Method and system for regulating participation of virtual power plant in day-ahead reserve auxiliary service based on similar entropy convex hull feasible region

Abstract

The application discloses a kind of virtual power plant based on similar entropy convex hull feasible region participates day in advance and is regulated and controlled method and system of reserve auxiliary service, the method is in 96 15 minute period, based on resource node physical characteristics, dynamic response and typical time sequence mode calculation multidimensional similarity entropy and weighted fusion, constructs similar neighborhood;Sample is filtered from similar neighborhood historical feasible record and is constructed convex hull, and intersection is formed with node physical feasible region to form similar entropy convex hull feasible region;Linear programming day in advance optimization model is established under the constraint of this feasible region, introduce lack of power supply and capacity penalty sequence that change with time period, and with electric quantity income settlement is carried out, output day in advance and is regulated and controlled plan with reserve declaration plan.The method can be solved using ADMM distributed and the convergence stability is improved by limiting the modification amplitude of the dual, improve the executability and economy of declaration plan.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant aggregation control and ancillary service trading technology in power systems, and particularly to a method for regulating virtual power plants participating in day-ahead standby ancillary services based on the feasible region of similar entropy convex hull. Background Technology

[0002] With the widespread integration of distributed power sources, energy storage, and adjustable loads on both the user and distribution sides, virtual power plants can aggregate various types of dispersed resources into a unified entity to participate in ancillary service transactions, such as day-ahead reserve increases. Day-ahead reserve increases typically require participating entities to submit available increase capacity within discrete time periods and meet requirements such as power response and duration during actual calls.

[0003] In existing technologies, one type of method mainly relies on mechanistic models for day-ahead optimization scheduling, such as establishing charging and discharging power and energy state constraints for energy storage, and establishing reduction boundaries and comfort constraints for adjustable loads. However, in engineering applications, aggregated resources come from diverse sources and their parameters are time-varying. Some constraints have implicit rules or "soft constraint" characteristics. Relying solely on static mechanistic models can easily lead to plans that are "mathematically feasible but not engineering feasible," thereby causing the risk of supply shortages.

[0004] Another type of method relies on historical experience or heuristic rules (such as applying for a discount based on the historical maximum available capacity). It is simple to implement but difficult to characterize cross-time coupling (such as energy storage SOC and load replenishment) and time-period difference risks. Especially during peak periods or continuous calls, it is difficult to balance the feasibility and economy of the plan.

[0005] Meanwhile, the recent increase in standby settlement generally includes assessment factors such as capacity revenue, electricity revenue, and penalties for supply shortages, and the risk cost of supply shortages may change at different times (e.g., peak / off-peak / valley). If the optimization model does not explicitly characterize the penalty weights for supply shortages and changes in time, it may lead to aggressive declarations and increase the risk of default. On the other hand, virtual power plants have a large number of nodes and complex constraints, requiring high solution efficiency in engineering and support for distributed solutions when necessary.

[0006] Therefore, the following technical problems have been raised and urgently require solutions: How to use historical feasible operation samples of similar nodes to form prior constraints of the feasible region while maintaining the solvability of linear programming, and combine them with physical constraints to improve the executability of day-ahead plans; How to introduce the power shortage and penalty weights that change with time in the settlement model, and settle revenue based on the actual delivered activated electricity, thereby reducing the risk of power shortage and improving overall economic efficiency. Summary of the Invention

[0007] The technical problem to be solved by this invention is: to address the issues of resource heterogeneity, time-varying constraints, strong cross-period coupling, difficulty in explicitly modeling historical feasible forms, and the changing costs of supply shortage risk when virtual power plants participate in day-ahead reserve adjustments, a regulation method for virtual power plants participating in day-ahead reserve ancillary services based on the feasible domain of similar entropy convex hull is provided.

