UCI, UC2, UCN and a processing unit UNT.

[0032] In the following description, the SYS system is considered advantageously to comprise several production units, here a number n of production units UPi, UP2, UPn. Similarly, the SYS system is considered advantageously to comprise several consumption units, for example a number N of consumption units UCi, UC2, UCN.

[0033] In the context of the invention, the SYS system can be considered a mixed portfolio composed of conventional production resources, namely the production units UPi, UP2, UPn, and consumption flexibility, namely the consumption units UCi, UC2, UCN. Managing such a portfolio aims to limit the imbalance between the supply and demand of electrical energy within the SYS system in order to achieve electrical energy savings and adapt to fluctuations in electrical production and consumption.

[0034] The production units UPi, UP2, UPn are arranged to supply electrical energy to the consumption units Uci, UC2, UCN. In other words, the units Production units UPi, UP2, UPn are electrically connected to consumption units Uci, UC2, UCN within the SYS system. Each production unit can be controlled in real time or follow a predetermined schedule.

[0035] The operation of each production unit can be characterized by a production profile and a production cost. The production profile and the production cost depend on the technical state, also called the state or operating mode, in which the production unit operates. Thus, for each production unit, technical constraints can be defined that constitute the set of possible technical states such that, at any given moment, the production unit in question operates according to one of these technical states.

[0036] More specifically, the jth production unit can be characterized by a function Pj which, for each technical state x1 of the technical constraints Ai of the jth production unit, associates a production profile Pj(xJ). The production profile Pj(xJ) makes it possible to determine, at a given instant tk of the predetermined future time period, the electrical energy production pj(xj)(tk) of the jth production unit operating in technical state xj.

[0037] Furthermore, it is possible to define a production profile for one or more production units UPi, UP2, UPn as a function of the production profile of each production unit. The production profile for one or more production units UPi, UP2, UPn makes it possible to determine, at a given time period, the total electrical energy production of one or more production units UPi, UP2, UPn. This production profile therefore depends on the respective technical states of the production units UPi, UP2, UPn.

[0038] The consumption units UCi, UC2, and UCN are arranged to be supplied with electrical energy by the production units UPi, UP2, and UPn. Each consumption unit typically comprises one or more industrial or domestic devices whose operation requires an electrical power supply. Such devices might be, for example, an electric storage water heater, a refrigerator, or an air conditioning system. As explained previously, the consumption units UCi, UC2, and UCN are flexible. In other words, the consumption units UCi, UC2, and UCN are capable of modulating their respective electrical energy consumption. In the SYS system described here, the consumption units UCi, UC2, and UCN modulate their electrical energy consumption at the request of the processing unit UNT, also called the flexibility operator.

[0039] The operation of each consumption unit is characterized by a consumption profile and a nominal transition probability. The consumption profile depends on the operating state of the consumption unit. The probability of The nominal transition value characterizes the probability of a consumption unit transitioning, at a given instant, from one operating state to another. In real time, the operation of each consumption unit can be modeled by random behavior based on the nominal transition probability. The states of each consumption unit can be assumed to be mutually independent and follow a known probability distribution.

[0040] More specifically, the i-th consumption unit can be characterized by a function c; which, to each state or operating mode X' of the i-th consumption unit, associates a consumption profile ci(Xi). The consumption profile c; (X') makes it possible to determine, at a given instant tk of the predetermined future time period, the electrical energy consumption ci(Xi)(tk) of the i-th consumption unit operating in state X*.

[0041] Furthermore, it is possible to define a consumption profile for one or more consumption units UCi, UC2, UCN based on the consumption profile of each individual consumption unit. The consumption profile of one or more consumption units UCi, UC2, UCN allows us to determine, at a given point in time, the total electrical energy consumption of one or more consumption units UCi, UC2, UCN. This consumption profile therefore depends on the respective states or operating modes of the consumption units UCi, UC2, UCN.

[0042] The operating specificities of each consumption unit can also be described by a local function.

[0043] Such a local function depends, for example, on the programmed use of equipment in the consumption unit, the technical characteristics of this equipment, and local contingencies. Furthermore, as explained later in the description, at least part of the local function associated with each consumption unit is common to the respective local functions of the other consumption units.

[0044] At the request of the UNT processing unit, each consumption unit can reduce its electricity consumption during periods of high demand on the SYS system. This reduction in electricity consumption can be achieved by shutting down industrial or domestic equipment at the consumption unit level for a few minutes or even several hours. This is known as demand response. Conversely, each consumption unit can increase its electricity consumption to absorb excess electricity production.

[0045] In the context of the present invention, the aim is to determine, within the framework of managing consumption flexibilities, a programming of the units of UCI, UC2, and UCN consumption is measured over a predetermined time period. Specifically, the predetermined time period is sampled into a series of time intervals, so that the operation of each consumption unit is determined for each instant within that period. Typically, the predetermined time period represents several hours or even one or more days, while the time interval is one or more minutes. This programming is determined prior to the predetermined time period and aims to establish an achievable average consumption level for the UCI, UC2, and UCN consumption units.

