An online feedback aggregation method for distributed energy resource time coupling flexibility

Abstract

The application discloses an online feedback aggregation method for time coupling flexibility of distributed energy resources, adopts a unified power-energy boundary model to describe independent flexibility regions of each distributed resource, performs arithmetic superposition on the resource boundaries to form a theoretical aggregation feasible region at a system level, introduces power scaling coefficients and energy scaling coefficients to parameterize and shrink the theoretical boundary with the aggregation power baseline as the center, extends a single-step scheduling instruction issued by a system operator into a complete forward-looking power trajectory in real-time operation, and obtains a tracking error by solving a decomposition problem; and the tracking error is used to construct an approximate gradient to perform online gradient descent update on the scaling coefficients. The process gradually shrinks the flexibility boundary into a conservative inner approximation of the real aggregation flexibility region. The application can solve the problem of insufficient accuracy or excessive calculation overhead of the aggregation flexibility evaluation under time coupling constraints in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of power system optimization and scheduling technology, specifically relating to an online feedback aggregation method for the time-coupled flexibility of distributed energy resources. Background Technology

[0002] Distributed energy resources (DERs), such as photovoltaics, energy storage systems, electric vehicles, and temperature-controlled loads, are becoming a key source of power system flexibility. Due to their large numbers and limited individual capacity, DERs are typically coordinated and dispatched in aggregate form. Accurately assessing aggregate flexibility is crucial for aggregators and system operators, especially in real-time operation where timely flexibility information is needed to support dispatch decisions.

[0003] Many DERs, such as energy storage systems, electric vehicles, and temperature-controlled loads, exhibit energy storage-like behavior, meaning that actions at a particular time step affect future flexibility. Therefore, aggregated flexibility forms a high-dimensional, time-coupled feasible region, making accurate characterization extremely challenging. Existing methods can be broadly categorized into time-coupled models and time-decoupled models. Time-coupled models can capture time dependencies and provide more accurate flexibility estimates, but determining their feasible boundaries is difficult, either requiring computationally intensive methods or leading to optimistic results that include infeasibility points. In contrast, time-decoupled methods define time-period power boundaries, forming a simple set of hyperrectangles suitable for online use, but this simplification reduces true flexibility and often produces overly conservative estimates. Furthermore, scheduling decisions based solely on instantaneous power boundaries ignore time coupling effects, potentially compromising the sustainability of flexibility provision.

[0004] Therefore, how to calculate the aggregation flexibility of time coupling online, so as to maintain both computational efficiency and reliability, is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] To address the problems of existing technologies, this invention provides an online feedback aggregation method for the time-coupled flexibility of distributed energy resources. It utilizes scheduling tracking error as feedback to derive approximate gradients of scaling coefficients in the flexibility model, achieving online gradient descent to update these parameters and reduce tracking error. Combined with state updates after each control action, it continuously outputs a real-time aggregated flexibility model with improved reliability.

[0006] To achieve the above objectives, the present invention provides the following solution: An online feedback aggregation method for distributed energy resources with time-coupled flexibility includes the following steps: Obtain the state information of each distributed resource within the aggregate at the current moment, and calculate the independent flexibility region of each resource in the future look-ahead time domain based on the power-energy boundary model; The power boundary and energy boundary of each resource are arithmetically superimposed to obtain the theoretical aggregated power boundary and energy boundary at the system level. Introducing power scaling factor and energy scaling factor Centered on the polymerization power baseline, the theoretical polymerization boundary is expressed as a deviation form and parametrically shrunk to construct an approximate polymerization flexibility region; The system receives the scheduling signal determined by the system operator within the approximate aggregation flexibility region and extends the single-step scheduling signal to a complete power trajectory covering the entire look-ahead time domain; The complete power trajectory is decomposed into independent flexibility regions of each resource to obtain a practically executable aggregated power trajectory and tracking error; When the tracking error exceeds the preset tolerance, identify the overly optimistic set of active constraints, construct the loss function, and calculate... and The approximate gradient is obtained, and the nonnegative projection is applied using the gradient descent method to update the parameters. The first step of the decomposed power command is executed, the status of each resource is updated, the time step is advanced, and the above process is repeated with the updated parameters and status to form a closed-loop control.

[0007] Preferably, the distributed resources include energy storage systems, photovoltaic power generation units, electric vehicle charging units, and temperature-controlled load units.

