Optical storage and charging integrated power station aggregation flexibility prediction method based on mechanism fusion data

Abstract

The invention relates to the technical field of power grid dispatching, in particular to an optical storage and charging integrated power station aggregation flexibility prediction method based on mechanism fusion data. The method specifically comprises the following steps: constructing a unified linearization model of optical storage and charging equipment operation constraint and power distribution network power flow, constructing a multi-phase power distribution network power flow equation in combination with a Kirchhoff's law, and performing linearization by adopting a fixed point method to form a unified aggregation model; on the basis of a closed-loop learning algorithm, a network physical information and data driving cooperation mechanism is utilized, and through dynamic screening of high-uncertainty samples, convex set property derivation and historical model transfer learning, multi-round iterative training is performed on a classifier to expand the coverage range of a flexible space on the premise of ensuring the space convexity of a feasible solution. According to the method, the power grid topological information is embedded into the classifier training process through matrix operation, efficient fusion of mechanism constraint and machine learning is achieved, and the problem that conservative property is too high due to traditional random sampling or fixed geometric hypothesis is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid dispatching, and particularly relates to a method for predicting the aggregated flexibility of a photovoltaic energy storage charging integrated power station that integrates mechanism data. Background Art

[0002] In the new power system, the access of a large number of distributed energy sources on the distribution network side makes the distribution network an energy resource that can be flexibly dispatched by the large power grid; widely coordinating the distributed energy resources of the distribution network and uniformly dispatching them with controllable loads to obtain a certain power margin is considered an effective way to release the power flexibility of the distribution network and achieve positive interaction between the transmission and distribution systems; among them, the photovoltaic energy storage charging integrated power station, as a new type of power facility that integrates photovoltaic power generation, energy storage system, and charging pile, its aggregated flexibility is of great significance for the stable operation and optimal dispatching of the power grid.

[0003] Commonly used in existing technical solutions is to first establish operation models for the photovoltaic energy storage charging devices under the substation respectively, and then construct a high-dimensional inscribed polyhedron of the flexibility solution space for the photovoltaic energy storage charging device cluster, so as to calculate the group-network collaborative optimization strategy; or directly model the aggregated flexibility at the substation level, make a preliminary assumption based on experience for the solution space of the aggregated flexibility, and then give the prediction of the aggregated flexibility of the power station by solving the probability optimization problem.

[0004] The existing technology has the following defects and problems in the prediction of the aggregated flexibility of the photovoltaic energy storage charging integrated power station: (1) Limitations of the physical model: There is a lack of unified processing of the models of photovoltaic, energy storage, and charging piles, resulting in a highly complex calculation for the flexibility space of distributed energy resources established; during the aggregation process, the non-linearity of the power flow cannot be processed, so the power flow relationship and network constraints are not considered; in addition, the solution model for the aggregated flexibility at the substation level does not consider the feasibility of decomposition, that is, the power dispatch trajectory in the aggregated flexibility space cannot be guaranteed to be achieved by appropriately dispatching the photovoltaic energy storage charging devices. (2) Insufficiency of the calculation method: The existing technology lacks the utilization of network physical information, resulting in insufficient speed, accuracy, and reliability of the prediction results; at the same time, the random sampling method requires a large amount of sample data, increasing the calculation burden and unable to meet the requirements of online applications. (3) Inaccurate characterization of the flexibility space: The aggregated flexibility space of the power station is the projection of the high-dimensional flexibility space of the photovoltaic energy storage distributed energy resources; the existing methods usually assume the solution space to be a fixed shape (such as a hypercube or an ellipsoid) according to experience and approximate it through an optimization problem. This method restricts the shape of the solution space, resulting in overly conservative prediction results, and some feasible solutions in the actual aggregated flexibility space of the power station are not included in the solution scheme, and the accuracy and reliability of the solution are relatively low. Summary of the Invention

[0005] The object of the present invention is to propose a method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that integrates mechanism and data in view of the problems existing in the background technology.

