Progressive vertex enumeration adjustable capacity dynamic evaluation method

Abstract

The present application relates to the technical field of new power system load scheduling, and discloses a gradual vertex enumeration adjustable capacity dynamic evaluation method, comprising the following steps: firstly, uniformly constructing a high-dimensional process electrical constraint set of multiple load coupling of a submerged arc furnace, polysilicon and cement; generating an initial vertex by power space quartering sampling; gradually approaching the nonlinear non-convex adjustable power feasible region of industrial load by a double gradual iteration mechanism of infeasible vertex gradient offset and feasible vertex centroid filling; setting a vertex improvement rate convergence criterion; after iteration convergence, solving two sets of adjustable capacity of second-level frequency modulation and medium and long-term peak shaving by layering, and realizing 15-minute rolling dynamic update; the scheme solves the defects of large linear approximation error, poor real-time grid traversal and inability to compatible multiple heterogeneous industrial loads, can directly support high-proportion new energy park source-load collaborative control and power grid peak shaving and frequency modulation business, has strong engineering adaptability, and has significant economic and power grid operation benefits.

Description

Technical Field

[0001] This invention relates to the field of novel power system load dispatching technology, specifically to a dynamic evaluation method for adjustable capacity using asymptotic vertex enumeration. Background Technology

[0002] In new power systems with a high proportion of renewable energy integration, loads in high-energy-consuming industrial parks such as electric arc furnaces, polysilicon reduction furnaces, and cement plants possess large-capacity flexible regulation potential, enabling them to participate in grid peak shaving and primary frequency regulation, and to smooth out random power fluctuations from wind and solar power. The three types of loads exhibit completely heterogeneous constraint characteristics: Electric arc furnaces: The electric arc exhibits time-varying resistance and harmonic-added reactance characteristics; the electrode lifting system has a dead zone, a 0.2s action delay, and a difference in lifting rate constraints; the power feasible region is strongly nonlinear and non-convex. Polysilicon reduction furnaces: Strong thermal-engineering coupling exists between the silicon rod radius, reaction temperature, and cooling water flow; power has hard upper and lower thresholds for safe production. Cement production lines: The rotary kiln is an uninterrupted rigid process; multi-level storage of raw materials and clinker forms a time-divisionally discrete adjustable range.

[0003] Current methods for assessing the adjustable capacity of industrial loads have the following drawbacks: Linear approximation method: This method simplifies the nonlinear power boundary to a straight line, ignoring the concave and convex feasible regions caused by process constraints. The total capacity assessment error of the industrial park exceeds 30%, resulting in distorted assessment results and directly leading to illegal shutdowns due to dispatch orders. Uniform grid traversal method: This method involves high-density discrete sampling of the power space. In high-dimensional multi-load scenarios, the computational load explodes, with a single calculation taking more than 1 second, which cannot meet the real-time requirements of second-level frequency regulation of the power grid. Two-dimensional polygon enumeration method: This method only supports two-dimensional power space solutions for single loads and cannot handle high-dimensional feasible regions coupled with multiple loads such as electric arc furnaces, polysilicon, and cement. Single time scale solution: This method only outputs the intraday peak-shaving capacity and cannot distinguish between short-term adjustable intervals for second-level frequency regulation and medium- and long-term adjustable intervals for peak regulation, making it difficult to adapt to the dispatching needs of the power grid in multiple scenarios.

[0004] Therefore, there is an urgent need to propose an adjustable capacity dynamic assessment algorithm that simultaneously takes into account multiple heterogeneous load coupling, non-convex feasible region fitting, millisecond-level real-time calculation, multi-timescale hierarchical evaluation, adapts to mixed high-energy-consuming parks, and balances accuracy and speed. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a novel dynamic evaluation method for adjustable capacity using progressive vertex enumeration, which solves the problems of large linear approximation errors, poor real-time performance of mesh traversal, and incompatibility with multiple heterogeneous industrial loads in existing evaluation methods.