[0008] Preferably, the power-energy boundary model is expressed in a unified form: for each resource i, at a step size of of Within the look-ahead time domain, its power trajectory must simultaneously satisfy the progressive upper and lower bound constraints of power and the cumulative upper and lower bound constraints of energy. The power boundary is determined by the physical operation constraints of each resource, while the energy boundary is characterized by the fastest and slowest feasible energy trajectories.

[0009] Preferably, the aggregate power baseline is the power curve of the aggregate when it does not participate in system-level control, and is composed of the sum of the resource baselines; the baseline of the energy storage system is zero output, the baseline of the electric vehicle is charging at maximum power from the current moment until the energy demand is met, the baseline of the photovoltaic unit is its maximum power generation curve, and the baseline of the temperature-controlled load is maintaining the set temperature.

[0010] Preferably, the parameterized shrinkage is performed by expressing the power constraint as follows: Expressing the energy constraint as , where A=[I,-I] T for Power constraint matrix, for An identity matrix of order L; C = [L, -L] Tfor Energy constraint matrix, for A lower triangular matrix of order 1, with all elements on the diagonal and below being of order 2. ; This is the power deviation width vector; The energy deviation width vector; These are the power and energy scaling factors, respectively; when When, the feasible region is an outer approximation of the theoretical superposition boundary; when At that time, the feasible region shrinks to contain only the baseline point and the tracking error is zero.

[0011] Preferably, the method for expanding the single-step scheduling signal into a complete power trajectory is as follows: using a random vector A linear programming model is constructed for the coefficients of the objective function, and the power in the first step is constrained to be equal to the scheduling signal value. The solution is obtained within the contracted aggregation flexibility region to obtain the complete power trajectory located on the boundary of the feasible region. Boundary points are more likely to induce larger tracking errors than interior points, thus exposing the infeasibility of the boundary more efficiently and improving the efficiency of parameter update.

[0012] Preferably, the method for identifying the active constraint set is as follows: Let The optimal solution to the decomposition problem is the executable trajectory. For the target trajectory, given the power constraint matrix The Okay, if Then it will be included in the active set. For the energy constraint matrix The Okay, if Then it will be included in the active set. .

[0013] Preferably, the loss function is defined as the loss function from the active constraint hyperplane to the executable trajectory. Normalized sum of squared distances: ;in The approximate gradient is: ; The approximate gradient is: The approximate gradient ignores right and The implicit dependency, this simplification factor It is reasonable that it is located on the boundary of the real flexibility area.

[0014] As a preferred option, the parameter update formula is: ,in and They are respectively and The step size and non-negative orthogonal projection ensure that the parameters are always within the feasible range; due to and Gradient descent is monotonically decreasing and bounded to zero, so convergence is guaranteed.

[0015] Preferably, the scheduling is executed within a rolling scheduling framework, with only the first power command executed at each step, and the status information of each resource updated, including the state of charge of the energy storage system, the cumulative charging energy of the electric vehicle, and the indoor temperature of the temperature-controlled load; once the tracking error remains below the tolerance for a period of time, and It can be fixed and no longer adjusted to reduce computational overhead.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention, through the aforementioned technical solution, realizes a closed-loop system of "reporting flexibility—receiving scheduling—executing and observing tracking errors—feedback correction." Compared with existing methods, this invention has the following advantages: First, the time-coupled model retains the impact of current actions on future flexibility, avoiding the defect of time decoupling methods ignoring coupling effects; Second, power and energy dual scaling factors achieve decoupled independent correction; Third, the explicit approximate gradient based on active constraints makes the correction mathematically optimal and converges quickly; Fourth, the aggregation process itself only involves basic arithmetic operations, resulting in extremely high computational efficiency, suitable for large-scale real-time DER operation. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of an online feedback aggregation method for the time-coupling flexibility of distributed energy resources according to an embodiment of the present invention; Figure 2 In a specific embodiment of the present invention, the tracking error and scaling factor ( and Evolution curve over time; Figure 3 This is a comparison diagram of the power trajectory implementation of the present invention and existing methods (PEB external approximation and PB internal approximation) in a specific embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1: This invention provides an online feedback aggregation method for the time-coupling flexibility of distributed energy resources, comprising the following steps: Obtain the state information of each distributed resource within the aggregate at the current moment, and calculate the independent flexibility region of each resource in the future look-ahead time domain based on the power-energy boundary model; The power boundary and energy boundary of each resource are arithmetically superimposed to obtain the theoretical aggregated power boundary and energy boundary at the system level. Introducing power scaling factor and energy scaling factor Centered on the polymerization power baseline, the theoretical polymerization boundary is expressed as a deviation form and parametrically shrunk to construct an approximate polymerization flexibility region; The system receives the scheduling signal determined by the system operator within the approximate aggregation flexibility region and extends the single-step scheduling signal to a complete power trajectory covering the entire look-ahead time domain; The complete power trajectory is decomposed into independent flexibility regions of each resource to obtain a practically executable aggregated power trajectory and tracking error; When the tracking error exceeds the preset tolerance, identify the overly optimistic set of active constraints, construct the loss function, and calculate... and The approximate gradient is obtained, and the nonnegative projection is applied using the gradient descent method to update the parameters. The first step of the decomposed power command is executed, the status of each resource is updated, the time step is advanced, and the above process is repeated with the updated parameters and status to form a closed-loop control.