[0006] The technical solution of the present invention: A method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that integrates mechanism and data includes the following specific implementation steps: S1. Establish an aggregated model of the photovoltaic-storage-charging integrated power station: Based on the operation constraint parameters of the photovoltaic power generation system, energy storage system, and charging pile, combined with the multi-phase power grid power flow equation, a three-phase power grid power flow model is constructed through Kirchhoff's law, and the fixed-point method is used to linearize the power flow equation to obtain a unified aggregated model including the voltage state vector, power vector, and system connection matrix; S2. Solve the initial aggregated flexibility space: Solve the feasible region of the photovoltaic-storage power through a two-stage optimization algorithm, and generate an initial sample set and mark the feasible scenarios with the goal of maximizing the volume of the flexibility space; S3. Screen high-value training samples, calculate the posterior probability of unlabeled samples, and dynamically select samples near the classification boundary based on the uncertainty measurement function to add to the training set; S4. Mechanism and data collaborative labeling, using network topology information and convex set properties, automatically label all samples within the convex hull of the initial sample set to expand the coverage range of the flexibility space; S5. Transfer learning to train the classifier, combine historical model parameters and newly labeled samples, and optimize the classifier through multiple rounds of iteration; S6. Judge the training rounds, and judge whether the set training rounds are reached according to the preset termination conditions; S7. Output the prediction result. If the preset rounds are reached, output the boundary of the aggregated flexibility space of the photovoltaic-storage-charging integrated power station for the real-time dispatching decision-making of the power grid.

[0007] Preferably, the operation constraint parameters include the upper and lower limits of photovoltaic power output, the charge and discharge power constraints of the energy storage and the continuity constraint of the state of charge, and the charging pile power constraint.

[0008] Preferably, the multi-phase power grid power flow equation is represented in matrix form as: ; ; ; where, diag() is matrix diagonalization; H is a 3N×3N diagonal matrix; is the conjugate vector of the Δ-connected current; s Y is the Y-connected node injection complex power vector; v is the node voltage vector; s Δis the complex power vector in Δ connection; Hv is the inter-phase voltage conversion; i is the total current vector; Y L0 is the zero-sequence admittance matrix; v0 is the zero-sequence voltage vector; Y LL is the positive- and negative-sequence admittance matrix; is the inter-phase difference operator; H T represents the transpose operation performed on matrix H.

[0009] Preferably, in the first stage, the maximum flexibility space is estimated through robust optimization; In the second stage, infeasible scenarios are eliminated to narrow the solution set, and an initial feasible sample set is generated and labeled as "1".

[0010] Preferably, the uncertainty metric function is defined as: ; where is the sample uncertainty metric value, quantifying the uncertainty of the classifier's prediction result for the i-th sample; represents the posterior probability that the sample belongs to the positive class, i.e., the label is 1, given ; is the predicted output of the i-th sample.

[0011] Preferably, the process of linearizing the power flow equation using the fixed-point method is as follows: ; where v t is the voltage state vector; A is the system connection matrix; p t is the power vector of the photovoltaic energy storage charging station; a t is the voltage offset constant term; i t is the branch current vector, representing the current distribution of each branch of the power grid at time t; B is the current coupling matrix; b t is the current correction constant term; p 0,t is the aggregated power scalar; d T is the weight row vector; g t is the power compensation term.

[0012] Preferably, in step S5: The transfer learning technique is adopted to use the historical model parameters as the initial weights, and incremental training is carried out in combination with the newly labeled samples. After each round of training, the classifier parameters are updated to improve the prediction accuracy.

[0013] Preferably, in step S6: The preset number of training rounds is 100 rounds, and the training is terminated when the classifier F1-score reaches 97%.