[0046] Furthermore, the present invention also allows real-time management, more precisely from one instant of the predetermined time period to the next instant, of consumption flexibilities to adapt to local constraints and hazards and to ensure that the average target consumption calculated upstream is reached.

[0047] Furthermore, the present invention proposes the determination of a production program, intended for the production units UPi, UP2, UPn consistent with the electrical energy needs of the consumption units UCi, UC2, UCN.

[0048] It is therefore proposed to link programming, prior to the predetermined time period, with real-time management of consumption flexibilities. This link is made possible by the UNT processing unit.

[0049] Within the SYS system, the UNT processing unit is a flexibility operator and therefore has the role of managing the production and consumption flexibilities of the SYS system.

[0050] Therefore, the UNT processing unit is configured to jointly program the operation, over a predetermined sampled time period, of the production units UPi, UP2, UPn and the consumption units UCi, UC2, UCN. The processing unit can also be configured to manage, in real time, during the predetermined time period, the operation of the consumption units UCi, UC2, UCN.

[0051] More specifically, in the present case, the UNT processing unit is configured to determine, during a programming phase, a control signal common to the consumption units UCb, UC2, and UCN, as well as a target average consumption of the consumption units UCh, UC2, and UCN. This target average consumption also makes it possible to determine a coherent production schedule.

[0052] The UNT processing unit is also configured to transmit the control signal to each consumption unit in order to update, at the level of each consumption unit, the nominal transition probability based on the received control signal and the local function characterizing the operating specifics of the consumption unit in question. In particular, at least part of the function local associated with a consumption unit depends on the control signal received and is therefore common to all consumption units UCi, UC2, UCN.

[0053] The UNT processing unit is further configured to transmit to each production unit a production schedule over a time period consistent with the target average consumption. Such a production schedule thus allows the production units UPi, UP2, UPn to be coordinated with the target average consumption of the consumption units UCi, UC2, UCN over the coming time period and to adapt their production accordingly to ensure a balance between the supply and demand of electrical energy.

[0054] As explained previously, the operation of a consumption unit is characterized by its nominal transition probability of moving, at a given instant, from one operating state to another. Thus, the determination and subsequent transmission of the control signal to the consumption units UCh UC2, UCN allows for the programming of the operation of each consumption unit. Typically, the time period over which the operation of each consumption unit is programmed corresponds to one day, and each programming is therefore carried out from one day to the next.

[0055] Furthermore, the UNT processing unit is further configured to, during a real-time management phase implemented over a predetermined time period, update the control signal in real time based on the target average consumption. Thus, during this real-time management phase, the UNT processing unit is further configured to update in real time the nominal transition probability of each consumption unit based on the updated control signal.

[0056] As illustrated in [Fig.1], the UNT processing unit comprises a MEM memory, a PROC processor and a COM communication module.

[0057] The MEM memory is arranged to store the instructions of a computer program whose implementation by the PROC processor results in the operation of the UNT processing unit.

[0058] The PROC processor thus makes it possible, during the programming phase, to determine the control signal and the average target consumption of the set of one or more consumption units UCh UC2, UCN in order to update the nominal transition probability of each consumption unit.

[0059] The PROC processor also allows, during the management phase implemented over the time period, to update in real time the control signal according to the target average consumption and to update in real time the nominal transition probability of each consumption unit on the basis of the control signal thus updated.

[0060] The COM communication module is arranged to communicate with the production units UPi, UP2, UPn and the consumption units UCI, UC2, UCN.

[0061] For example, the COM communication module makes it possible to receive, from each production unit, the technical constraints and the cost of production characterizing the production unit in question.

[0062] The COM communication module also allows the control signal determined by the PROC processor to be transmitted to each consumption unit.

[0063] Similarly, the COM communication module makes it possible to transmit, to each production unit, the production program consistent with the target average consumption.

[0064] During the real-time management phase over the predetermined time period, the COM communication module allows the control signal updated at every instant to be transmitted to each consumption unit.

[0065] It is known to a person skilled in the art that there are many different types of data communication networks, for example radio communication networks, cellular or non-cellular, and that depending on the embodiment, the COM communication module may integrate one or more communication sub-modules, for example radio frequency communication, and be configured for the transmission and reception of radio frequency signals, according to one or more technologies, such as TDMA, FDMA, OFDMA, CDMA, or one or more radio communication standards, such as GSM, EDGE, CDMA, UMTS, HSPA, LTE, LTE-A, WiFi (IEEE 802.11) and WiMAX (IEEE 802.16), or their variants or evolutions, currently known or developed later.