[0008] To solve the above technical problems, the present invention adopts the following technical solution:

[0009] First, this invention proposes a method for regulating virtual power plants participating in day-ahead standby ancillary services based on the feasible region of the convex hull of similar entropy, comprising the following steps:

[0010] S1: Establish a day-ahead scheduling time period set t=1,2,…,96, with a time period length Δt=15 minutes (i.e., 0.25 hours); obtain the physical parameters, historical operation / execution data, day-ahead forecast data, and market settlement parameters for each node within the virtual power plant. Market settlement parameters must include at least: capacity price. (RMB / MW), Electricity Price (Yuan / MWh) and capacity penalty sequence varying over time (RMB / MW).

[0011] S2: For any pair of nodes (i,j), construct similarity indices in the dimensions of physical features, up-adjustment backup dynamic response, and typical time-series pattern, respectively; calculate the similarity entropy after normalizing the similarity to a probability distribution, and then normalize the similarity entropy of each dimension and fuse them according to the weight to obtain the weighted similarity entropy S(i,j).

[0012] S3: Select similar neighbors for each node i based on S(i,j). A hybrid strategy of "threshold filtering + Top-K supplementation / truncation" is preferred to ensure the stability of the neighborhood size.

[0013] S4: From similar neighborhoods Candidate samples are extracted from historical records, and a feasible sample set is obtained by filtering based on the feasibility criteria of node type. And construct the feasible region of the convex hull of similar entropy:

[0014] ,

[0015] In the formula, Let be the feasible region of the convex hull of the similarity entropy of node i; This is the set of feasible decision samples obtained from similar neighborhoods. Let i be the set of similar neighboring nodes of node i; Let i be the physical feasible region of node i (consisting of constraints such as power, energy, ramp rate, and comfort).

[0016] The convex hull constraint is constructed segment by segment in 15-minute intervals, and the local vector for each interval t is... Introducing convex combination weights :

[0017] ;

[0018] In the formula, Let be the local decision vector of node i in time period t; Let be the local vector of the m-th historical feasible sample of node i in time period t; The weights are convex combination weights (dimensionless), which are non-negative and sum to 1, and are used to realize the linear expression of "belonging to the convex hull". Let be the number of historical feasible samples available for node i in time period t.

[0019] Thus, the prior feasible region constraint of "belonging to the convex hull set" is realized by linear constraints.

[0020] S5: At each node Establish and solve a day-ahead linear programming optimization model under constraints. Model introduction: Application for increased reserve capacity. (MW), Activation Power (MW), power shortage (MW) and actual delivered activated power (MWh), satisfying , , , .

[0021] The objective function should at least include capacity revenue, electricity revenue, and capacity penalty terms:

[0022] ,

[0023] In the formula, The capacity price for time period t; The activation power price for time period t; The unit price of capacity penalty for time period t; This represents the linear cost or penalty term for node i during time period t, which can be used to represent losses, comfort losses, action penalties, etc. The solution outputs the virtual power plant's day-ahead reserve requirement plan and the control plans for each node's baseline power, state trajectory, etc.

[0024] Secondly, this invention proposes a control system for virtual power plants participating in day-ahead reserve ancillary services based on the feasible region of similar entropy convex hull, comprising:

[0025] The data acquisition module is used to establish a day-ahead scheduling time period set, and to obtain the standby-related physical parameters of each resource node in the virtual power plant, historical standby execution data, and day-ahead forecast data.

[0026] The similarity entropy calculation module is used to construct the similarity between any two nodes in the dimensions of physical features, up-adjusted backup dynamic response, and typical time series pattern, respectively. After normalizing the similarity to form a probability distribution, the similarity entropy is calculated. The similarity entropy of each dimension is normalized and fused according to the weight to obtain the weighted similarity entropy.

[0027] The similar neighborhood selection module is used to select similar neighborhoods for each node based on the weighted similarity entropy.

[0028] The convex hull feasible region construction module is used to select feasible decision samples that meet the constraints from the historical date-upgraded backup records of similar neighborhoods, construct the prior convex hull set, and intersect it with the physical feasible region of node i to obtain the similarity entropy convex hull feasible region of the node.