[0022] As one embodiment of the present invention, the distributed resources include an energy storage system, a photovoltaic power generation unit, an electric vehicle charging unit, and a temperature-controlled load unit.

[0023] As one embodiment of the present invention, the power-energy boundary model is expressed in a unified form: for each resource i, at a step size of of Within the look-ahead time domain, its power trajectory must simultaneously satisfy the progressive upper and lower bound constraints of power and the cumulative upper and lower bound constraints of energy. The power boundary is determined by the physical operation constraints of each resource, while the energy boundary is characterized by the fastest and slowest feasible energy trajectories.

[0024] Let the current time be Consider a step size of of Looking ahead to the next time domain, among which Index future time steps. At each time step... DER aggregators report their time-domain performance to upper-level system operators. The flexible area on Operators The internal system determines the scheduling signal and sends it back to the aggregator, which then executes the scheduling and updates its internal state variables. This is due to the flexibility of the reporting area. Possibly related to real aggregation flexibility The difference will result in a non-zero tracking error. The aggregator then adjusts the flexibility region and moves to the next time step based on the observed tracking error and the updated resource status.

[0025] set up For the set of DERs in the aggregate, | |=N. For each resource ,definition For its power trajectory in the look-ahead time domain, Its independent flexibility area. True aggregate flexibility. For the aggregate power trajectory The feasible region, calculated by Minkowski: The classic power-energy boundary model is used to characterize the independent flexibility set. The power boundary is determined by physical operational constraints, the energy boundary is defined by the fastest and slowest feasible energy trajectories, and the independent flexibility region is represented as: in, and These are the lower and upper bounds of the power. and These are the corresponding lower and upper bounds of the cumulative energy.

[0026] Due to the individual model Accurately calculating aggregation flexibility has exponential complexity, and aggregators need a simplified approximate model. Report to higher-level operators. Using an approximate model means based on... The calculated scheduling trajectory may not be accurately decomposed into feasible trajectories for each DER, resulting in tracking errors. For the scheduling trajectory... The tracking error is defined as: in, For DER at time The actual achievable aggregate power trajectory.

[0027] As one embodiment of the present invention, the power and energy boundaries of all DERs within the aggregator are summed: However, the feasible set obtained by this method is only... External approximations can lead to significant tracking errors. To address this, this invention proposes a closed-loop aggregation method that uses the tracking error as a feedback iterative correction boundary, adjusting online until the error is reduced to within an acceptable tolerance.

[0028] To achieve this correction, a polymerization power baseline is introduced. This represents the operating conditions when the aggregate does not participate in system-level control (e.g., zero energy storage output, electric vehicle charging at maximum power, photovoltaic power generation at maximum, temperature-controlled load maintaining set temperature). Using a baseline as a reference, the power and energy boundaries are expressed as deviations and adjusted using scaling factors. in for identity matrix for A lower triangular matrix with all elements on the diagonal and below it being... , These are the power and energy scaling factors, respectively.

[0029] set up Let P be the feasible region defined by the above constraints. When At that time, it reverts to an overlay boundary, with ;when At that time, P degenerates to the baseline. ,have Furthermore, the tracking error is zero. Therefore, α and β can be gradually decreased from 1 until the error meets the preset tolerance.

[0030] As one implementation of the present invention, in actual operation, the system operator only provides the current step scheduling signal. Decomposition requires the complete trajectory. Aggregators need to build coverage before decomposition. The trajectory is designed to maintain forward-looking flexibility. Assume the received signal... Compared to It is feasible, that is The extended trajectory is obtained by solving the following optimization problem: in, for The problem involves sampling random vectors in a subspace of dimensions 2 to K along random directions. Boundary points, because they are more likely to induce larger tracking errors, help to expose infeasibility and improve performance. and The update efficiency and the obtained trajectory This is then used to calculate the tracking error, thus obtaining the tracking error. and executable trajectory .

