Decentralized AGC Method and System for Virtual Power Plant Based on Exact Diffusion Algorithm

Abstract

The present invention discloses a decentralized AGC method and system for a virtual power plant based on an exact diffusion algorithm. According to the regulation upper and lower limits of each resource and the actual line capacity constraint, the present invention designs an AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance; decouples the Lagrange multipliers of the model, and uses the exact diffusion algorithm to achieve fast and accurate solution of the instructions; finally, an intelligent agent communication model with communication noise. The exact diffusion algorithm is rewritten and a gain function is introduced to enable the exact diffusion algorithm to have the ability to translate noise and improve the robustness of the algorithm.

Description

Technical Field

[0001] The present invention belongs to the field of automatic generation control of virtual power plants, and particularly relates to a decentralized AGC method and system for virtual power plants based on an accurate diffusion algorithm. Background Art

[0002] Virtual power plant technology is an important solution for internal coordinated control and optimized management of flexible distributed resources. It widely promotes the utilization of internal flexible resources, contributes to grid balance, reduces the dependence on traditional fossil fuel power generation modes, and accelerates the advancement of the "zero carbon emission" goal. In order to maintain the stability of real-time frequency and tie-line exchange power, automatic generation control (AGC) is a necessary tool for balancing input disturbances. With the trend of digitalization and decentralization in the operation of energy systems, the centralized system has limited ability to process massive data and is also difficult to meet the privacy requirements of units. Therefore, the decentralized AGC system has become a new trend. The control cycle of AGC is short, requiring the decentralized algorithm to have the advantages of fast convergence speed and high accuracy.

[0003] In addition, as the power generation resources mainly based on clean energy such as photovoltaic and wind turbines in virtual power plants gradually increase, their intermittency and volatility will cause stronger frequency fluctuations and power mismatches. Since the AGC control cycle is short (1 - 16 s) and the economic dispatch (ED) control cycle is long (5 - 15 min), when the penetration rate of renewable energy in virtual power plants is high, the situation of exceeding the line capacity may occur. Currently, the line capacity is rarely taken as a major constraint in the optimization solution of AGC instruction decomposition. In addition, the AGC instruction allocation process usually only targets the lowest frequency fluctuation, rarely taking the operating cost and frequency modulation quality as dual objectives to comprehensively design the instruction allocation algorithm, and ignoring the analysis of the AGC control effect of virtual power plants under the combined action of operating cost and frequency modulation quality.

[0004] In addition, the decentralized system will encounter the situation of communication noise, threatening the stability of AGC. Therefore, it is also essential to discuss communication noise and suppression methods in distributed algorithms. Summary of the Invention

[0005] The present invention makes improvements in view of the problems existing in the above-mentioned prior art, and provides a virtual power plant decentralized AGC method and system based on an exact diffusion algorithm, constructs a multi-agent AGC control system for a virtual power plant with a three-layer structure, and models communication lines and electrical connections; and takes the power adjustment upper and lower limits of each frequency modulation resource and the actual line capacity constraint as important constraint conditions, designs an AGC instruction decomposition model with the optimal operation cost and the highest frequency modulation performance as the dual optimization objectives, and realizes the reduction of the operation cost and the improvement of the frequency modulation performance. For this model, the Lagrange multipliers in the optimization model are decoupled, and the exact diffusion algorithm is used to quickly and accurately solve the instructions in each control period. Finally, in view of the problem of communication noise in the decentralized system, a gain function is introduced into the exact diffusion algorithm to effectively suppress the noise.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions, including:

[0007] A virtual power plant decentralized AGC method based on an exact diffusion algorithm, including:

[0008] Obtain the current operating parameters, control parameters of the previous control period, and frequency modulation performance evaluation related parameters of each distributed resource in the virtual power plant;

[0009] Take each distributed resource as an AGC unit, and perform primary frequency modulation and secondary frequency modulation based on the current operating parameters and frequency modulation performance evaluation related parameters of each obtained AGC unit to jointly maintain the frequency stability of the virtual power plant; among them, the secondary frequency modulation is to solve the actual power instructions of each distributed resource in the next control period by using an AGC instruction decomposition model with the optimal operation cost and the highest frequency modulation performance as the dual optimization objectives based on the current operating parameters, control parameters of the previous control period, and frequency modulation performance evaluation related parameters of each distributed resource in the virtual power plant; then each distributed resource, that is, the AGC unit, controls the power according to the corresponding actual power instruction.

[0010] Among them, the constraints considered when solving the AGC instruction decomposition model with the optimal operation cost and the highest frequency modulation performance as the dual optimization objectives include power supply and demand balance constraints, power adjustment upper and lower limit constraints of each distributed resource, and line capacity constraints.