[0066] The operation of the UNT processing unit, and more generally of the mixed portfolio modeled by the SYS system, will now be described with reference to [Fig.2] which illustrates the programming phase and then with reference to [Fig.3] which illustrates the management phase, implemented following the programming phase.

[0067] In the process described below, the UNT processing unit is responsible for managing the production and consumption flexibilities of the SYS system for a predetermined time period sampled into a succession of instants. In other words, the UNT processing unit is responsible for the joint programming, for coordinated management, of the production of the production units and the consumption of the consumption units.

[0068] During an SI step, the UNT processing unit receives at least one consumption simulation from each consumption unit of the SYS system.

[0069] The UNT processing unit can also receive data or information necessary for the construction of the program, for the coming period of time, from the production units UPi, UP2, UPn and the consumption units UCI, UC2, UCN.

[0070] Each consumption simulation of a consumption unit covers the predetermined time period. Each simulation relating to a consumption unit is parameterized by a nominal transition probability characterizing the probability of the consumption unit switching, at a given instant of the time period, from one operating state to another.

[0071] Advantageously, the UNT processing unit receives a high number of consumption simulations for each processing unit. In the following description of the process, the UNT processing unit receives the same number M of consumption simulations for each consumption unit. In other words, in the SYS system illustrated in [Fig. 1], the UNT processing unit receives M consumption simulations for consumption unit UC1, M consumption simulations for consumption unit UC2, and M consumption simulations for consumption unit UCN.

[0072] Consumption simulations are for example transmitted directly by the consumption units UCi, UC2, UCN to the processing unit UNT, which receives them via the communication module COM.

[0073] During an S2 step, the UNT processing unit determines, based on the received consumption simulation(s), a target average consumption of the set of one or more consumption units UCi, UC2, UCN and a control signal.

[0074] More precisely, the target average consumption and the control signal are determined by the PROC processor of the UNT processing unit by executing the instructions of the computer program stored in the MEM memory.

[0075] Furthermore, advantageously, during step S2, at least one production profile is determined for each production unit. The production profile(s) are determined jointly with the target average consumption and the control signal.

[0076] The production profile of a production unit depends on the technical state in which the production unit is operating and makes it possible to determine, at any given time, the quantity of electrical energy produced by the production unit in question. A set of possible technical constraints can be determined for each production unit and provided to the processing unit UNT.

[0077] Similarly, the consumption profile of a consumption unit depends on the state of operation of the consumption unit and makes it possible to determine, at each instant, the amount of electrical energy consumed by the consumption unit considered.

[0078] As explained previously, the predetermined time period over which the operation of the production units UPi, UP2, UPn and the consumption units UCI, UC2, UCN is programmed is sampled into a succession of instants. Consequently, the target average consumption and the control signal are also sampled, their respective values ​​being known for each time sample, therefore for each instant.

[0079] It should be noted that the target consumption and the control signal can be extended by continuity over the predetermined time period, for example in the form of step functions.

[0080] The programming aims overall to minimize the following criterion which characterizes the imbalance between the supply and demand for electrical energy within a SYS system and which the UNT processing unit must remedy through the management of consumption flexibilities:

[0081] [Math. 16] J(x, p) = E = x' ) - E jc + E " = ] / . (xJ) + ÿ E j where: - E is the mathematical expectation; -f0 is a function that penalizes a difference between the production profile of one or more production units and the consumption profile of one or more consumption units; - n is the number of production units; - Pj(xj) is the production profile of the j-th production unit when it operates in a technical state xj; - N is the number of consumption units; - Ci being a function which, to a state of the i-th consumption unit, associates a consumption profile, Ci(X') being the value of the consumption of the i-th consumption unit operating in the state X'; -fi(xj) is a production cost of the j-th production unit when it operates in a technical state xi; - k is a predetermined constant, allowing in particular to penalize deviations in consumption from the nominal profile; - D( / / ', fp) is the pseudo Kullback-Leibler distance between pi, which corresponds to the controlled probability distribution of the i-th consumption unit, and which corresponds to the nominal probability distribution, therefore in the absence of control, of the i-th consumption unit.

[0082] In the preceding criterion to be minimized, the first term corresponds to a failure penalty, the second term corresponds to a total production cost of the set of one or more production units UPi, UP2, UPn and the third term corresponds to a penalty for deviation from the nominal consumption of the set of one or more consumption units UCi, UC2, UCN.

[0083] As illustrated in [Fig. 2], step S2 of determining consumption target average and control signal is implemented according to an iterative algorithm.

[0084] At the input of this algorithm, the iteration rank p is initialized to 1 and the control signal is also initialized to an initial control signal 20.

[0085] Typically, the initial control signal 20 used to initialize the control signal prior to the implementation of the iterative algorithm is the control signal obtained at the end of a time period preceding the time period under consideration. In other words, the last control signal obtained during the time period preceding the upcoming time period for which programming is required can be used to initialize the control signal.