[0031] As one embodiment of the present invention, tracking error The two optimization problems mentioned above depend on and In principle, it can be done through Gradient updates can be used to update parameters, but this dependency is implicit, requiring differentiation of nested functions, which is computationally expensive and numerically unstable. Inspired by approximate gradient methods in power flow analysis, this invention constructs an approximate gradient to achieve efficient and stable parameter updates.

[0032] set up To find the optimal solution to the decomposition problem, if ,but Falling Externally, this indicates that the approximate boundary is overly optimistic, and it is directed towards... Contraction can reduce tracking error. Let... and These are sets of constraints in power and energy constraints, identified by comparing projections along each constraint direction: in and The first and second constraints are respectively the first constraint matrix A and the second constraint matrix C. row and number Row vectors, the goal is to reduce these boundaries and Squared distance between them: right and The gradient is: in, It also implicitly depends on and This dependency is ignored in this invention. The rationale for this simplification is that... Located in the true flexibility zone On the boundary, therefore along Direction adjustment and This can also improve feasibility. Under all these approximations, the above gradient serves as the tracking error. right and The approximate gradient.

[0033] The parameters are then updated via the gradient descent step: in, and They are respectively and The step size and non-negative orthogonal projection ensure that the parameters are always within the feasible range.

[0034] As one embodiment of the present invention, initialization , Set step size , and tolerance At each time step Update each DER model and calculate the superimposed power-energy boundary, then use and Build an aggregation model and report it to the upper-layer operator. Receive scheduling signal. Then, the extended trajectory is generated and the decomposition problem is solved to obtain... and tracking error .like Identify active constraints, calculate approximate gradients, and update. and Finally, execute Update DER status, incrementing. Repeat the above process.

[0035] In each step, and Updated by tracking error feedback, because and It is monotonically decreasing and bounded to zero, so convergence is guaranteed. The main computation comes from solving the extended problem (…). Linear programming and decomposition problems Quadratic programming problems can all be solved efficiently by commercial solvers (such as Gurobi). In practice, once... Continue for a sufficient period of time, and It can be fixed to reduce computational overhead.

[0036] Example 2 like Figure 1 As shown, this invention provides an online feedback aggregation method for the time-coupling flexibility of distributed energy resources, comprising: 1) Distributed resource unit flexibility modeling and overlay boundary calculation; 2) Introduce scaling factors to construct parameterized aggregation flexibility regions and report them to upper-layer operators; 3) Receive scheduling signals and expand them into a complete power trajectory covering the look-ahead time domain; 4) Solve the power decomposition problem to obtain the executable trajectory and tracking error; 5) Identify active constraints, calculate approximate gradients, and update scaling factors; 6) Execute the current step power command, update the resource status, and roll forward.

[0037] In step 1), it is assumed that there are N heterogeneous distributed resources in the aggregate, and the time step is... The forward-looking time domain includes Step. For each resource Calculate the future based on its current state upper limit of step power Lower power limit Energy Upper Boundary and lower energy level This defines its independent flexibility zone. The power boundary is determined by the physical constraints of each resource, while the energy boundary is characterized by the fastest and slowest feasible energy trajectories. By arithmetically superimposing the power and energy boundaries of each resource, the theoretical aggregate power boundary at the system level is obtained. , and theoretical convergence energy boundary , .

[0038] In step 2), since the feasible region formed by directly superimposed boundaries is an external approximation of the true aggregation flexibility, it may lead to significant tracking errors. Therefore, a power scaling factor is introduced. and energy scaling factor Based on the polymerization power baseline The theoretical boundary is parameterized and contracted around the central point. The aggregated power baseline is the sum of the resource baselines, representing the operating power curve of the aggregate when it does not participate in system-level control. The energy storage system baseline is zero output, the electric vehicle baseline is charging at maximum power from the current moment until energy demand is met, the photovoltaic unit baseline is its maximum power generation curve, and the temperature-controlled load baseline is maintaining the set temperature. Power and energy constraints are expressed in deviation form: , ,in .when When it is restored to the superimposed boundary (external approximation); when The time degenerates into a region containing only baseline points (zero error), resulting in a shrunken area of ​​aggregate flexibility. Submitted to the upper-level system operator.