[0006] To achieve the above objectives, the present invention employs the following technical solution: A progressive vertex enumeration adjustable capacity dynamic evaluation method is used for high-precision dynamic evaluation of adjustable capacity across multiple time scales in submerged arc furnaces, polysilicon reduction furnaces, and cement production line mixed industrial parks. The method includes the following steps: Step S1: Collect real-time operating parameters of the submerged arc furnace, polysilicon reduction furnace, and cement production line in the industrial park, establish various load power constraint equations respectively, superimpose the common operating constraints of the power distribution network, and integrate them to form a high-dimensional inequality constraint set. ; Step S2: The active power domain of the three types of loads is sampled into four equal parts, and the Cartesian product is used to generate the original vertex pool. Extremely illegal vertices are coarsely screened out to obtain the initial iterative vertex set. ; Step S3: For the current iteration vertex set Batch substitution of all vertices into the constraint set Verification and partitioning into feasible vertex subsets and infeasible vertex subset And record the constraint violation gradient of infeasible vertices; Step S4: Perform adaptive reverse offset correction on infeasible vertices based on constraint gradients, and generate transitional candidate vertices which are then stored in the candidate pool. ; Step S5: Traverse feasible vertices pairwise and calculate their geometric centroids as new boundary vertices, storing them in the candidate pool. And repeat the feasibility check on the candidate pool vertices, retaining valid new vertices; Step S6: Merge the existing feasible vertices with the newly validated vertices, perform deduplication and redundant interior point filtering, and streamline to generate the vertex set for the next iteration. ; Step S7: Calculate the global improvement rate of the vertices in this round. If the convergence is not achieved, the process is repeated in step S3; if the convergence is achieved, the iteration is stopped and the adjustable feasible region of the park power is obtained by fitting all feasible vertices. Step S8: Apply frequency modulation and peak shaving scenario constraints respectively, and solve the adjustable capacity of frequency modulation at the second level and the adjustable capacity of peak shaving in the medium and long term in the day in a layered manner. Step S9: Read the real-time operating status of the load control terminal every 15 minutes and update the constraint set. Repeat steps S1 to S8 to achieve adjustable capacity rolling dynamic evaluation.

[0007] Furthermore, in step S1, the formula for calculating the active power of the electric arc furnace is: ; The fitting relationship between arc harmonic additional reactance is as follows: ; By superimposing electrode adjustment dead zone, smelting temperature, and transformer overload constraint, the power range of the electric arc furnace is obtained. .

[0008] Furthermore, in step S4, the formula for calculating the adaptive offset step size is as follows: , The corrected vertex expression is ,in Let be the constraint function that is most severely violated at the vertex. Let be the gradient vector of this constraint.

[0009] Furthermore, in step S5, the feasible formula for calculating the geometric centroid of the vertex is as follows: , Let be any two feasible vertices.

[0010] Furthermore, in step S7, the formula for calculating the global vertex improvement rate is: ; Convergence threshold ,when Determine if the iteration converges.

[0011] Furthermore, in step S8, a hard constraint is added to the second-level single frequency regulation scenario: the active power of the cement production line is fixed at the current baseline load. Only minor power adjustments are allowed for submerged arc furnaces and polysilicon plants; for medium- and long-term peak-shaving scenarios, the timing constraints on cement storage are lifted, and all three types of loads participate in power regulation.

[0012] Furthermore, in step S2, for any one-dimensional power interval , Interval width , The nodes for dividing into four equal parts are: .

[0013] Furthermore, in step S6, the Euclidean distance threshold for vertex deduplication is set to 0.2MW. Two points with a distance less than the threshold are identified as duplicate vertices and merged.

[0014] Furthermore, an LDC-320 industrial load control terminal is used to collect real-time operating data on silicon rod radius, cooling water temperature, raw and clinker inventory, and submerged arc furnace electrode current, which is used to dynamically update the constraint set. .

[0015] Furthermore, the total calculation time for a single complete iteration is less than 200ms, the assessment error of the total adjustable capacity of the park does not exceed 5%, and the assessment error of the single type of load capacity does not exceed 3%.