[0029] The day-ahead optimization solution and reporting output module is used to establish and solve the virtual power plant day-ahead reserve adjustment optimization model under the constraint of the feasible region of the similar entropy convex hull of each node. It outputs the reserve adjustment reporting amount, baseline power plan, energy storage energy status trajectory, and adjustable load reduction reserve of each node during the scheduling period, and summarizes them to form the virtual power plant day-ahead reserve adjustment reporting plan.

[0030] Compared with the prior art, the present invention, by adopting the above technical solution, has at least the following technical effects:

[0031] (1) Select similar neighborhoods by similar entropy and construct a convex hull prior feasible region using the historical feasible samples of the most recent 30 days, and intersect with the physical feasible region to improve the executability of the current day plan;

[0032] (2) Introduce the power shortage and capacity penalty sequence that varies with time period in the settlement model, and settle the settlement based on the actual delivered activated power, so that the risk cost is reflected differently in peak / flat / valley time periods, and the risk of default due to power shortage is reduced.

[0033] (3) It maintains the linear programming structure, has high solution efficiency, and can be extended to distributed solution to adapt to large-scale aggregation scenarios. Attached Figure Description

[0034] Figure 1 This is a flowchart of the method.

[0035] Figure 2 This diagram illustrates the calculation of similarity entropy and the selection of similar neighborhoods.

[0036] Figure 3 A schematic diagram for finding the intersection of the convex hull feasible region and the physical feasible region.

[0037] Figure 4 This is a schematic diagram of the ADMM distributed solution framework. Detailed Implementation

[0038] 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.

[0039] First refer to Figure 1 As shown, this invention proposes a method for regulating a virtual power plant participating in day-ahead reserve ancillary services based on the feasible region of the convex hull of similar entropy, including the following implementation steps:

[0040] 1. Set time period, symbols, and units.

[0041] The current scheduling time period set is t=1,2,…,96, with a time interval Δt=15 minutes, or Δt=0.25 hours. The node set is i=1,2,…,N (preferably N=50).

[0042] Price and penalty parameters include: capacity price (RMB / MW), Electricity Price (Yuan / MWh) and capacity penalty sequence varying over time (yuan / MW), of which The penalty sequence can be set in segments according to peak / flat / valley or configured for each time period.

[0043] 2. Similarity Entropy Calculation and Similar Neighborhood Selection

[0044] refer to Figure 2 As shown, for any pair of nodes (i,j), similarity is constructed in the dimensions of physical features, adjusted backup dynamic response, and typical time-series pattern, respectively; the similarity is normalized to form a probability distribution and then the similarity entropy is calculated; the similarity entropy of each dimension is normalized and weighted and fused according to the weight to obtain the weighted similarity entropy S(i,j).

[0045] For each node i, select similar neighbors based on S(i,j). A hybrid strategy of threshold screening and Top-K completion is preferred: threshold screening... Minimum number of neighborhoods K min =10, cut off Top-K and take K=12.

[0046] 3. Construction of feasible region of convex hull due to similarity entropy

[0047] refer to Figure 3 As shown, from similar neighborhood Extracting candidate samples from historical records Construct the convex hull and connect it to the physical feasible region of the nodes. Find the intersection to obtain the feasible region of the convex hull of the similarity entropy. :

[0048] ,

[0049] In the formula, Let be the feasible region of the convex hull of the similarity entropy of node i; This is the set of feasible decision samples obtained from similar neighborhoods. Let i be the set of similar neighboring nodes of node i; Let i be the physical feasible region of node i (consisting of constraints such as power, energy, ramp rate, and comfort).

[0050] To implement the constraint "belongs to the convex hull set" in linear programming, the local vector at each time interval t is... Introducing convex combination weights And apply linear convex combination constraints:

[0051] ,

[0052] In the formula, Let be the local decision vector of node i in time period t; Let be the local vector of the m-th historical feasible sample of node i in time period t; The weights are convex combination weights (dimensionless), which are non-negative and sum to 1, and are used to realize the linear expression of "belonging to the convex hull". Let be the number of historical feasible samples available for node i in time period t.

[0053] accomplish And together with cross-time physical constraints, constitute .

[0054] 4. The settlement model has been optimized recently.