[0014] Preferably, a method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that integrates mechanism and data is applicable to a photovoltaic-storage-charging integrated power station with a single-phase or three-phase distribution network, and the prediction results can be decomposed to lower-level distributed energy devices.

[0015] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects: The present invention designs a method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that integrates mechanism and data, and has the following beneficial technical effects: (1) Establishing a unified model for photovoltaic-storage-charging: The present invention considers different constraints of equipment operation and the power flow relationship of the distribution network to form a unified linear power flow model, thereby ensuring the feasibility of decomposing the results within the aggregated flexibility space and meeting the actual requirements in power grid dispatching. The unified model of photovoltaic-storage-charging equipment and power flow established by the present invention has expandability and broad universality, facilitating the use of power system measurement data and more ensuring the actual requirement that the aggregated flexibility results of the power station can be decomposed to lower-level distributed resources; (2) Integrating mechanism and data-driven: The present invention combines power grid physical information and data-driven technology, uses an extensible matrix operation to train a classifier, and optimizes it using network information to classify sample points, thereby quickly expanding the flexibility space, giving play to the high efficiency of the data-driven method, and improving the efficiency and reliability of calculation; (4) Closed-loop learning algorithm: The present invention continuously searches for the maximum inner approximation of the aggregated flexibility space by screening and optimizing in the sample pool, using network information and knowledge accumulation about the sample space, reducing the conservativeness of the approximation of the solution space during the solution process; (5) Improving prediction efficiency and reliability: The classifier trained by the present invention can quickly classify sample points and finally give the aggregated flexibility space, with high calculation efficiency and can be applied to the dispatching and control of photovoltaic-storage-charging integrated power stations in real time; (6) Enhancing the completeness and accuracy of the predicted space: The present invention combines cyber-physical information and data-driven technology, and cooperates with a closed-loop learning algorithm. The present invention can more accurately depict the flexibility space of the system, reduce the conservativeness of the prediction results, and compared with the prior art, can predict feasible solutions outside the conservative solution space obtained by the prior art, discover more scenarios where the power system can actually operate stably, and has better robustness and adaptability. Description of the Drawings

[0016] Figure 1 It is the method flow chart of a method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that integrates mechanism and data proposed by the present invention; Figure 2 It is the comparison situation between the results of the embodiments of the present invention and the random sampling technique; Figure 3This is the relationship curve between F1-score and the number of test rounds during the implementation of the present invention. Specific implementation manner