[0016] Compared to existing technologies, the beneficial effects of the proposed progressive vertex enumeration adjustable capacity dynamic evaluation method, using the above-mentioned scheme, employ a gradient offset + centroid filling dual iterative mechanism to accurately characterize the concave and convex segmented power boundaries of submerged arc furnaces, polysilicon, and cement loads. Compared to the linear approximation method, the evaluation error is reduced by 85%, with a total capacity error of ≤5% and a single load error of ≤3%. By adopting a four-part initial vertex sampling and vertex simplification strategy, the time taken for a single complete iteration of the 100 MW hybrid industrial park is less than 200ms, which meets the second-level data upload requirements of the power grid's primary frequency regulation. For the first time, electric arc electrical constraints, thermal coupling constraints, and storage timing constraints are incorporated into the same high-dimensional enumeration framework, which is compatible with electric arc furnaces, polysilicon, and cement mixed parks, and can be extended to process industries such as electrolytic aluminum and chlor-alkali. It can simultaneously output two adjustable capacities for short-term frequency regulation and medium-to-long-term peak regulation, adapting to various dispatching scenarios such as peak regulation, frequency regulation, and new energy fluctuation smoothing in new power systems. All capacity solutions strictly meet all production hard constraints, and the evaluation results will not cause equipment downtime or product scrapping, ensuring the safety of engineering implementation. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the progressive vertex enumeration adjustable capacity dynamic evaluation method of the present invention. Detailed Implementation

[0018] To facilitate understanding of the present invention by those skilled in the art, specific embodiments of the present invention are described below with reference to the accompanying drawings. Preferred embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the present invention.

[0019] It should be noted that when a component is said to be "fixed to" another component, it can be directly attached to the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component.

[0020] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0021] One embodiment of the present invention is as follows: Figure 1As shown, this asymptotic vertex enumeration adjustable capacity dynamic evaluation method is applicable to the fields of load-side dispatching in new power systems and industrial flexible load demand response. It can be used for high-precision dynamic evaluation of adjustable capacity across multiple time scales in submerged arc furnaces, polysilicon reduction furnaces, cement production lines, and mixed industrial parks. The method includes the following steps: Step S1: Construct a high-dimensional constraint set of multi-load coupling. Collect real-time operating parameters of the submerged arc furnace, polysilicon reduction furnace, and cement production line in the industrial park, establish single-load power constraint equations respectively, superimpose the common constraints of the distribution network operation, and integrate them to form a high-dimensional inequality constraint set. ; Specifically, constraints on electric arc furnaces: Arc active power: ; The fitting relationship between arc harmonic additional reactance is as follows: ; By superimposing electrode adjustment dead zone, smelting temperature, and transformer overload constraint, the power range of the electric arc furnace is obtained. .

[0022] Specifically, polysilicon reduction constraints: Based on a thermal equilibrium model, constraints are imposed on a reaction temperature of 1000~1080℃ and a minimum cooling water flow rate, resulting in... , Where is the radius of the silicon rod. This refers to the cooling water temperature.

[0023] Specifically, cement process constraints: Based on the STN state-task network, non-schedulable rotary kilns and scheduleable grinding equipment are distinguished, and time-sharing power intervals are generated based on the raw material and clinker storage capacities. .

[0024] Specifically, overall constraints within the park: the park has a total merit. By superimposing constraints on bus voltage, power factor, and distribution capacity, the constraint set is integrated as follows: ; Step S2: Quadruple sampling to generate the initial vertex set Determine the power domain of the three types of loads. ; Each power interval is divided into four equal parts. Split Node: , ; The Cartesian product of 3D segmented nodes generates the original vertex pool; The original vertex coarse constraint set Remove severely violating extreme vertices to obtain the initial iterative vertex set. .

[0025] Step S3: Batch partitioning of vertex feasibility, input the first... Wheel vertex set For each three-dimensional vertex Substitute each constraint into the constraint function: all A vertex is determined to be feasible and assigned to a subset. ; any Vertices deemed infeasible are relegated to a subset. ; Synchronously record the constraint violation gradients corresponding to each infeasible vertex. Used for vertex offset correction.

[0026] Step S4: Correction of infeasible vertex gradient reverse offset. For any non-compliant vertex, filter the constraint with the greatest degree of violation. Calculate the gradient ; Adaptive offset step size: ; Vertex inward correction formula: ; All corrected vertices are stored in a temporary candidate pool. .

[0027] Step S5: Fill and complete the non-convex boundary with the centroids of feasible vertices. Traversal All two-point combinations inside Calculate the geometric centroid: ; Add all centroid vertices to the candidate pool ; Repeat the feasibility check for all vertices in the candidate pool, and only retain new feasible vertices that meet the constraints.