[0086] The iterative algorithm is defined as follows: [Math. 17] , . . , . . / / \ V n / i \ f \ V v' M i ci r J- \ At i (4) -(h) + P p' ' E • iP pf + E , LL P+ | \ K / p \ K / ttp \ KJ - P ap J \ ** z 1 q - ] py KJ * \ \ r JK / with : [Math. 18] vp = argminv { / Q ( r ) + E QÀp ( tk )} [Math. 19] *Jp = argmin^ / ( f . (xJ ) - EX ( 4 ) P -MW} * ' 1 l J 1 J \ / \ / J [Math.20] / AJ T / » x \ 1' \\ _ exP z E ^,4(4)+>y4 4)] l0" E 4 texp E 4,4OO+ rjL'”4)) [Math.21] ET / VZ 4 A \tt'\ p + ippa - uz - iq - ip ïV / \ J Or : -Xp(tk) is the value at time tk of the control signal at iteration p, the control signal being initialized to a predetermined initial control signal Xo; - pp is a positive real number; -f0 is a function that penalizes a difference between a production profile of one or more production units and a consumption profile of one or more consumption units; - T is the number of moments in the succession of moments in the predetermined time period; - vp is a failure profile at iteration p, vp(tk) is the value at time tk of the failure profile; - J, is a real number characterizing a zero-energy deflection constraint of the set of one or more consumption units in relation to a nominal consumption; - n is the number of production units; - Ai is a set of technical constraints constituting the set of technical states in which the jth production unit can operate; - Pj(xj) is the production profile of the j-th production unit when it operates in a technical state xi, Pj(xj)(tk) being the value at time tk of the production profile Pj(xJ); -fj(xj) is a production cost of the j-th production unit when it is operating in a technical state x1; - N is the number of consumption units; - M is the number of consumption simulations for each consumption unit; - k is a predetermined constant; - exp(-) is the exponential function; and - U^tk) is the value at time tk of the consumption of the i-th consumption unit according to the q-th consumption simulation

[0087] Furthermore, the iterative algorithm is interrupted when a predefined stopping criterion is met.

[0088] In the example illustrated in [Fig.2], the stopping criterion is defined as follows: [Math.22] It v - E"}P (xJUZN || <e Il p J = P y \ p 7 I = iq = 1 p II Or : - || ■ || is a standard; and - e is a real positive.

[0089] Of course, a person skilled in the art understands that other stopping criteria can be defined. For example, the stopping criterion for the iterative algorithm could be reaching or exceeding a predetermined number of iterations. The formula for the stopping criterion proposed above makes it possible to characterize that the difference between the failure profile, relating in particular to the penalty for the deviation between the production profile and the consumption profile, on the one hand, and the difference between the production profile and the consumption profile, on the other hand, must be arbitrarily as small as possible.

[0090] When the stopping criterion is met, the iterative algorithm terminates, thus stopping at a given iteration p. The algorithm has then generated as many control signals as there were iterations, not counting the initial control signal Xo. The desired control signal must then be found among all these control signals. For example, if a number a were required, the control signal determined at the end of step S2 is the initial control signal Xo or one of the control signals among the number n of control signals generated during the iterative algorithm.

[0091] For example, the selected control signal corresponds to the control signal, among all the control signals generated up to the last iteration satisfying the stopping criterion, for which a jump in duality is minimal.

[0092] In other words, the rank p* of the selected control signal satisfies: [Math.23] p * = argmin { Gap(p)} p with: [Math.24] Gap(p) = / ^ e - e--foh)- e - E ", + E:, E " EE " , E ",« /

[0093] The Gap function corresponds to the duality jump of the control signal Xp obtained at iteration p of the algorithm.

[0094] The control signal determined 2* thus corresponds to the control signal determined at iteration p*, i.e.: [Math.25] / = 2 * p

[0095] The rank p* of the iteration at which the control signal allows to reach, among all the control signals generated, a minimum duality jump also allows to determine the average target consumption of the set of one or more consumption units UCi, UC2, UCN.

[0096] As explained previously, the target average consumption allows for the determination not only of the consumption of the consumption units UCi, UC2, and UCN, but also of the production schedule of the production units UPi, UP2, and UPn. More precisely, the control signal transmitted to the consumption units UCi, UC2, and UCN aims to achieve, for each consumption unit, the target average consumption over the predetermined time period. The target average consumption is then used to determine a production schedule for the production units UPi, UP2, and UPn that is consistent with the target average consumption, so that they adapt their production, over the predetermined time period, to the electrical energy requirements of the consumption units UCi, UC2, and UCN.