[0011] Further, among them, the current operating parameters include the current power level of the AGC unit, the AGC power adjustment upper and lower limits at this moment, etc.;

[0012] Among them, the frequency modulation performance evaluation indicators include response ramp rate, response time, and response accuracy evaluation indicators, which are respectively expressed as follows:

[0013] The frequency modulation quality formula is:

[0014]

[0015] In the formula, represents the response ramp rate evaluation index of the i-th AGC unit, is the actual ramp rate of the i-th AGC unit, is the actual power output adjustment amount of the i-th AGC unit, is the starting moment of ramp rate calculation, is the ending moment of ramp rate calculation, is the average standard regulation rate of the frequency modulation resources of the virtual power plant;

[0016]

[0017] In the formula, represents the response time evaluation index of the i-th AGC unit, is the actual delay error of the i-th AGC unit, is the moment when the i-th AGC unit responds, is the moment when the regulation command is sent to the i-th AGC unit, is the allowable delay error for the frequency modulation resources to respond to the regulation command;

[0018]

[0019] In the formula, represents the regulation accuracy index, is the actual regulation error of the i-th distributed resource, is the moment when the i-th distributed resource enters the target dead zone, is the ending moment of the regulation accuracy calculation of the i-th distributed resource, P standard is the allowable regulation error for the frequency modulation resources to respond to the AGC controller regulation command.

[0020] Comprehensively These three evaluation indexes, the comprehensive evaluation index of the frequency modulation quality of the i-th distributed resource of the virtual power plant is s i , s i The specific formula of is as follows:

[0021]

[0022] Furthermore, the objective function of the AGC instruction decomposition model is expressed as follows:

[0023]

[0024] In the formula, C(i) represents the objective function of the i-th AGC unit, ΔT is the implementation period of each control interval, t represents the current time point, t = ΔT, 2*ΔT,..., N*ΔT, and N is the total number of control periods; ai , b i , c i is the operating cost coefficient of each AGC unit, represents the actual power command of the i-th AGC unit at time t, represents the initial power of the i-th AGC unit at time t; is the regulation mileage command of the i-th AGC unit at time t; λ is the specified mileage price, σ i = 1 - s i represents the performance deficiency index.

[0025] Furthermore, the exact diffusion algorithm is used to solve the AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance in each control cycle, specifically including:

[0026] Rewrite the objective function, power supply-demand balance constraint, and line capacity constraint of the AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance into the form of a Lagrangian dual function, and decouple the Lagrange multipliers therein;

[0027] Initialize the Lagrange multipliers, iteration step size, and number of iterations;

[0028] Update the Lagrange multipliers using the exact diffusion algorithm based on the currently obtained operating parameters of each distributed resource and the relevant parameters for frequency modulation performance evaluation;

[0029] Calculate the actual power command of the i-th frequency modulation AGC unit at the best time t corresponding to the (k + 1)-th iteration round according to the Lagrange multiplier update result of the intermediate variable Based on the intermediate variable Combine the upper and lower limits of power adjustment of each distributed resource for correction to obtain the actual power command of the i-th frequency modulation AGC unit at the best time t corresponding to the (k + 1)-th iteration round;

[0030] Repeat the iteration according to the Lagrange multiplier update result until the set number of iterations is reached to obtain the final actual power command of the i-th frequency modulation AGC unit at time t.

[0031] Furthermore, the rewriting of the objective function, power supply-demand balance constraint, and line capacity constraint of the AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance into the form of a Lagrangian dual function, and the decoupling of the Lagrange multipliers therein are expressed as follows:

[0032]

[0033] λ = [λ1 ,…, λ i ,…, λ n ;

[0034]

[0035] τ l = τ l,1 ,…, τ l,i ,…, τ l,n ;

[0036] In the formula, λ i is the Lagrange multiplier of the decoupled equality constraint, τ l,i respectively represent the Lagrange multipliers of the decoupled inequality constraints, C(i) represents the objective function of the i-th AGC unit, is the regulation mileage command of the i-th AGC unit; λ i , τ l,i respectively represent the Lagrange multiplier of the equality constraint and the Lagrange multiplier of the inequality constraint corresponding to the i-th AGC unit; ρ li represents the power contribution of the i-th AGC unit to line l. If the i-th AGC unit contributes power to line l, then ρ li = 1, otherwise ρ li = 0; P ARR is the area control demand value of the AGC control system; n is the number of AGC units, and m is the total number of lines.

[0037] Furthermore, based on the obtained current operating parameters of each AGC unit and the parameters related to the frequency modulation performance evaluation, the Lagrange multiplier is updated using the exact diffusion algorithm, and the update formula is as follows:

[0038]

[0039] φ λ,i (k + 1)= ψ λ,i (k + 1)+ λ i (k)- ψ λ,i (k)

[0040]

[0041]

[0042] where k is the iteration number of the exact diffusion algorithm;

[0043] where, its calculation method is specifically:

[0044]

[0045] Among them, 1 N is the identity matrix, and Φ W =[ω i,j ∈ R n×n represents the set of line weights. If (i, j) ∈ Φ E , then ω i,j > 0, indicating that the ith AGC unit and the jth AGC unit are neighbors of each other; if then ω i,j = 0, and Φ E represents the set of communication lines; the specific weight definition is as follows:

[0046]

[0047] Among them, n i , n j respectively represent the number of neighbor nodes of the AGC unit i and the AGC unit j.

[0048] λ i (k) is the intermediate variable of the kth update of λ i , u i is the iteration step, and ψ λ,i (k + 1) is the intermediate variable matrix of the Lagrange multiplier λ updated by the gradient method. ψ λ,i (0)=λ i (0), and φ λ,i (k + 1) is the intermediate variable matrix of the Lagrange multiplier λ after correcting the imbalance; are respectively the intermediate variable matrices of the Lagrange multiplier and τ l,i updated by the gradient method, and are respectively the intermediate variable matrices of the Lagrange multiplier and τ l,i after correcting the imbalance.