[0028] Step S6: Simplify the vertices to generate the next iteration set, and merge the original feasible vertices. With the new vertex that passes the verification; Set Euclidean distance threshold Merge duplicate vertices whose distance is less than the threshold; Remove redundant interior points that are entirely within the feasible region and do not participate in the boundary formation; Output the simplified vertex set Iteration counting .

[0029] Step S7: Iterative convergence determination, defining the global vertex improvement rate: ; in, The coordinates of the vertex in the previous round. This represents the theoretical maximum power of the park. Set convergence threshold : like If the feasible region boundary converges, the iteration terminates; Return to step S3 to start a new round of iteration; After the iteration terminates, a high-dimensional convex hull fitting is performed on all feasible vertices to obtain the complete adjustable power feasible region of the park.

[0030] Step S8: Solve for the adjustable capacity of the park on different time scales, with second-level frequency modulation capacity: additional constraints (Cement power is fixed), solve for short-term adjustable capacity. Adjustable capacity ; Medium- and long-term peak-shaving capacity: Relax the time series constraints of cement storage, fully utilize the convergence feasible region, and solve for the intraday upscaling capacity. Adjustable capacity .

[0031] Step S9: Rolling dynamic update, collect real-time operating data from the LDC-320 industrial load control terminal every 15 minutes, and update the constraint set. Repeat the entire process from S1 to S8 to achieve dynamic online assessment of adjustable capacity.

[0032] In a specific embodiment, a hybrid industrial park is used as an example. The park includes 3 sets of electric arc furnaces, 4 sets of 24-pair polycrystalline silicon reduction furnaces, and 1 new dry process cement production line with a total installed capacity of 126MW. It is equipped with LDC-320 industrial load control terminals, RTDS real-time digital simulation platform, and park edge computing server.

[0033] Implementation steps: Constraint modeling (S1): The edge server reads real-time data from the terminal: electrode current of the submerged arc furnace, silicon rod radius, cooling water temperature, and raw / clinker inventory. Substituting these data into the corresponding constraint formulas, a three-dimensional high-dimensional constraint set is generated. ,variable Mineral heat always has benefits. Polycrystalline silicon has a total active power. Cement always has its uses.

[0034] Initial vertex generation (S2): The three-dimensional power interval is divided into four equal parts to generate 125 original vertices. After coarse screening, 78 initial vertices are retained. .

[0035] The iteration loop executes S3~S6, completing vertex partitioning, offset correction, centroid filling, and simplification in each round.

[0036] Convergence criterion: When iterating to the 12th round Once the convergence condition is met, the iteration stops.

[0037] Capacity calculation: Frequency regulation scenario: Cement power is fixed to the reference load, and the output can be adjusted up by 6.2MW and down by 5.7MW in a short time; Peak shaving scenario: With the relaxation of warehousing constraints, the total capacity of the park can be increased by 18.5MW and decreased by 16.9MW. Rolling update: The terminal operating status is refreshed every 15 minutes, the entire algorithm is re-executed, and adjustable capacity data is continuously sent to the Xiangyang power grid dispatching master station.

[0038] The implementation results show that the single calculation time is 168ms, which meets the real-time requirements of frequency regulation; compared with the actual load regulation test on site, the total capacity assessment error is 4.1%; it can directly support three types of services: primary frequency regulation of the power grid, local wind and solar power stabilization, and intraday optimized dispatch.

[0039] The asymptotic vertex enumeration adjustable capacity dynamic evaluation method in this embodiment proposes an initial vertex sampling mechanism that divides the power space into four equal parts, completely enclosing the high-dimensional non-convex feasible region with the fewest vertices, significantly reducing the number of iterations and balancing accuracy and computational speed. It designs an asymptotic iteration strategy of "infeasible vertex gradient inward offset + feasible vertex centroid filling" to address the industry pain points of concave and discontinuous feasible regions in industrial loads, making them difficult to fit. It constructs an integrated high-dimensional constraint solution framework for three types of heterogeneous loads: submerged arc furnaces, polysilicon, and cement, uniformly handling multiple types of production constraints including electrical, thermal, and temporal constraints. It proposes a hierarchical constraint solution method to distinguish between two sets of adjustable capacities: second-level frequency regulation and hour-level peak regulation, adapting to multi-timescale scheduling operations of the power grid. It designs a 15-minute rolling update mechanism to dynamically refresh the feasible region based on real-time terminal operating conditions, solving the problem that static evaluation cannot match dynamic changes in production.