[0017] A method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that fuses data by mechanisms proposed by the present invention, as Figure 1 shown, includes the following specific implementation steps: S1. Establish an aggregation model for the photovoltaic-storage-charging integrated power station, specifically: Comprehensively considering the operating characteristics of photovoltaic power generation, energy storage systems, and charging piles, as well as the interactions and influences among them. For photovoltaic power generation, considering the operating characteristics in multiple time periods, the following constraints are established: ; Among them, and respectively represent the upper and lower limits of photovoltaic output; i represents the node number; t represents the time period number; represents the actual photovoltaic output power of node i, time period t, and phase sequence Ψ; Ψ represents the phase sequence number, which is used to describe the operating states of different phases in a single-phase / three-phase system. Since there may be single-phase operation in the distribution network, in order to ensure the universality of the model, Ψ is introduced to represent the phase sequence number; It should be noted that photovoltaic-storage-charging refers to the photovoltaic power generation system, energy storage system, and charging pile equipment under the substation. Aggregation means mapping the actual flexibility space of power generation and power consumption of these devices to the substation under the premise of meeting the equipment operation constraints and network topology and power flow constraints, so as to predict the power flexibility space at the substation level; Flexibility refers to the maximum space of the substation power in all scenarios that can be realized in the dispatching under the premise of ensuring the stable operation of the actual system; The energy storage device also considers the upper and lower limit constraints: ; Among them, represents the lower limit of the energy storage power; represents the charging and discharging power of the energy storage device; represents the upper limit of the energy storage power; In addition, according to the characteristics of the energy storage system, the following SOC (State of Charge) constraints are considered: ; Among them, represents the state of charge of the energy storage at node i and time period t; κ i represents the self-discharge coefficient; represents the SOC value of the previous time period, reflecting the time continuity of the energy storage state; represents the initial situation of the state of charge; represents the final state of charge; Δt is the time period interval; is the lower limit of SOC; is the upper limit of SOC; The constraints of the charging pile equipment are similar to those of the energy storage equipment; For the power flow relationship, the following processing is carried out: Without loss of generality, considering a polyphase distribution network, the power flow of the polyphase distribution network is constructed according to Kirchhoff's law: ; where j is the node number of the distribution network, that is, the specific position in the power grid topology (including but not limited to transformers, buses, load connection points); a, b, and c are the phase identifiers in the three-phase system, corresponding to phase A, phase B, and phase C respectively; is the complex power between phases ab at node j; is the complex power between phases bc at node j; is the complex power between phases ca at node j; is the injected complex power of phase a at node j; is the injected complex power of phase b at node j; is the injected complex power of phase c at node j; is the voltage phasor of phase a at node j; is the voltage phasor of phase b at node j; is the voltage phasor of phase c at node j; is the conjugate phasor of the current between phases ab at node j; is the conjugate phasor of the current between phases bc at node j; is the conjugate phasor of the current between phases ca at node j; is the conjugate phasor of the injected current of phase a at node j; is the conjugate phasor of the injected current of phase b at node j; is the conjugate phasor of the injected current of phase c at node j; Furthermore, it can be written in matrix form as: ; ; ; where diag() is to diagonalize the matrix; H is a 3N×3N diagonal matrix; is the conjugate vector of the Δ-connected current; s Y is the vector of the injected complex power of the Y-connected nodes; v is the node voltage vector; s Δ is the vector of the Δ-connected complex power, that is, the power of the three-phase equipment in the Δ connection mode; Hv is the conversion of the interphase voltage, which converts the node voltage vector into the interphase voltage through the H matrix; i is the total current vector, that is, the combined current of all nodes or branches in the power grid; Y L0is the zero-sequence admittance matrix, which describes the relationship between the zero-sequence current components (such as ground faults) and voltages in a three-phase system; v0 is the zero-sequence voltage vector, that is, the zero-sequence components of the three-phase voltages; Y LL is the positive-sequence / negative-sequence admittance matrix, which describes the admittance characteristics of a three-phase system under symmetric operating conditions; is the inter-phase difference operator, which is used to convert three-phase voltages / currents from phasors to inter-phase quantities; H T represents the transpose operation performed on matrix H; Applying the fixed-point method to linearize the general power flow model, the following power flow model can be obtained: ; where, v t is the voltage state vector, which contains the voltage magnitudes or phase angles of each node at time t; A is the system connection matrix, and the element values are determined by the power grid topology (including but not limited to line impedances, transformer turns ratios); p t is the power vector of the photovoltaic-battery-charging pile system, including but not limited to the output powers of photovoltaic, energy storage, and charging piles at time t; a t is the voltage offset constant term, including but not limited to the fixed impacts of uncontrollable loads and line losses; i t is the branch current vector, which represents the current distribution of each branch of the power grid at time t; B is the current coupling matrix, which is constructed from including but not limited to line admittances and node connection relationship parameters; b t is the current correction constant term, including but not limited to the current components of uncontrollable loads and line loss compensation; p 0,t is the aggregated power scalar, that is, the total equivalent power interaction of the photovoltaic-battery-charging pile system with the power grid at time t; d T is the weight row vector, and the elements are the aggregation coefficients of the powers of each device; g t is the power compensation term, including but not limited to corrections such as line losses and power differences of uncontrollable loads; Accordingly: p t captures the output powers of the photovoltaic-battery-charging pile system. The coefficient parameters reflect the system connection mode, and the constant term parameters reflect the power flow losses on the line and brought by uncontrollable loads; Combining the voltage and current limit constraints, the aforementioned model can be summarized into this unified aggregated model: Wp ≤ z, p0 = Dp + b; where, W is the constraint coefficient matrix, which describes the linear relationship between the powers of the photovoltaic-battery-charging pile system and the system operation limits; z is the constraint limit vector, which contains the right-side thresholds of each constraint condition; p is the power vector of the photovoltaic-battery-charging pile system; p0 is the system balance power vector, that is, the power demand of the key nodes or regions of the power grid; D is the power distribution matrix, which describes the contribution relationship of the powers of the photovoltaic-battery-charging pile system to the system balance power; b is the balance compensation term, including but not limited to the power components of uncontrollable power sources or fixed loads; Accordingly, the model expresses the conditions for the stable operation of the system. Through this model, the response power of the energy resources of the photovoltaic-battery-energy storage charging system can be quickly normalized, and the power aggregation of the integrated photovoltaic-battery-energy storage charging power station can be achieved.