[0049] Furthermore, for the case with communication noise, an improved algorithm is proposed based on the above-mentioned exact diffusion algorithm by introducing a gain function method to effectively shift the noise. The exact diffusion algorithm updates the Lagrange multiplier, and the update formula is expressed as follows:

[0050]

[0051]

[0052] Among them, k is the iteration number of the exact diffusion algorithm, 1 N is the identity matrix, and Φ W =[ω i,j ∈ R n×nDenote the set of line weights. If (i, j) ∈ Φ E , then ω i,j > 0, indicating that the i-th AGC unit and the j-th AGC unit are neighbors to each other; if then ω i,j = 0, and Φ E denotes the set of communication lines; λ i (k) is an intermediate variable updated at the k-th time of λ i , u i is the iteration step size, and ψ λ,i (k + 1) is an intermediate variable matrix obtained by updating the Lagrange multiplier λ using the gradient method, and ψ λ,i (0) = λ i (0), and φ λ,i (k + 1) is an intermediate variable matrix of the Lagrange multiplier λ after correcting the imbalance; are respectively the intermediate variable matrices obtained by updating the Lagrange multipliers and τ l,i using the gradient method, are respectively the intermediate variable matrices of the Lagrange multipliers and τ l,i after correcting the imbalance; η ij is Gaussian white noise following a normal distribution, and c λ (k) is the gain function when updating λ i , and it needs to satisfy the following two necessary conditions: ① ② is the gain function when updating , and it needs to satisfy the following two necessary conditions: ① ② is the gain function when updating τ l,i , and it needs to satisfy the following two necessary conditions: ① ②

[0053] A virtual power plant decentralized AGC system based on an exact diffusion algorithm, comprising:

[0054] Communication layer: The communication layer consists of n agents, and each agent corresponds to an AGC unit one by one; each agent calculates the frequency modulation performance of the local AGC unit according to the power data provided by the corresponding intelligent acquisition device, is responsible for communication between neighbors, and updates the actual power command of the local AGC unit in the next control cycle according to the virtual power plant decentralized AGC method based on the exact diffusion algorithm, and gives the result to the middle control layer;

[0055] Control layer: The control layer consists of n controllers. The controllers control the device layer according to the actual power instructions for the next control period calculated by the corresponding agents in the communication layer.

[0056] Device layer: The device layer consists of AGC units and corresponding intelligent acquisition devices. The intelligent acquisition devices are used to obtain the current operating parameters, control parameters of the previous control period, and parameters related to the frequency regulation performance evaluation of the corresponding AGC units.

[0057] An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method for decentralized AGC of a virtual power plant based on an exact diffusion algorithm is implemented.

[0058] A storage medium containing computer-executable instructions, which implement the method for decentralized AGC of a virtual power plant based on an exact diffusion algorithm when executed by a computer processor.

[0059] A computer program product, comprising a computer program / instructions, which implement the steps of the method for decentralized AGC of a virtual power plant based on an exact diffusion algorithm when executed by a processor.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] (1) A three-layer structure of the AGC control system of the virtual power plant is proposed, and the communication topology and grid structure are modeled to facilitate decentralized control.

[0062] (2) In the AGC instruction decomposition problem, a multi-objective optimization problem considering operating costs and frequency regulation performance is constructed, and the line capacity constraint is used as one of the constraints of the multi-objective optimization problem to prevent the frequency regulation unit from exceeding the line capacity during power adjustment.

[0063] (3) The line capacity constraint that was originally coupled together is decoupled, and the exact diffusion algorithm is used for solution. The exact diffusion algorithm has higher convergence accuracy and convergence speed compared with the consensus algorithm and the traditional diffusion algorithm. The proposed method is not prone to single-point failures, has high scalability and flexibility, and protects the privacy of the AGC units.

[0064] (4) The noise occurring in the communication process is taken into account, and an intelligent agent communication model containing communication noise is constructed. The exact diffusion algorithm is rewritten and a gain function c(k) is introduced to enable the exact diffusion algorithm to have the ability to translate noise and improve the robustness of the algorithm. Description of the Drawings

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings of embodiments can be obtained based on these drawings.

[0066] Figure 1 It is a control scheme diagram of a virtual power plant decentralized AGC method based on an accurate diffusion algorithm of the present invention;

[0067] Figure 2 It is an algorithm flow chart for decentralized solution using the accurate diffusion algorithm;

[0068] Figure 3 It is a result diagram of the power adjustment values of each AGC unit solved by the accurate diffusion algorithm;

[0069] Figure 4 It is a system frequency fluctuation diagram under a given input perturbation, with the abscissa being time (s);

[0070] Figure 5 It is a system operating cost change diagram under a given input perturbation, with the abscissa being time (s). Detailed implementation manners

[0071] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the following further details the present invention in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0072] This application will use the terms commonly adopted by those skilled in the art to describe various aspects of the illustrative embodiments to convey the essence of their work to other technicians in the field. However, it is obvious to those skilled in the art that some alternative embodiments can be practiced using parts of the described aspects. For the purpose of explanation, specific numbers, materials and configurations are elaborated to provide a thorough understanding of the illustrative embodiments. However, it is obvious to those skilled in the art that alternative embodiments can be implemented without specific details. In other cases, some well-known features are omitted or simplified to avoid confusing the illustrative embodiments.