[0097] The target average consumption is given by the following formula: [Math.26] EW y- M jqi, q{ \ Or : - r(tk) is the value at time tk of the average target consumption of the set of one or more consumption units; and - p* is the rank of the iteration, among all the iterations up to the last iteration satisfying the stopping criterion, for which a jump in duality is minimal.

[0098] During step S3, the processing unit UNT transmits the determined control signal to each consumption unit. The transmitted control signal is thus common to the consumption units UCi, UC2, and UCN of the SYS system. The control signal is transmitted to each consumption unit in order to achieve the target average consumption over the coming time period.

[0099] The control signal is for example transmitted by the COM communication module of the UNT processing unit.

[0100] Also during this S3 step, the processing unit transmits, via the COM communication module, a production program to each production unit of the SYS system. Since the consumption units UCi, UC2, and UCN are configured to follow the control signal and achieve the target average consumption, a supplementary production program, conforming to this target average consumption, must be transmitted. In this way, the SYS system can achieve a balance between the supply and demand of electrical energy over the coming period. The difference between the amount of electrical energy produced and the amount of electrical energy consumed is thus minimized.

[0101] Advantageously, the production program further includes, for each production unit, at least one production profile. As explained previously, the production program transmitted to the jth production unit may include a function pj which, for each technical state xj of the technical constraints A j of the jth production unit, associates a production profile Pj(xj). The production profile Pj(xj) makes it possible to determine, at a given time tk of the predetermined future time period, the electrical energy production pj(xj)(tk) of the jth production unit operating in technical state xj.

[0102] During an S4 step, the nominal transition probability is updated, at the level of each consumption unit, on the basis of the received control signal and the local function characterizing the operating specificities of the consumption unit considered.

[0103] As explained previously, the local function associated with a consumption unit depends, for example, on the programmed use of equipment The consumption unit, the technical characteristics of this equipment, and local contingencies. Furthermore, the local function of a consumption unit also depends on the control signal received, so the local function can only be determined upon receipt of the control signal from the processing unit UNT, and in particular from the communication module COM.

[0104] Typically, the nominal transition probability of the i-th consumption unit is updated as follows during the programming phase: [Math.27] p^.x^dYjv 1 ^ „r) with : [Math.28] V'Ct X) = J P'^. X, dY)V'(t t ^, Y)exp ( - Or : - Xj, X2 and Y are possible states of a consumption unit, dX], dX2 and dY are the differentials of such states, - PodtbXbdXd is the nominal transition probability characterizing the probability of the i-th consumption unit to switch, at time 4, from one operating state X2 to another operating state X2; - QditbXudX^ is the nominal transition probability updated during the initial programming phase characterizing the probability of the i-th consumption unit to switch, at time 4, from one operating state X2 to another operating state X2; - VdthX) is the value at time 4 of the local function characterizing the operating specificities of the i-th consumption unit and technical characteristics of said equipment, when the i-th consumption unit operates in state X; - exp(-) is the exponential function; - N is the number of consumption units; - k is a predetermined constant; - X(tk) is the value at time 4 of the determined control signal; and - Ci being a function which, to a state of the i-th consumption unit, associates a consumption profile, Ci(Y) being the value of the consumption of the i-th consumption unit operating in state Y.

[0105] It should be noted that the states of the consumption units UCi, UC2, UCN are not necessarily discrete states and may be continuous. Thus, some industrial or domestic equipment may operate in discrete states with a first a state corresponding to the "on" mode and a second state corresponding to the "off" mode. However, some industrial or domestic equipment can operate in continuous states, such as an air conditioning system or an electric storage water heater which, during heating mode, can operate at different temperatures.

[0106] Therefore, it is relevant in the preceding formula for updating the nominal transition probability to consider the differential of the operating states to take into account the fact that the operating states of the consumption units UCi, UC2, UCN may be continuous and not discrete.

[0107] Step S4, with the updating of the nominal transition probability of each consumption unit, concludes the programming phase of the management process according to the invention. At this stage, each consumption unit has a nominal transition probability for the upcoming predetermined time period. This nominal transition probability varies from one instant of the predetermined time period to another, depending on the control signal, itself sampled over the predetermined time period, and the local function, which also depends on the control signal. As explained previously, the real-time operation of each consumption unit can be modeled by random behavior based on the nominal transition probability. Thus, the operation of each industrial or domestic piece of equipment at the level of each consumption unit depends on the nominal transition probability of the consumption unit.

[0108] The set of consumption units UCi, UC2, UCN aims to reach the target average consumption, also transmitted during step S3 by the COM communication module of the UNT processing unit, determined at the end of the iterative algorithm.

[0109] Furthermore, within the framework of the management process according to the invention, the programming phase can advantageously be supplemented by a real-time management phase, implemented during the predetermined time period and therefore subsequent to the programming phase. Such a management phase allows for coherent integration with the programming phase.