[0055] Introducing activation power variables into recent optimization models With activated power variable :

[0056] ;

[0057] In the formula, In order to apply for an increase in standby capacity.

[0058] Introducing power supply variables Based on this, the actual delivered activation power is defined. :

[0059] ;

[0060] The objective function is preferably in linear form and includes at least a capacity revenue item, a power revenue item calculated based on the actual delivered activated power, and a capacity penalty item calculated based on the power shortage and varying over time.

[0061] ,

[0062] In the formula, The capacity price for time period t; The activation power price for time period t; The unit price of capacity penalty for time period t; This is the linear cost or penalty term for node i in time period t, which can be used to represent losses, comfort losses, action penalties, etc.

[0063] 5. General physical constraints

[0064] For any node i and time period t, a power upper limit coupling constraint can be applied:

[0065] ,

[0066] Climbing constraints:

[0067] ,

[0068] In the formula, Let be the baseline power of node i in time period t; The upper limit of the net power injected into the grid by node i during time period t; To increase the reserve declaration quantity; The upward climbing ability of point i.

[0069] 6. Implementation methods for energy storage nodes

[0070] Physical constraints and duration of energy storage nodes: Introducing charging power to energy storage node i∈B With discharge power And satisfy power boundary constraints:

[0071] ;

[0072] In the formula, and These are the upper limits of charging and discharging power for energy storage node i, respectively.

[0073] Define net grid-connected power (positive for grid injection):

[0074] ;

[0075] SOC Dynamics and Boundary Constraints:

[0076] ,

[0077] ,

[0078] In the formula, For charging efficiency; For discharge efficiency; Let i be the lower limit of the SOC of energy storage node i; This represents the upper limit of the SOC of energy storage node i.

[0079] To ensure the increased delivery capacity, a deliverable capacity constraint is imposed:

[0080] ,

[0081] In the formula, This represents the discharge space that can still be increased at the current discharge baseline.

[0082] To meet the minimum duration requirement of 2 hours (8 segments), an energy coverage constraint is applied based on the actual delivered activation power:

[0083] ,

[0084] In the formula The energy state of energy storage node i in time period t; A coefficient related to efficiency; This refers to the actual activated power delivered.

[0085] To improve feasibility and explicitly constrain charge / discharge patterns, the time-by-time convex hull local vector of the energy storage node is:

[0086] ;

[0087] And apply convex hull constraints .

[0088] When historical data does not contain the missing supply field but includes metering and baseline, the historical actual activation power can be estimated based on the positive deviation of the metering from the baseline, and truncated based on the historical declared reserve capacity:

[0089] ,

[0090] This leads to the historical actual activation power:

[0091] ,

[0092] In the formula, The estimated real-intersection activation power for sample m in time period t; The reserve capacity was increased for sample m during time period t. The measured net grid-connected power of sample m during time period t; The baseline net grid-connected power of sample m in time period t; This is the estimated actual activation power of sample m in time period t.

[0093] 7. Implementation method of adjustable load nodes

[0094] Introducing power reduction for adjustable load nodes i∈L Power replenishment With debt energy state The outstanding debt status is as follows:

[0095] ,

[0096] Boundary conditions and nonnegativity:

[0097] ,

[0098] Power and debt limit boundary:

[0099] ;

[0100] Given baseline load With the bottom line sequence changing over time Actual load:

[0101] ;

[0102] To ensure that the actual delivered activation power does not exceed the space that can be further reduced:

[0103] ;

[0104] To incorporate the "reduction-replenishment-debt" pattern into the historical prior, the local vector of the convex hull is defined for each time period:

[0105] ;

[0106] And apply convex hull constraints: .