[0073] The present invention provides a decentralized AGC method for a virtual power plant based on an exact diffusion algorithm. In the virtual power plant, distributed resources serve as AGC units, and each AGC unit is provided with a corresponding agent, controller, and intelligent acquisition device. The intelligent acquisition device is used to obtain the current operating parameters, control parameters of the previous control cycle, and parameters related to frequency modulation performance evaluation of the corresponding AGC unit. The agent calculates the frequency modulation performance of the corresponding AGC unit based on the power data provided by the corresponding intelligent acquisition device, is responsible for communication between neighbors, and updates the actual power command for the corresponding AGC unit in the next control cycle according to the decentralized AGC method for a virtual power plant based on an exact diffusion algorithm of the present invention, and gives the result to the intermediate controller. The controller controls the device layer according to the actual power command for the next control cycle calculated by the corresponding agent in the communication layer. Specifically, in the decentralized AGC method for a virtual power plant based on an exact diffusion algorithm of the present invention, the power adjustment upper and lower limits of each distributed resource and the actual line capacity constraint are used as important constraint conditions, and an AGC command decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance is designed to achieve the reduction of operating cost and the improvement of frequency modulation performance. For this model, the Lagrange multipliers in the optimization model are decoupled, and the exact diffusion algorithm is used to achieve fast and accurate solution of the command in each control cycle. Finally, aiming at the problem of communication noise in the decentralized system, a gain function is introduced into the exact diffusion algorithm to effectively suppress the noise.

[0074] This embodiment is introduced by taking a virtual power plant composed of n distributed resources as an example. A decentralized AGC method for a virtual power plant based on an exact diffusion algorithm, as Figure 1 shown, includes:

[0075] S1, obtaining the current operating parameters, control parameters of the previous control cycle, and parameters related to frequency modulation performance evaluation of each distributed resource in the virtual power plant;

[0076] Among them, the distributed resources in the virtual power plant serve as each AGC unit, and communication and line connection can be carried out between each AGC unit according to requirements to realize the utilization of internal flexible resources and control the grid balance; in a specific implementation scheme, the communication topology model can be represented by an undirected graph G = {Φ N , Φ E , Φ W}. Among them, Φ N = {1, 2,..., n}, representing the set of agent nodes in the virtual power plant. At the same time, since the agent and the AGC unit numbers correspond one by one, n also represents the number of AGC units; Φ E represents the set of communication lines, and (i, j) ∈ Φ E indicates that nodes i and j can communicate; Φ W = [ω i,j ∈ Rn×n Denote the set of line weights. If (i, j) ∈ Φ E , then ω i,j > 0, indicating that they are neighbors of each other; if then ω i,j = 0. The specific weight definition is as follows:

[0077]

[0078] where n i , n j represent the number of neighbor nodes of node i and node j respectively.

[0079] Furthermore, the grid structure formed by line connections can be represented by ρ = [ρ li ∈ R m×n , where l = 1, …, m, m represents the total number of lines in the virtual power plant, ρ represents the power contribution matrix of the distributed resource, i.e., the AGC unit i, to line l. If the AGC unit i contributes to the power of line l, then ρ li = 1; otherwise, ρ li = 0.

[0080] The parameter data obtained in the present invention is mainly used for primary frequency modulation and secondary frequency modulation solution calculations. Generally, the current operating parameters of each distributed resource include the current power level of the AGC unit, the upper and lower limits of AGC power adjustment at this moment, and the control parameters of the previous control cycle include the actual power command and the regulation mileage command of the previous control cycle, etc. The parameters related to frequency modulation performance evaluation include the response ramp rate, response time, regulation accuracy evaluation index, etc., which are respectively represented as follows:

[0081]

[0082] In the formula, represents the response ramp rate evaluation index of the i-th AGC unit, is the actual ramp rate of the i-th AGC unit, is the actual power output adjustment amount of the i-th AGC unit, is the starting moment of the ramp rate calculation of the i-th AGC unit, is the ending moment of the ramp rate calculation of the i-th AGC unit, is the average standard regulation rate of the frequency modulation resources in the virtual power plant;

[0083]

[0084] In the formula, represents the response time evaluation index of the i-th AGC unit, is the actual delay error of the i-th AGC unit, is the moment when the i-th AGC unit responds and acts. is the moment when the adjustment command is sent to the i-th AGC unit. is the delay error allowed for the frequency modulation resource to respond to the adjustment command.

[0085]

[0086] In the formula, represents the adjustment accuracy index. is the actual adjustment error of the i-th AGC unit. is the moment when the i-th AGC unit enters the target dead zone. is the termination moment of the adjustment accuracy calculation of the i-th AGC unit, P standard is the allowed adjustment error for the frequency modulation resource to respond to the AGC controller adjustment command.

[0087] Comprehensively These three evaluation indicators, the comprehensive evaluation index of the frequency modulation quality of the i-th distributed resource in the virtual power plant is s i , s i The specific formula of is as follows:

[0088]

[0089] S2, each distributed resource is used as an AGC unit, and primary frequency modulation and secondary frequency modulation are performed based on the currently obtained operating parameters of each distributed resource and the parameters related to the frequency modulation performance evaluation to jointly maintain the frequency stability of the virtual power plant; among them, the secondary frequency modulation is based on the currently obtained operating parameters of each distributed resource in the virtual power plant, the control parameters of the previous control cycle, and the parameters related to the frequency modulation performance evaluation, and the actual power command of each distributed resource in the next control cycle is obtained by solving the AGC instruction decomposition model with the dual optimization goals of the optimal operating cost and the highest frequency modulation performance.