[0110] Note that, in the absence of a real-time management phase, the nominal transition probability of each consumption unit can be programmed as follows over the predetermined time period: [Math.28] JP\Jtk,X.dX2\V\tk+[,X2) NS! \ dX2 - ,,-z, y / «P \ 1 / y (db x û /

[0111] Such a nominal transition probability aims to calculate a preliminary deviation by minimizing the following quantity: [Math.29] y .v / \ D ( ) + E £ = 0Z ( 4 ) ( Ci ) \tk) with: [Math.3O] = E[c,(xi) ] = J Ci{X)fàdX)

[0112] The management phase implemented over the predetermined time period is described below with reference to [Fig. 3]. Such a management phase aims to update the control signal in real time to update, also in real time, the nominal transition probability at the level of each consumption unit of the SYS system.

[0113] During the management phase, the entire set of one or more production units UPi, UP2, UPn follows the production program transmitted by the processing unit UNT. The production of each production unit is thus programmed over the predetermined upcoming time period.

[0114] During an S5 step, the UNT processing unit, and more specifically the PROC processor, updates the control signal in real time according to the target average consumption.

[0115] To do this, the UNT processing unit provides, at each instant of the predetermined time period, an average consumption at the next instant of the set of one or more consumption units UCi, UC2, UCN as a function of the updated nominal transition probability of each consumption unit.

[0116] The UNT processing unit then updates, at each instant, the control signal based on a comparison between the expected average consumption at the next instant and the target average consumption.

[0117] In the embodiment illustrated in [Fig.2], the update of the control signal during step S5 of the management phase is implemented as follows:

[0118] The UNT processing unit constructs a predictive consumption model that allows approximating, based on the control signal at a given time, the average consumption predicted at the next time.

[0119] In other words, a function F is generated by the processing unit UNT. This function F takes as input the value of the control signal at a given time and returns the average consumption of the set of one or more consumption units UCi, UC2, UCN expected for the next time.

[0120] Once the predictive consumption model has been built, it is modified by learning at each instant of the predetermined time period.

[0121] At each instant of the predetermined time period, the UNT processing unit compares an actual average consumption with a predicted average consumption at that instant for all one or more consumption units UCi, UC2, UCN. Such a comparison makes it possible to measure the difference or distance between the approximate average consumption and the actual average consumption at that instant.

[0122] Then, the UNT processing unit, and more precisely the PROC processor, adapts by learning, at each instant, the predictive consumption model, therefore the function F, on the basis of the comparison carried out.

[0123] Finally, the processing unit updates the control signal at every instant by reciprocally applying the predictive consumption model to the target average consumption at the next instant. The reciprocal application F1 thus allows adaptation to the constraints, uncertainties, and specific characteristics of the consumption units UCi, UC2, and UCN of the SYS system.

[0124] In other words, at the end of step S5, the control signal is updated as follows: [Math.31] =r\r(t k+i )) where is the value at time tk of the updated control signal.

[0125] During an S6 step, the UNT processing unit transmits, at each instant of the predetermined time period, the updated control signal to each consumption unit.

[0126] As explained previously, the updated control signal is transmitted, for example, by the COM communication module. It should be noted that the control signal updated during the real-time management phase is common to all consumption units UCi, UC2, UCN, as is the control signal obtained at the end of the programming phase.

[0127] Finally, during an S7 step, the nominal transition probability is updated, at the level of each consumption unit, based on the updated control signal received.

[0128] For example, the nominal transition probability can be updated as follows during the management phase:

[0129] [Math.32] Ql dx2 ^x^dx^p (-B(4)c(*2)) J X3, dY^exp ( - Or : - Qi(tioX],dX2) is the nominal transition probability updated during the phase of management characterizing the probability of the i-th consumption unit to switch, at time tk, from an operating state Xi to another operating state X2; - Ç(tk) is the value at time tk of the updated control signal.

[0130] A person skilled in the art understands that, during the predetermined time period in which the real-time management phase is implemented, each consumption unit of the SYS system operates according to the nominal transition probability updated in real time. Thus, at each instant of the time period, the control signal is updated by the PROC processor at the UNT processing unit. The modified control signal is transmitted to each consumption unit. The nominal transition probability is also updated at each instant, based on the received control signal. The operation of each consumption unit is parameterized by the nominal transition probability. In other words, the consumption unit transitions from one state or operating mode to another in accordance with the updated nominal transition probability.At the level of the i-th consumption unit, the function c; allows us to determine, for each operating state, the associated consumption profile ci(Xi). The consumption profile ci(X') allows us to determine, at each instant tk of the time period, the consumption ci(Xi)(tk) of the i-th consumption unit.

[0131] Similarly, the production units UPi, UP2, UPn follow the production schedule, over the time period, transmitted by the processing unit UNT. As explained previously, this production schedule is consistent with the target average consumption and helps to improve the balance between the supply and demand of electrical energy.