[0018] S2. Solve the aggregation flexibility optimization problem, initialize the classification space, and randomly select samples, specifically: Express the aggregation flexibility space as set P0. For all p0 ∈ P0, a p can be found to satisfy the unified aggregation model. Therefore, the optimization problem can be expressed as: ; In the formula, P0 is the set of feasible regions; volume() is the volume function. The goal is to expand its volume through optimization so that the system has higher robustness or flexibility when the parameters change. Maximizing the volume means expanding the coverage of the feasible solutions; p(p0) is the dependent variable, that is, the decision variable or intermediate variable associated with p0; D is the linear transformation matrix, which describes the mapping relationship from the dependent variable p to the main variable p0; b is the offset vector, which is used to adjust the result after linear transformation; represents the constraint condition; Considering the aggregation optimality and decomposition feasibility, this optimization problem can be solved to obtain the solution space through the principle of adaptive robust optimization. The optimization problem is solved in two stages. In the first stage, the flexibility space is maximally estimated. In the second stage, the non-decomposable power scenarios estimated in the first stage are excluded, and the maximum estimate is reduced, so as to ensure the adaptability and practicality of the estimation result to the actual scheduling; Continuously optimize in the two-stage calculation to obtain the data-driven initial flexibility solution space; The problem formula is described as follows: ; Among them, and are the upper and lower limit vectors of the power state; 1 is the all-1 vector, with the same dimension as p0, which is used to calculate the interval width; 0 is the all-0 vector, with the same dimension as p, and the inner-layer optimization objective function is the constant 0; Accordingly, it is stipulated that the scenarios within this initial space are feasible samples, marked as "1", so as to obtain the initial classifier; the samples outside the remaining space are not marked temporarily, providing a basis for subsequent sample selection and classification; randomly select from the unmarked sample pool to participate in subsequent training.

[0019] S3. Calculate the posterior probability and filter the samples, specifically: In each round, seek to find the samples that are the most uncertain about the classifier, that is, the samples that contain the most new knowledge about the sample space; Calculate the posterior probability of the classification space result. The closer the posterior probability is to 1, the more feasible the sample scenario is, and the closer the probability is to 0, the more infeasible it is; Consider the following monotonic function to measure the posterior probability, i.e., the uncertainty measurement function: ; where is the sample uncertainty measurement value, which is used to quantify the uncertainty of the classifier's prediction result for the i-th sample; represents the posterior probability that the sample belongs to the positive class (labeled 1) given ; is the predicted output of the i-th sample; Accordingly: M expresses the uncertainty of the posterior probability P. Select the sample with the largest M, that is, the sample whose posterior probability is closer to 0.5 participates in the subsequent training process; in this way, the samples with the most training value for the classifier are filtered out; as for the initial number of samples in each round and the number of samples selected by filtering, they are customizable and are hyperparameters of the model; this step calculates the probabilities of the samples belonging to different classes, identifies and excludes those samples with low information content to improve the sample quality, thereby improving the training efficiency of the algorithm.