[0107] Estimation of historical actual activated power: When historical data does not contain the supply shortage field but includes metering and baseline, the historical actual activated power reduction (equivalent to an upward adjustment of delivered power) can be estimated using the positive difference between the baseline and the metering, and truncated according to the historical declared reserve capacity:

[0108] This leads to the historical actual activation volume (equivalent to an upward adjustment of the delivered volume):

[0109] ;

[0110] Adjustable load (load is positive):

[0111] ;

[0112] In the formula, The estimated actual activation power of sample m in time period t is equivalent to the "actual reduced delivery power" for the load. The reserve capacity was increased for sample m during time period t. The measurement load of sample m in time period t; The baseline load of sample m in time period t; The estimated actual activated power for sample m in time period t is equivalent to the "actual reduced delivered power" for the load.

[0113] 8. Distributed Solution Model (ADMM) and Stability Enhancement

[0114] refer to Figure 4 As shown, to improve scalability, optimization variables can be decomposed into local variables by node and global consistency variables can be set. ADMM is used to alternately update local subproblems, global consistency variables and dual variables.

[0115] A global consensus-oriented correction term is introduced into the dual variable update, and the correction magnitude is limited or truncated to suppress oscillations and improve convergence stability. The stopping criterion can be based on the original residual and the dual residual being less than a threshold or reaching the maximum number of iterations.

[0116] This embodiment also proposes a virtual power plant day-ahead reserve optimization reporting and control system for implementing the above method, including: a data acquisition module, a similarity entropy calculation module, a similar neighborhood selection module, a convex hull feasible region construction module, a day-ahead optimization solution module, and a reporting output module; wherein the convex hull feasible region construction module is used to construct... The solution module has recently been optimized for use in Under constraints, the standby optimization model is adjusted and the application plan is output before the solution date. Specific implementation details have been described in the methodology section and will not be repeated here.

[0117] The foregoing has shown and described the basic principles, main features, and advantages of this disclosure. Those skilled in the art should understand that this disclosure is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this disclosure. Various changes and modifications can be made to this disclosure without departing from its spirit and scope, and all such changes and modifications fall within the scope of this disclosure as claimed.

Claims

1. A method for regulating a virtual power plant participating in day-ahead standby ancillary services based on the feasible region of the convex hull of similar entropy, characterized in that, Includes the following steps: S1. Establish a day-ahead scheduling time period set, and obtain the standby-related physical parameters, historical standby execution data and day-ahead forecast data of each resource node in the virtual power plant; S2. For any two nodes (i,j), construct the similarity between the nodes in the dimensions of physical features, up-adjustment backup dynamic response, and typical time series pattern, respectively. After normalizing the similarity to form a probability distribution, calculate the similarity entropy. Normalize the similarity entropy of each dimension and fuse them according to the weight to obtain the weighted similarity entropy S(i,j). S3. Based on the weighted similarity entropy, select similar neighborhoods for each node i. ; S4, from similar neighborhoods From the historical records of the previous day, feasible decision samples that meet the constraints are selected, a prior convex hull set is constructed, and its intersection with the physical feasible region of node i is calculated to obtain the similarity entropy convex hull feasible region of node i. ; S5. Feasible region of the convex hull of the similarity entropy at each node. Under constraints, a virtual power plant day-ahead reserve adjustment optimization model is established and solved. The output includes the reserve adjustment application amount, baseline power plan, energy storage energy status trajectory, and adjustable load reduction reserve amount of each node during the scheduling period. The results are then summarized to form the virtual power plant day-ahead reserve adjustment application plan.

2. The method according to claim 1, characterized in that: In step S3, a hybrid strategy of threshold filtering and Top-K complementation is used for similar neighborhoods. First, those that meet the criteria are selected. The candidate set; when the number of candidates is less than K min Sort by S(i,j) and fill in the gaps from the remaining nodes up to K. min When the number of candidate nodes exceeds K, the top-K nodes are selected as the final similar neighborhoods based on the sorting by S(i,j). The total number of nodes is N=50. K min =10, K=12.

3. The method according to claim 1, characterized in that: In step S4, the feasible region of the convex hull of the similarity entropy satisfies: ， In the formula, Let i be the feasible region of the convex hull of the similarity entropy. This is the set of feasible decision samples obtained from similar neighborhoods. Let i be the set of similar neighboring nodes of node i; Let i be the physical feasible region of node i, which consists of constraints composed of power, energy, climbing ability, and comfort.