[0132] Thus, domestic or industrial equipment located at the level of each consumption unit modulates its consumption of electrical energy, supplied by the set of one or more production units UPh UP2, UPn by modifying their state or mode of operation in real time by following the nominal transition probability updated at each reception of the control signal, also updated in real time.

[0133] The use of the predictive model makes it possible to get closer to the target average consumption determined during the programming phase while taking into account, for each unit of consumption, the specific local data.

Claims

Demands

1. A method, implemented by computer means, for the coordinated management, over a predetermined time period sampled into a succession of instants, of a set of one or more production units (UPi, UP2, UPn) and a set of one or more consumption units (UCi, UC2, UCN) supplied with electrical energy by said set of one or more production units, said method comprising, during a programming phase: - receiving (SI) at least one consumption simulation of each consumption unit, each consumption simulation of a consumption unit covering said time period and being parameterized by a nominal transition probability characterizing the probability of said consumption unit switching, at a given instant of said time period, from one operating state to another, - determining (S2), as a function of the consumption simulation(s) received,a target average consumption of one or more consumption units and a control signal sampled over said time period, - transmit (S3), to each consumption unit, said control signal to update (S4), at the level of each consumption unit, the nominal transition probability based on the received control signal and a local function characterizing the operating specifics of the consumption unit considered, and - transmit (S3), to each production unit, a production program over said time period consistent with said target average consumption.

2. Management method according to claim 1, wherein the production program includes at least, for each production unit, a production profile determined jointly with the target average consumption and the control signal.

3. A method according to claim 2, wherein the sampled control signal is determined during the programming phase according to an iterative algorithm defined as follows: [Math.32] 3 +i (^) Where +p G' ta)- E " .P tata ta EA । EMP + ] v K ! p \ K. p\ P y KJ J-*1 j\ P / \ ^ / =1^^=1 / 7 \ K / / with : [Math.33] {T \ / 0( v) + E À . = 0 ^(4)v(rJ [Math.34] has j = aremin j J \ f ( x J ) - E f ( tk ) P ) ! pax eA IJ j ' k = Ü pv K7'j y / \ / I [Math.35] [Math.36] V-' T jV rT A / î . q 1 A \ y = y +p L k_0L}L ~n)u M p + 1 pph ”ui — iq — ip in / y / Or : kp(tk) is the value at time tk of the control signal at iteration p, the control signal being initialized to a predetermined initial control signal k0; - pp is a positive real number; -fo is a function that penalizes a difference between a production profile of one or more production units and a consumption profile of one or more consumption units; - T is the number of moments in the succession of moments in the predetermined time period; - VP is a failure profile at iteration p, Vp(tk) is the value at time tk of the failure profile; - Y is a real number characterizing a zero energy deviation constraint of the set of one or more consumption units with respect to a nominal consumption; - n is the number of production units; - A> is a set of technical constraints constituting the set of technical states in which the jth production unit can operate; - Pj(xj) is the production profile of the j-th production unit when it operates in a technical state xi, Pj(xj)(tk) being the value at time tk of the production profile Pj(xJ); -f / xi) is a production cost of the j-th production unit when it operates in a technical state x1; - N is the number of consumption units; - M is the number of consumption simulations of each consumption unit; - k is a predetermined constant; - exp(-) is the exponential function; and - tk) is the value at time tk of the consumption of the i-th consumption unit according to the q-th consumption simulation, the iterative algorithm being interrupted when a predefined stopping criterion is met.

4. A method according to claim 3, wherein the stopping criterion is defined as follows: [Math.37] ii V- H / i \ V-' NV' M in GQ ■■ v - EJ? (x ) + E ■ , E < e 11 PJ = Ij \ p J l = 1 « = 1 p 11 where: - || • || is a norm; and - e is a positive real number

5. A method according to claim 3 or 4, wherein the initial control signal used to initialize the control signal prior to the implementation of the iterative algorithm is the control signal obtained at the end of a period of time preceding the period of time considered.

6. A method according to any one of claims 3 to 5, wherein the control signal determined at the end of the iterative algorithm is defined as follows: [Math.38] 2 —À * p where p* is the rank of the iteration, among all iterations up to the last iteration satisfying the stopping criterion, for which a duality jump is minimal, i.e.: [Math.39] p* = argmin {Gap(p)} p with: [Math.40] Gap(p) = / 0( E - EL E-f0Gp) - E - En + XVO-r Ef nEN ,EM j = ij\ rP.) = ipp N} \ '

7. A method according to any one of claims 3 to 6, wherein the target average consumption is defined as follows: [Math.41] Or : - r(tk) is the value at time tk of the average target consumption of the set of one or more consumption units; and - p* is the rank of the iteration, among all iterations up to the last iteration satisfying the stopping criterion, for which a duality jump is minimal, i.e.: [Math.42] p* = argmin { Gap(p)} p with : [Math.43] c^p(pi =. / „(£y,P.- ÏME' / fï'')- / „(>■„)- - E + E'^^'WhL.E:,E