[0020] S4. Mechanism fusion and classification marking of network information, specifically: Perform the mechanism fusion of network information, and use the network knowledge formed by the integrated photovoltaic energy storage and charging power station model established in the previous step S1 to prune the sample space, so that the points from a certain area can be marked without calculation; Mark all the samples in the initial aggregated flexibility space set P0 obtained from the optimization problem in step S2 as 1; Mark the samples filtered by step S3 under the classifier obtained after step S2. Those belonging to the flexibility space are marked as "1", otherwise marked as "0"; Then expand the flexibility space. Since is a convex constraint, and P0 is the projection of the solution space of on p0, therefore, P0 is a convex set; Thus, the convex hull of any member of P0 must be a subset of P0; Accordingly: All the samples contained in the convex hull of the sample set marked as "1" by steps S2 and S3 should be directly marked as 1, so as to aggregate the flexibility space, that is, the solution set of the problem can be rapidly expanded in each round of training, which benefits from the use of network information.

[0021] S5. Train the classifier by combining the historical model and the newly marked samples, specifically: Use the enlarged training set generated in step S4, which includes the newly selected samples and the original samples, to train the model. Utilize the parameters from the historical model and use transfer learning to accelerate the training. Then, continuously grow by merging the selected high-value training data points identified in each round by step S3 to cover a wider range of operating scenarios and conditions.

[0022] S6. Determine whether the set number of rounds is reached, specifically: Preset the number of training rounds, such as Figure 2 and Figure 3 As shown, in this embodiment, the performance effect of the proposed method is verified using the actual operating values of the Southern California Edison network. According to the test results, when the set number of rounds is around 100, the flexible domain boundary can be determined with an accuracy of not less than 97%, and compared with the general rectangular and ellipsoidal assumptions, the conservativeness of this boundary can be greatly reduced.

[0023] S7. End after reaching the number of rounds to obtain the maximum prediction, specifically: End after reaching a certain number of training rounds to obtain the maximum prediction of the aggregated flexibility space; after sufficient training, the classifier can accurately predict the aggregated flexibility space of the integrated energy storage and charging station under different operating conditions, providing decision-making support for power grid dispatching and optimization.

[0024] Accordingly: Through steps S1 to S7, the method of the present invention can effectively predict the aggregated flexibility of the integrated energy storage and charging station, providing a scientific basis for the real-time dispatching and operation optimization of the actual power grid, and improving the operation efficiency and reliability of the power grid.

[0025] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the knowledge scope of those skilled in the art.

Claims

1. A method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that fuses data based on mechanisms, characterized in that, It includes the following specific implementation steps: S1. Establish an aggregation model for the integrated photovoltaic energy storage and charging station: Based on the operation constraint parameters of the photovoltaic power generation system, energy storage system, and charging pile, combined with the multi-phase power grid power flow equation, a three-phase power grid power flow model is constructed through Kirchhoff's law, and the fixed-point method is used to linearize the power flow equation to obtain a unified aggregation model including the voltage state vector, power vector, and system connection matrix; S2. Solve the initial aggregation flexibility space: Solve the feasible region of the photovoltaic energy storage and charging power through a two-stage optimization algorithm, and generate an initial sample set and mark the feasible scenarios with the goal of maximizing the volume of the flexibility space; S3. Screen high-value training samples, calculate the posterior probability of unlabeled samples, and dynamically select samples near the classification boundary based on the uncertainty measurement function to add to the training set; S4. Mechanism and data collaborative labeling, using network topology information and convex set properties, automatically label all samples within the convex hull of the initial sample set to expand the coverage of the flexibility space; S5. Transfer learning to train the classifier, combine historical model parameters with newly labeled samples, and optimize the classifier through multiple rounds of iteration; S6. Judge the training rounds, and judge whether the set training rounds are reached according to the preset termination conditions; S7. Output the prediction result. If the preset rounds are reached, output the aggregation flexibility space boundary of the integrated photovoltaic energy storage and charging station for the real-time dispatching decision-making of the power grid.