[0090] The schematic diagram of the method for the virtual power plant to participate in AGC control in the present invention is as Figure 1 shown, ΔP out is the total adjustment amount of the actual power output of all AGC units, s represents the complex variable in the Laplace transform, Δf represents the frequency deviation of the AGC control system, M represents the equivalent inertia constant, D represents the equivalent damping coefficient, ACE represents the area control error, ΔP loadThe load disturbance representing the system input, B represents the frequency deviation coefficient. The AGC control system maintains the frequency stability of the virtual power plant under the combined action of primary frequency modulation and secondary frequency modulation. For primary frequency modulation, once the system frequency deviates from the rated value, the units capable of primary frequency modulation will automatically control the increase or decrease of the active power of the units according to the frequency deviation of the AGC control system to limit the change of the grid frequency, which is a droop control with a frequency offset. For secondary frequency modulation, after receiving the load disturbance input to the system, the ACE is calculated, and then the area control demand ARR is obtained through the action of the PI controller. In the present invention, through the communication and calculation of the communication layer proxy, the optimal solution of the optimization problem is decentralizedly solved, the AGC commands of each AGC unit are calculated, and then each distributed resource in the device layer is guided by the control layer to adjust to the specified power. In each control cycle, with the lowest operating cost and the highest frequency modulation performance, a dual-objective optimization AGC command decomposition model is designed:

[0091] The proposed AGC command decomposition model takes into account the lowest operating cost and the highest frequency modulation performance. Among them, the operating cost objective function C1 is specifically as follows:

[0092]

[0093] Among them, C1(i) represents the operating cost objective function of the i-th AGC unit, ΔT is the implementation period of each control interval, generally 1 - 16 s, t represents the current time point, t = ΔT, 2*ΔT,..., N*ΔT, and N is the total number of control cycles. a i , b i , c i are the operating cost coefficients of each AGC unit, represents the actual power command of the i-th AGC unit at time t, represents the initial power of the i-th AGC unit at time t; is the regulation mileage command of the i-th AGC unit at time t.

[0094] Furthermore, in order to encourage resources with high frequency modulation performance to make greater resource adjustments, it is necessary to give them a larger adjustment command input. The frequency modulation performance objective function C2 is specifically as follows:

[0095]

[0096] Among them, C2(i) represents the frequency modulation performance objective function of the i-th AGC unit, λ is the specified mileage price, s i is the frequency modulation performance score of each AGC unit, which is linearly weighted and obtained by real-time evaluation of three indicators: response ramp rate, response time, and response accuracy, as shown in Equation (8).

[0097] To ensure that C2 is convenient for subsequent gradient-based solution, Equation (10) is written in the following form:

[0098]

[0099] where σ i = 1 - s i , since s i ∈[0, 1], so σ i ∈[0, 1], representing the Performance Insufficiency Index, which measures the part where the frequency modulation performance score does not reach the maximum value. In addition, the squared absolute value function maintains the property of a convex function and is differentiable everywhere.

[0100] The objective function of the AGC instruction decomposition model proposed by the present invention can be specifically written as:

[0101]

[0102] where C(i) = C1(i) + C2(i).

[0103] Furthermore, a set of constraint functions is constructed. The construction of the set of constraint functions considers the power supply and demand balance constraint, the upper and lower limits of power adjustment of each AGC unit, and the line capacity constraint.

[0104] Furthermore, the power supply and demand balance constraint is:

[0105]

[0106] where P ARR (t) is the regional control demand value of the AGC control system at time t.

[0107] Furthermore, the upper and lower limits of resource adjustment constraint are:

[0108]

[0109] where and respectively represent the lower and upper limits of power adjustment of the i-th AGC unit at time t.

[0110] Furthermore, the line capacity constraint is:

[0111]

[0112] where and P l are respectively the lower and upper limits of the power flow of the l-th line, and m is the total number of lines in the virtual power plant.

[0113] In a specific embodiment, the Lagrange multipliers in the optimization model can be decoupled, and an exact diffusion algorithm can be used to achieve fast and accurate solution of instructions in each control cycle, specifically including:

[0114] First, write equations (12), (13), and (14) in step S2 in the form of the Lagrangian dual function, and then update the Lagrange multipliers through a distributed algorithm to achieve optimization solution, which is a common means of solving by the distributed algorithm. The rewritten Lagrangian function is as follows:

[0115]

[0116] Among them, λ is the specified mileage price, serving as the Lagrange multiplier for equality constraints, and τ l ≥0 are the Lagrange multipliers for inequality constraints. However, since the power supply-demand balance constraint and the line capacity constraint in the optimization problem are a coupled constraint, that is, the constraint requires the combination of the powers of several AGC units and cannot be solved decentralized for this optimization problem. Therefore, the constraint is decoupled:

[0117]

[0118] λ is decomposed into λ = [λ 1 , …, λ i , …, λ n . Similarly, among them, and τ l are decomposed into τ l = τ l,1 , …, τ l,i , …, τ l,n . The model for decomposing distributed AGC instructions can be obtained by the above decoupling method:

[0119]

[0120] If is the optimal solution of the original problem, when the power constraint (14) is not considered, needs to satisfy the following conditions for the optimal solution:

[0121]

[0122] Furthermore, the algorithm flow chart for decentralized solution using the exact diffusion algorithm is as Figure 2 shown, specifically as follows:

[0123] Obtain relevant parameters, including calculating and updating the frequency modulation performance score by introducing the power measurement values at the device layer; update the communication line set according to the current communication topology structure, and update the line weight set according to the grid structure;

[0124] Initialize the Lagrange multiplier, the iteration step size u, and the number of iterations k;

[0125] Based on the currently obtained operating parameters of each distributed resource and the relevant parameters for frequency regulation performance evaluation, use the exact diffusion algorithm to update the Lagrange multiplier. The method for updating the Lagrange multiplier by the exact diffusion algorithm is as follows:

[0126]

[0127] where k is the number of iterations of the exact diffusion algorithm, is an intermediate variable, ψ λ,i (0) = λ i (0), 1 N is the identity matrix, λ i (k) is the intermediate variable of λ i at the k-th update, u i is the iteration step size, ψ λ,i (k + 1) is the intermediate variable matrix of the Lagrange multiplier λ updated by the gradient method, φ λ,i (k + 1) is the intermediate variable matrix of the Lagrange multiplier λ after correcting the imbalance. It is necessary to ensure that the intermediate variable of λ is always > 0 during the update. and τ l,i in each round of update and λ i are similar, which is expressed as follows:

[0128]

[0129] Similarly, are the intermediate variable matrices of the Lagrange multipliers and τ l,i updated by the gradient method respectively, are the intermediate variable matrices of the Lagrange multipliers and τ l,i after correcting the imbalance respectively. Similarly, it is necessary to ensure that and τ l,i the intermediate variables are always > 0.

[0130] In the algorithm flowchart, after each round of Lagrange multiplier update is completed, it is necessary to calculate the actual power command of the i-th frequency regulation AGC unit at the best t moment corresponding to the round k + 1 according to Equation (24) of the intermediate variable and correct it according to the upper and lower limit constraints of to obtain the final The specific formula is as follows:

[0131]

[0132] Furthermore, the proposed power correction calculation method is as follows:

[0133]

[0134] Repeat the iteration according to the Lagrange multiplier update result until the set number of iterations is reached, and obtain the actual power command of the i-th frequency modulation AGC unit at the best t moment.

[0135] In a preferred embodiment, based on the exact diffusion algorithm in step S3, a method of introducing a gain function into the exact diffusion algorithm is proposed to effectively suppress noise:

[0136] Introducing noise into the proposed exact diffusion algorithm, the third equation of Equation (21) can be rewritten in the following form:

[0137]

[0138] Furthermore, a gain function c(k) is introduced into the exact diffusion algorithm to suppress the influence of communication noise, and its expression is as follows:

[0139]

[0140] η ij is Gaussian white noise subject to a normal distribution, and c λ (k) needs to satisfy the following two necessary conditions: ① ② The noise suppression method adopted in this embodiment is τ l,i The same principle applies, and the expression is as follows:

[0141]

[0142] The following further illustrates the effect of the present invention in combination with a specific embodiment:

[0143] In this embodiment, the AGC control system of a virtual power plant composed of 5 distributed resources, namely 1 photovoltaic, 1 wind turbine, 1 small hydropower, 1 energy storage, and 1 adjustable load, is used to verify the effectiveness of the method and system. Modeling is carried out on the MATLAB / Simulink platform, and a 30MW disturbance is input at 200s of system simulation. Figure 3 The instruction allocation situation of the 5 AGC units in the AGC control system is shown. It can be seen that there are differences in the frequency modulation performance and operating costs among different AGC units, which will affect the output size of the power. In addition, the method of centralized fixed ratio allocation and the traditional diffusion algorithm are compared with the method proposed in the present invention. Figure 4The frequency deviation is calculated by three methods. It can be seen that the centralized fixed - ratio allocation method cannot improve the frequency modulation performance because there is no dynamic optimization process, resulting in the largest frequency deviation. The accuracy of the traditional diffusion algorithm is lower than that of the precise diffusion algorithm, and it is also impossible to calculate the optimal solution of the frequency modulation performance. In contrast, the proposed method has the smallest frequency deviation for the system, proving the improvement of the proposed method in terms of frequency modulation performance. Figure 5 By comparing the operating costs of the three methods, it is found that the method proposed in the present invention also has the smallest operating cost, proving the reduction of the proposed method in terms of operating cost.

[0144] The present invention also provides a decentralized AGC system for a virtual power plant based on the precise diffusion algorithm. The system has a three - layer structure, specifically from top to bottom:

[0145] Communication layer: The communication layer is composed of agents (multi - agent system), which mainly has three major tasks: 1) Calculate the frequency modulation performance of the AGC unit according to the power data provided by the intelligent acquisition device; 2) Be responsible for communication between neighbors; 3) Update local variables through the above - mentioned precise diffusion algorithm and give the power adjustment result to the middle control layer.

[0146] Control layer: It includes n controllers. The end - side controllers of each AGC unit control the device layer according to the power adjustment value calculated by the communication layer. At the same time, the primary control is also completed in the control layer.

[0147] Device layer: The device layer mainly includes AGC units and intelligent acquisition devices. The agents in the communication layer extract power measurement values from the device layer.