[0039] In step 3), the system operator... The system determines and issues the current step scheduling signal. Because subsequent decomposition requires a complete... Step power trajectory to maintain look-ahead flexibility, aggregator with random vector To define the coefficients of the objective function, construct a linear programming model, constraining the power in the first step to equal... In the region of polymer flexibility after shrinkage Solve within the inner region to obtain the complete power trajectory located on the feasible region boundary. The reason for using boundary sampling is that boundary points are more likely to induce larger tracking errors than interior points, thus exposing the infeasibility of approximate boundaries more efficiently and improving parameter update efficiency.

[0040] In step 4), the target trajectory obtained in step 3) is... Decomposed into independent flexibility regions for each resource. Specifically, to minimize the aggregate power trajectory. With the target trajectory Differences between For the goal, and Solving this quadratic programming problem yields a practically executable aggregate power trajectory. and tracking error .

[0041] In step 5), when the tracking error Exceeding the preset tolerance When this occurs, it indicates that the approximate boundary is too optimistic and parameter correction is needed. First, identify the set of active constraints: for the power constraint matrix... The Okay, if Then it will be included in the active set. For the energy constraint matrix The Okay, if Then it will be included in the active set. Define the loss function. To active constrain the hyperplane to the executable trajectory Calculate the normalized sum of squared distances, and calculate them separately. right and Approximate gradient: Finally, the parameters are updated using gradient descent: Non-negative orthogonal projection ensures that the parameters are always within the feasible range.

[0042] In step 6), the aggregator executes the first-step power command obtained from the decomposition. It also updates the status information of each resource, including the state of charge of the energy storage system, the cumulative charging energy of electric vehicles, and the indoor temperature of the temperature-controlled load. The time step advances to... Steps 1) to 6) are repeated with the updated scaling factor and resource status to form a closed-loop online feedback aggregation method for the time-coupled flexibility of distributed energy resources. Because and During gradient descent, the gradient is monotonically decreasing and bounded to zero, ensuring convergence. However, once the tracking error remains below the tolerance for a period of time, and It can be fixed and no longer adjusted to reduce computational overhead.

[0043] The method described in this invention will be further illustrated below with a specific simulation embodiment. The simulation employs high temporal resolution and a time step size of [missing information]. Minutes, with a look-ahead time of 2 hours ( The simulation period is from 00:00 to 22:00. The aggregate includes 50 electric vehicles, 40 temperature-controlled loads, 5 energy storage units, and 5 photovoltaic units, totaling 100 heterogeneous devices. In each step, Random sampling was performed within the feasible power range to simulate upper-level scheduling. The simulation was implemented in a Python environment on a desktop computer (Intel i7-14790F, 32GB RAM), with an average computation time of 4.12 seconds per step, meeting the real-time running requirements of 1-minute resolution.

[0044] The specific parameters for each resource are as follows: Energy Storage Units (ESS): 5 units in total. ESS-1: Maximum charge / discharge power = 100kW, rated capacity = 600kWh, charge / discharge efficiency = 0.95, initial SOC = 300kWh; ESS-2: Maximum charge / discharge power = 150kW, rated capacity = 800kWh, charge / discharge efficiency = 0.95, initial SOC = 300kWh; ESS-3 / 4 / 5 parameters are the same as ESS-1 / 2 / 1, respectively. The energy boundary of the energy storage system is adjusted for efficiency from the grid perspective, with a baseline of zero output.

[0045] Photovoltaic (PV) power generation units: 5 units in total, with power scaling factors of 50, 70, 50, 70, and 50 respectively. The output can be adjusted within 80% to 100% of the predicted maximum value.

[0046] Electric vehicle charging units (EVs): 50 units in total, divided into two categories based on charging behavior. Category A (80%): Charging after get off work in the evening, arrival time follows a normal distribution with a mean of 19:00 and a standard deviation of 2 hours, departure time is around 8:00 the next day, reaching a mean SOC of 50%, with a target SOC of 90% or 99%. Category B (20%): Charging during working hours in the daytime, arrival time is around 8:30, and parking for 7-9 hours. All EVs have a rated charging power of 7kW, a charging efficiency of 0.95, and battery capacities randomly selected from 50 / 60 / 80kWh. Energy boundaries are calculated using the ASAP / ALAP strategy, and the baseline uses the maximum power charging strategy from the current moment.

[0047] Temperature Controlled Load Units (TCLs): 40 units in total. Building thermal parameters are read from actual building data, including the equivalent heat loss coefficient (HBLD), equivalent heat capacity (CBLD), and rated heat pump power (PRT). A uniform energy efficiency ratio (COP) of 3.85 is set, and the set temperature... Temperature control dead zone The baseline adopts a strategy of maintaining the set temperature.