2. The method for predicting the aggregation flexibility of a photovoltaic-storage-charging integrated power station that fuses data by mechanism according to claim 1, wherein The operation constraint parameters include the upper and lower limits of photovoltaic output, the charge and discharge power constraints of the energy storage and the continuity constraint of the state of charge, and the power constraint of the charging pile.

3. A method for predicting the aggregated flexibility of a photovoltaic-storage-charging integrated power station that fuses data based on mechanisms according to claim 1, characterized in that The multi-phase power grid power flow equation is expressed in matrix form as: ； ； ； where diag() is matrix diagonalization; H is a 3N×3N diagonal matrix; is the conjugate vector of the Δ-connected current; s Y is the complex power injection vector of the Y-connected nodes; v is the node voltage vector; s Δ is the complex power vector of the Δ connection; Hv is the phase voltage conversion; i is the total current vector; Y L0 is the zero-sequence admittance matrix; v0 is the zero-sequence voltage vector; Y LL is the positive-sequence and negative-sequence admittance matrix; is the phase difference operator; H T denotes the transpose operation on matrix H.

4. A method for predicting the aggregation flexibility of an integrated photovoltaic energy storage and charging power station that fuses data based on mechanisms according to claim 1, wherein In the first stage, the maximum flexibility space is estimated through robust optimization; In the second stage, infeasible scenarios are eliminated to narrow the solution set, and an initial feasible sample set is generated and marked as "1".

5. A method for predicting the aggregation flexibility of an integrated photovoltaic energy storage charging power station that fuses data based on mechanisms according to claim 1, characterized in that, The uncertainty measurement function is defined as: ； Among them, is the sample uncertainty measurement value, quantifying the uncertainty of the classifier's prediction result for the i-th sample; represents that when given the posterior probability that the sample belongs to the positive class, that is, the label is 1; is the predicted output of the i-th sample.

6. The method for predicting the aggregation flexibility of a photovoltaic energy storage and charging integrated power station that fuses data based on mechanisms according to claim 1, wherein The process of linearizing the power flow equation using the fixed-point method is: ； Among them, v t is the voltage state vector; A is the system connection matrix; p t is the power vector of the photovoltaic energy storage charging system; a t is the voltage offset constant term; i t is the branch current vector, representing the current distribution of each branch of the power grid at time t; B is the current coupling matrix; b t is the current correction constant term; p 0,t is the aggregated power scalar; d T is the weight row vector; g t is the power compensation term.

7. A method for predicting the aggregation flexibility of an integrated photovoltaic energy storage and charging power station that fuses data based on a mechanism, as claimed in claim 1, wherein In step S5: The transfer learning technology is adopted to use the historical model parameters as the initial weights, and incremental training is carried out in combination with the newly labeled samples. The classifier parameters are updated after each round of training to improve the prediction accuracy.

8. A method for predicting the aggregated flexibility of a photovoltaic energy storage charging integrated power station that fuses data based on a mechanism, as described in claim 1, wherein In the said step S6: The preset training rounds are 100 rounds, and the training is terminated when the F1-score of the classifier reaches 97%.

9. A method for predicting the aggregation flexibility of an integrated photovoltaic energy storage and charging power station that fuses data based on mechanisms according to any one of claims 1 to 8, characterized in that, A method for predicting the aggregation flexibility of an integrated photovoltaic energy storage and charging station by fusing mechanism and data is applicable to integrated photovoltaic energy storage and charging stations with single-phase or three-phase distribution networks, and the prediction results can be decomposed to lower-level distributed energy devices.

Images