8. A method according to any one of the preceding claims, wherein the nominal transition probability is updated as follows during the programming phase: [Math.44] J Pi^, x^dYp^, r) with : [Math.45] X) = JP^i,. X. Y)ap ( - ,)<:,( / )) Y / Or : - X], X2 and Y are possible states of a consumption unit, dXh dX2 and dY are the differentials of such states, - PodtbXj^Xp is the nominal transition probability characterizing the probability of the i-th consumption unit to pass, at time tk, from an operating state X] to another operating state X2; - QdttbXydXd is the nominal transition probability updated during the initial programming phase characterizing the probability of the i- oo th dX2 - i-th consumption unit to switch, at time tk, from one operating state X2 to another operating state X2; - V'(tk,X) is the value at time tk of the local function characterizing the operating specifics of the i-th consumption unit and technical characteristics of said equipment, when the i-th consumption unit operates in state X; - exp(-) is the exponential function; - N is the number of consumption units; - k is a predetermined constant; - V(tk) is the value at time tk of the determined control signal; and - Ci being a function which, to a state of the i-th consumption unit, associates a consumption profile, Ci(Y) being the value of the consumption of the i-th consumption unit operating in state Y.

9. A method according to any one of the preceding claims, wherein the local function characterizing the operating specificities of a consumption unit depends on a programmed use of industrial equipment of said consumption unit, on technical characteristics of said industrial equipment and on local contingencies.

10. A method according to any one of the preceding claims, further comprising, following the programming phase, a management phase implemented over said time period including the real-time updating (S5) of the control signal as a function of said target average consumption; and the real-time updating (S7) of the nominal transition probability of each consumption unit based on the updated control signal.

11. A method according to claim 8, further comprising, following the programming phase, a management phase implemented over said time period, comprising the real-time updating of the control signal as a function of said average target consumption; and the real-time updating of the nominal transition probability of each consumption unit based on the updated control signal, wherein the nominal transition probability is updated as follows during the management phase: [Math.46] J. V ,VLX,. dX^exp ( - ^ f ( t„ ) q ( X, ) ) y A , ^A2 — J Q'0(lf XdY)exp (-^(4¼^)) where: - Q(tbXbdX2) is the nominal transition probability updated during the management phase characterizing the probability of the i-th consumption unit to switch, at time 4, from an operating state X] to another operating state X2; - C(tk) is the value at time tk of the updated control signal.

12. A method according to claim 10 or 11, wherein the management phase is implemented as follows: - forecasting, at each instant of said time period, an average consumption at the next instant of the set of one or more consumption units as a function of the updated nominal transition probability of each consumption unit, - updating, at each instant, the control signal on the basis of a comparison between the forecasted average consumption at the next instant and the target average consumption, and - transmitting (S6), at each instant, said updated control signal to each consumption unit to update, at the level of each consumption unit, the nominal transition probability on the basis of the updated control signal received.

13. A method according to claim 12, wherein the updating of the control signal during the management phase is implemented as follows: - construct (S51) a predictive consumption model enabling the approximation, as a function of the control signal at a given instant, of the average consumption predicted at the next instant, then, at each instant of the predetermined time period: - compare an actual average consumption and an average consumption predicted at said instant of the set of one or more consumption units, - adapt (S52) by learning, at each instant, said predictive consumption model on the basis of said comparison, and - update (S53), at each instant, the control signal by reciprocal application of the predictive consumption model to the target average consumption aimed at the next instant.

14. A method according to any one of claims 10 to 13, wherein, during the management phase implemented over said time period, each consumption unit operates according to the nominal transition probability updated in real time, while each production unit follows the received production schedule.

15. Computer program, comprising instructions for implementing the method according to any one of the preceding claims when said instructions are implemented by at least one processor (PROC).

16. Processing unit (PU) for the coordinated management, over a predetermined period of time sampled into a succession of instants, of a set of one or more production units (UPb UP2, UPn) and a set of one or more consumption units (UCb UC2, UCN) supplied with electrical energy by said set of one or more production units, said processing unit being configured to, during a programming phase: - receive at least one consumption simulation for each consumption unit, each consumption simulation for a consumption unit covering said time period and being parameterized by a nominal transition probability characterizing the probability of said consumption unit switching, at a given instant of said time period, from one operating state to another, - to determine, based on the consumption simulation(s) received, a target average consumption for one or more consumption units and a control signal sampled over said time period, - transmit, to each consumption unit, said control signal in order to update, at the level of each consumption unit, the nominal transition probability based on the received control signal and a local function characterizing the operating specifics of the consumption unit in question, and - transmit, to each production unit, a production program over the said period of time consistent with the said target average consumption.

Images