[0048] Simulation key parameters: initial Power learning rate Energy learning rate Error tolerance Used to determine whether a parameter update is triggered.

[0049] like Figure 2 The diagram illustrates the evolution of tracking error and scaling factor. In the initial stage, due to... The external approximation of the corresponding superimposed boundary results in frequent and significant tracking errors. As online feedback correction progresses, and As the error is gradually reduced, it becomes sparse and gradually diminishes, and the flexibility range tends to be conservative. Ultimately, It converged to 0.711. The convergence to 0.722 indicates that approximately 70% of the original flexibility is sufficient to almost eliminate tracking error. It is worth noting that... and Each decrease in ...

[0050] like Figure 3As shown, to demonstrate the advantages of this method, the achieved power trajectory is compared with existing methods, including: i) the PEB external approximation method, which directly uses the superimposed boundary without scaling; and ii) the PB internal approximation method, which is a time-decoupled power boundary method. Under the same per-unit scheduling signal, compared with the PEB external approximation, this method produces a moderately contracted power envelope, reflecting... and The contraction effect is to reduce tracking error. Although the PB approximation method performs similarly to this method in the initial stage, its envelope contracts excessively over time, indicating that it is overly conservative compared to this method, which weakens the flexibility of the polymer supply.

[0051] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. An online feedback aggregation method for distributed energy resources with flexible time coupling, characterized in that, include: Obtain the state information of each distributed resource within the aggregate at the current moment, and calculate the independent flexibility region of each resource in the future look-ahead time domain based on the power-energy boundary model; The power boundary and energy boundary of each resource are arithmetically superimposed to obtain the theoretical aggregated power boundary and energy boundary; Introducing power scaling factor and energy scaling factor Centered on the polymerization power baseline, the theoretical polymerization boundary is expressed as a deviation form and parametrically shrunk to construct an approximate polymerization flexibility region; The system receives the scheduling signal determined by the system operator within the approximate aggregation flexibility region and extends the single-step scheduling signal to a complete power trajectory covering the entire look-ahead time domain; The complete power trajectory is decomposed into the independent flexibility regions of each resource to obtain the actual executable aggregated power trajectory and tracking error; When the tracking error exceeds the preset tolerance, identify the overly optimistic set of active constraints, construct the loss function, and calculate... and The approximate gradient is obtained, and the nonnegative projection is applied using the gradient descent method to update the parameters; The first step of the decomposed power command is executed, the status of each resource is updated, the time step is advanced, and the above process is repeated with the updated parameters and status to form a closed-loop control.

2. The online feedback aggregation method for time-coupled flexibility of distributed energy resources as described in claim 1, characterized in that, The distributed resources include energy storage systems, photovoltaic power generation units, electric vehicle charging units, and temperature-controlled load units.

3. The online feedback aggregation method for the time-coupled flexibility of distributed energy resources as described in claim 2, characterized in that, The aggregate power baseline is the power curve of the aggregate when it does not participate in system-level control, and is composed of the sum of the resource baselines; the baseline of the energy storage system is zero output, the baseline of the electric vehicle is charging at maximum power from the current moment until the energy demand is met, the baseline of the photovoltaic unit is its maximum power generation curve, and the baseline of the temperature-controlled load is maintaining the set temperature.

4. The online feedback aggregation method for the time-coupled flexibility of distributed energy resources as described in claim 3, characterized in that, The specific method of expanding the single-step scheduling signal into a complete power trajectory is as follows: using a random vector A linear programming model is constructed for the coefficients of the objective function, and the power in the first step is constrained to be equal to the scheduling signal value. The model is then solved within the contracted aggregation flexibility region to obtain the complete power trajectory located on the boundary of the feasible region.

5. The online feedback aggregation method for the time-coupled flexibility of distributed energy resources as described in claim 4, characterized in that, The specific set of active constraint identification is as follows: Let... The optimal solution to the decomposition problem is the executable trajectory. For the target trajectory, given the power constraint matrix The Okay, if Then it will be included in the active set. For the energy constraint matrix The Okay, if Then it will be included in the active set. .

6. The online feedback aggregation method for the time-coupled flexibility of distributed energy resources as described in claim 5, characterized in that, The loss function is defined as the trajectory from the active constraint hyperplane to the executable trajectory. Normalized sum of squared distances: ;in The approximate gradient is: ; The approximate gradient is: .

Images