[0148] Corresponding to the foregoing embodiment of a decentralized AGC method for a virtual power plant based on the precise diffusion algorithm, the present invention also provides an electronic device.

[0149] An electronic device provided by an embodiment of the present invention includes one or more processors for implementing a decentralized AGC method for a virtual power plant based on the precise diffusion algorithm in the above - mentioned embodiment.

[0150] The embodiment of the electronic device of the present invention can be applied to any device with data - processing capabilities, and the any device with data - processing capabilities can be a device or apparatus such as a computer.

[0151] The device embodiments can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a logically defined device, it is formed by a processor in any device with data processing capabilities reading the corresponding computer program instructions in the non-volatile memory into the memory for execution. In terms of hardware, it includes a processor, memory, network interface, and non-volatile memory. In addition, any device with data processing capabilities where the device in the embodiment is located usually includes other hardware according to the actual functions of the device with data processing capabilities, which will not be elaborated here.

[0152] For the implementation processes of the functions and roles of each unit in the above device, please refer to the implementation processes of the corresponding steps in the above method for details, which will not be elaborated here.

[0153] For the device embodiments, since they basically correspond to the method embodiments, please refer to the relevant descriptions in the method embodiments for the relevant parts. The device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0154] The embodiments of the present invention also provide a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements a virtual power plant decentralized AGC method based on an exact diffusion algorithm in the above embodiments.

[0155] The computer-readable storage medium can be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be any device with data processing capabilities, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device. Further, the computer-readable storage medium can also include both the internal storage unit of any device with data processing capabilities and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store data that has been output or will be output.

[0156] The above are only the preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A decentralized AGC method for virtual power plants based on an exact diffusion algorithm, characterized in that: include: Obtain the current operating parameters of each distributed resource in the virtual power plant, the control parameters of the previous control cycle, and the frequency regulation performance evaluation index; wherein the frequency regulation performance evaluation index includes the response ramp rate, response time, and response accuracy; Each distributed resource is used as an AGC unit, and primary and secondary frequency regulation are performed based on the current operating parameters and frequency regulation performance evaluation indicators of each AGC unit, so as to jointly maintain the frequency stability of the virtual power plant; wherein, the secondary frequency regulation is based on the current operating parameters of each distributed resource in the virtual power plant, the control parameters of the previous control cycle, and the frequency regulation performance evaluation indicators, and the AGC instruction decomposition model with the optimal operating cost and the highest frequency regulation performance as the dual optimization goals is used to solve and obtain the actual power instruction of each distributed resource in the next control cycle; then each distributed resource, i.e., the AGC unit, controls the power according to the corresponding actual power instruction; The constraints considered when solving the AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance include power supply and demand balance constraints, upper and lower limit constraints of power adjustment of each distributed resource, and line capacity constraints; The objective function of the AGC instruction decomposition model is expressed as follows: Where n represents the number of AGC units, C(i) represents the objective function of the i-th AGC unit, ΔT is the implementation period of each control interval, t represents the current time point, t = ΔT, 2*ΔT, …, N*ΔT, N is the total number of control periods; a i ,b i ,c i is the operating cost coefficient of each AGC unit, represents the actual power command of the i-th AGC unit at time t, represents the initial power of the i-th AGC unit at time t; is the adjustment mileage command of the i-th AGC unit at time t; σ i represents the performance deficiency indicator; σ i =1-s i ,s i represents the comprehensive evaluation index of frequency regulation quality of the i-th distributed resource in the virtual power plant; s i The specific formula is as follows: In the formula, Represents the response ramp rate evaluation index of the i-th AGC unit, is the actual ramp rate of the ith AGC unit, is the actual power output adjustment of the i-th AGC unit, Calculate the starting time for the ramp rate, The end time of the climbing rate calculation is The average standard regulation rate of the frequency regulation resources of the virtual power plant; In the formula, represents the evaluation index of the response time of the i-th AGC unit, is the actual delay error of the i-th AGC unit, is the time when the i-th AGC unit responds to the action, is the time when the adjustment command is issued to the i-th AGC unit, The delay error allowed for the frequency modulation resource to respond to the modulation command; In the formula, Represents the adjustment accuracy index, is the actual adjustment error of the i-th distributed resource, is the moment when the i-th distributed resource enters the target dead zone, is the termination time of the i-th distributed resource adjustment accuracy calculation, P standard The allowable adjustment error of the frequency modulation resource in response to the adjustment command of the AGC controller; The AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance is solved in each control cycle by using an accurate diffusion algorithm, specifically including: Rewrite the objective function of the AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency regulation performance, power supply and demand balance constraint and line capacity constraint into the form of Lagrangian dual function, and decouple the Lagrangian multipliers therein; Initialize Lagrange multiplier, iteration step and number of iterations; Based on the current operating parameters and frequency modulation performance evaluation indicators of each distributed resource, the Lagrange multiplier is updated using an accurate diffusion algorithm; According to the Lagrange multiplier update result, the actual power instruction of the i-th frequency modulation AGC unit at time t corresponding to the optimal iteration round k+1 is calculated. The intermediate variable Based on intermediate variables Combined with the upper and lower limits of each distributed resource power adjustment constraints, the actual power instruction of the i-th frequency modulation AGC unit at time t corresponding to the iteration round k+1 is corrected; Repeated iterations are performed according to the Lagrange multiplier update result until the set number of iterations is reached, and the final optimal actual power instruction of the i-th frequency modulation AGC unit at time t is obtained.