4. The method according to claim 3, characterized in that: The constraints belonging to the convex hull set are expressed in convex combination form, that is, for each time period or each local variable block, there exists a non-negative weight λ such that the decision variable is equal to the weighted sum of the feasible samples, and the sum of the weights is 1.

5. The method according to claim 4, characterized in that: Construct convex hull constraints segment by segment in 15-minute intervals, and for each interval t, construct the local vector. Apply Together with the cross-time period coupling constraint, it forms the similarity entropy convex hull feasible region constraint of node i. Indicates from node similar neighborhood The set of historical feasible decision samples obtained through screening.

6. The method according to claim 1, characterized in that: For any node i and time period t, increase the reserve declaration quantity. Compared with baseline power At least one of the following must be met: a）、 ； b）、 ； In the formula, Let be the baseline power of node i in time period t; The upper limit of the net power injected into the grid by node i during time period t; To increase the reserve declaration quantity; The climbing ability of point i is adjusted upwards.

7. The method according to claim 6, characterized in that: In step S5, an activation power variable is introduced into the day-ahead standby optimization model. With activated power variable and satisfy as well as ,in Hours; further introduce power supply shortage variables and satisfy Define the actual delivered activation power. ; The objective function should at least include capacity revenue, electricity revenue, and capacity penalty terms: ， In the formula, The capacity price for time period t; The activation power price for time period t; The unit price of the capacity penalty for time period t; Let be the linear cost or penalty term for node i in time period t, used to represent loss, comfort loss, and action penalty.

8. The method according to claim 7, characterized in that: When the node is energy storage, the 8 time period coverage constraints corresponding to the minimum standby duration H=2 hours are expressed through energy coverage as follows: ， In the formula, The energy state of energy storage node i in time period t; Let i be the lower limit of the SOC of energy storage node i; A coefficient related to efficiency; This refers to the actual activated power delivered. When the node is an adjustable load, the increase in reserve is represented by the load reduction reserve, and meets the upper limit of the load reduction range, the upper limit of the duration, and / or comfort constraints.

9. The method according to claim 1, characterized in that: In step S5, the Alternating Direction Multiplier Method (ADMM) is used to solve the virtual power plant day-ahead reserve optimization model in a distributed manner. A correction term oriented towards global consensus is introduced in the dual variable update, and the correction magnitude is limited or truncated to suppress oscillations and improve convergence stability. The node contribution weight is calculated based on the weighted similarity entropy, and the local information of each node is weighted and aggregated according to the contribution weight to obtain a global knowledge representation. The global knowledge representation is used to guide the local optimization or distributed solution process.

10. A control system for virtual power plants participating in day-ahead standby ancillary services based on the feasible region of similar entropy convex hull, characterized in that, include: The data acquisition module is used to establish a day-ahead scheduling time period set, and to obtain the standby-related physical parameters of each resource node in the virtual power plant, historical standby execution data, and day-ahead forecast data. The similarity entropy calculation module is used to construct the similarity between any two nodes in the dimensions of physical features, up-adjusted backup dynamic response, and typical time series pattern, respectively. After normalizing the similarity to form a probability distribution, the similarity entropy is calculated. The similarity entropy of each dimension is normalized and fused according to the weight to obtain the weighted similarity entropy. The similar neighborhood selection module is used to select similar neighborhoods for each node based on the weighted similarity entropy. The convex hull feasible region construction module is used to select feasible decision samples that meet the constraints from the historical date-upgraded backup records of similar neighborhoods, construct the prior convex hull set, and intersect it with the physical feasible region of node i to obtain the similarity entropy convex hull feasible region of the node. The day-ahead optimization solution and reporting output module is used to establish and solve the virtual power plant day-ahead reserve adjustment optimization model under the constraint of the feasible region of the similar entropy convex hull of each node. It outputs the reserve adjustment reporting amount, baseline power plan, energy storage energy status trajectory, and adjustable load reduction reserve of each node during the scheduling period, and summarizes them to form the virtual power plant day-ahead reserve adjustment reporting plan.

Images