[0040] It should be noted that the above-mentioned technical features can be combined with each other to form various embodiments not listed above, all of which are considered to be within the scope of this invention specification; and, for those skilled in the art, improvements or modifications can be made based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. An asymptotic vertex enumeration adjustable capacity dynamic evaluation method, used for high-precision dynamic evaluation of adjustable capacity across multiple time scales in submerged arc furnaces, polysilicon reduction furnaces, and cement production line mixed industrial parks, characterized by: Includes the following steps: Step S1: Collect real-time operating parameters of the submerged arc furnace, polysilicon reduction furnace, and cement production line in the industrial park, establish various load power constraint equations respectively, superimpose the common operating constraints of the power distribution network, and integrate them to form a high-dimensional inequality constraint set. ; Step S2: The active power domain of the three types of loads is sampled into four equal parts, and the Cartesian product is used to generate the original vertex pool. Extremely illegal vertices are coarsely screened out to obtain the initial iterative vertex set. ; Step S3: For the current iteration vertex set Batch substitution of all vertices into the constraint set Verification and partitioning into feasible vertex subsets and infeasible vertex subset And record the constraint violation gradient of infeasible vertices; Step S4: Perform adaptive reverse offset correction on infeasible vertices based on constraint gradients, and generate transitional candidate vertices which are then stored in the candidate pool. ; Step S5: Traverse feasible vertices pairwise and calculate their geometric centroids as new boundary vertices, storing them in the candidate pool. And repeat the feasibility check on the candidate pool vertices, retaining valid new vertices; Step S6: Merge the existing feasible vertices with the newly validated vertices, perform deduplication and redundant interior point filtering, and streamline to generate the vertex set for the next iteration. ; Step S7: Calculate the global improvement rate of the vertices in this round. If the convergence is not achieved, the process is repeated in step S3; if the convergence is achieved, the iteration is stopped and the adjustable feasible region of the park power is obtained by fitting all feasible vertices. Step S8: Apply frequency modulation and peak shaving scenario constraints respectively, and solve the adjustable capacity of frequency modulation at the second level and the adjustable capacity of peak shaving in the medium and long term in the day in a layered manner. Step S9: Read the real-time operating status of the load control terminal every 15 minutes and update the constraint set. Repeat steps S1 to S8 to achieve adjustable capacity rolling dynamic evaluation.

2. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S1, the formula for calculating the active power of the electric arc furnace is: ； The fitting relationship between arc harmonic additional reactance is as follows: ； By superimposing electrode adjustment dead zone, smelting temperature, and transformer overload constraint, the power range of the electric arc furnace is obtained. .

3. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S4, the formula for calculating the adaptive offset step size is as follows: , The corrected vertex expression is ,in Let be the constraint function that is most severely violated at the vertex. Let be the gradient vector of this constraint.

4. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S5, the feasible formula for calculating the geometric centroid of the vertex is as follows: , Let be any two feasible vertices.

5. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S7, the formula for calculating the global vertex improvement rate is: ； Convergence threshold ,when Determine if the iteration converges.

6. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S8, a hard constraint is added to the second-level single frequency regulation scenario: the active power of the cement production line is fixed at the current baseline load. Only minor power adjustments are allowed for submerged arc furnaces and polysilicon plants; for medium- and long-term peak-shaving scenarios, the timing constraints on cement storage are lifted, and all three types of loads participate in power regulation.

7. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S2, for any one-dimensional power interval , Interval width , The nodes for dividing into four equal parts are: 。 8. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, In step S6, the Euclidean distance threshold for vertex deduplication is set to 0.2MW. Two points with a distance less than the threshold are identified as duplicate vertices and merged.

9. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, The LDC-320 industrial load control terminal is used to collect real-time operating data on silicon rod radius, cooling water temperature, raw and clinker inventory, and electric arc furnace electrode current, which are then used to dynamically update the constraint set. .

10. The method for dynamically evaluating adjustable capacity through asymptotic vertex enumeration according to claim 1, characterized in that, The total time for a single complete iteration calculation is less than 200ms, the assessment error of the total adjustable capacity of the park does not exceed 5%, and the assessment error of the single type of load capacity does not exceed 3%.

Images