2. The method according to claim 1, characterized in that The objective function of the AGC instruction decomposition model with the dual optimization objectives of optimal operating cost and highest frequency modulation performance, power supply and demand balance constraint and line capacity constraint are rewritten into the form of Lagrangian dual function, and the Lagrangian multipliers therein are decoupled, which is expressed as follows: λ=[λ1,…,λ i ,…,l n ] t l ＝[ t l,1 ,…, t l,i ,…, t l,n ] In the formula, λ i is the Lagrange multiplier of the decoupled equality constraint, t l,i They represent the Lagrange multipliers of the decoupled inequality constraints, C(i) represents the objective function of the i-th AGC unit, is the adjustment mileage command of the i-th AGC unit; λ i , t l,i They represent the Lagrange multipliers of equality constraints and inequality constraints corresponding to the i-th AGC unit respectively; ρ li represents the power contribution of the i-th AGC unit to line l. If the i-th AGC unit contributes to the power of line l, then ρ li =1, otherwise ρ li =0;P ARR is the regional control demand value of the AGC control system; n is the number of AGC units, and m is the total number of lines.

3. The method according to claim 2, characterized in that Based on the current operating parameters of each AGC unit and the frequency modulation performance evaluation index, the Lagrange multiplier is updated using the precise diffusion algorithm. The updated formula is as follows: f λ,i (k+1)=ψ λ,i (k+1)+λ i (k)-ψ λ,i (k) Where k is the number of iterations of the exact diffusion algorithm; in, The specific calculation method is: Among them, 1 N is the identity matrix, Φ W =[ω i,j ]∈R n×n Represents the line weight set, if (i,j)∈Φ E , then ω i,j >0, it means that the i-th AGC unit and the j-th AGC unit are neighbors; if Then ω i,j =0,Φ E Represents a set of communication lines; the specific weights are defined as follows: Among them, n i ,n j Respectively represent the number of neighbor nodes of AGC unit i and AGC unit j; λ i (k) is λ i The intermediate variable u updated for the kth time i is the iteration step size, ψ λ,i (k+1) is the intermediate variable matrix updated by the Lagrange multiplier λ using the gradient method, ψ λ,i (0) = λ i (0),φ λ,i (k+1) is the intermediate variable matrix after the Lagrange multiplier λ corrects the imbalance; They are the Lagrange multipliers and t l,i The intermediate variable matrix updated using the gradient method, They are the Lagrange multipliers and t l,i The intermediate variable matrix after correcting the imbalance.

4. The method according to claim 2, characterized in that: The exact diffusion algorithm updates the Lagrange multiplier, and the update formula is expressed as follows: f λ,i (k+1)=ψ λ,i (k+1)+λ i (k)-ψ λ,i (k) Where k is the number of iterations of the exact diffusion algorithm, 1 N is the identity matrix, Φ W =[ω i,j ]∈R n×n Represents the line weight set, if (i,j)∈Φ E , then ω i,j >0, it means that the i-th AGC unit and the j-th AGC unit are neighbors; if Then ω i,j =0,Φ E represents the set of communication lines; i (k) is λ i The intermediate variable u updated for the kth time i is the iteration step size, ψ λ,i (k+1) is the intermediate variable matrix updated by the Lagrange multiplier λ using the gradient method, ψ λ,i (0) = λ i (0),φ λ,i (k+1) is the intermediate variable matrix after the Lagrange multiplier λ corrects the imbalance; They are the Lagrange multipliers and t l,i The intermediate variable matrix updated using the gradient method, They are the Lagrange multipliers and t l,i The intermediate variable matrix after correcting the imbalance; η ij is Gaussian white noise that follows a normal distribution, c λ (k) is the updated λ i The gain function when , needs to meet the following two necessary conditions: ① ② For Update The gain function when , needs to meet the following two necessary conditions: ① ② For Update t l,i The gain function when , needs to meet the following two necessary conditions: ① ② 5. A decentralized AGC system for a virtual power plant based on an accurate diffusion algorithm, characterized in that: include: Communication layer: The communication layer is composed of n intelligent agents, which correspond to each AGC unit one by one; each intelligent agent calculates the frequency modulation performance of the local AGC unit according to the power data provided by the corresponding intelligent collection device and is responsible for communication between neighbors and updating the actual power instruction of the next control cycle of the local AGC unit according to the virtual power plant decentralized AGC method based on the precise diffusion algorithm according to any one of claims 1 to 4, and gives the result to the middle control layer; Control layer: The control layer consists of n controllers, which control the device layer according to the actual power command of the next control cycle calculated by the corresponding intelligent agent in the communication layer; Equipment layer: The equipment layer consists of AGC units and corresponding intelligent acquisition devices. The intelligent acquisition devices are used to obtain the current operating parameters of the corresponding AGC units, the control parameters of the previous control cycle, and the frequency modulation performance evaluation indicators.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements a decentralized AGC method for a virtual power plant based on an exact diffusion algorithm as described in any one of claims 1-4.

7. A storage medium comprising computer executable instructions, which, when executed by a computer processor, implement a decentralized AGC method for a virtual power plant based on an exact diffusion algorithm as described in any one of claims 1 to 4.

8. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of a decentralized AGC method for a virtual power plant based on an exact diffusion algorithm as described in any one of claims 1 to 4 are implemented.

Images