A method and device for evaluating power grid forward scheduling under multiple operation scenarios

Abstract

This invention discloses an evaluation method and apparatus for forward-looking dispatching of power grids under multiple operating scenarios. The method includes: obtaining the topology classification result of the current agent based on the classification results of all topology changes in the power grid, and obtaining evaluation net load scenario data based on the net load scenario data during the training of the current agent; obtaining multi-operation generalized scenario data; obtaining the forward-looking dispatching agent decision results under the multi-operation generalized scenario based on the power grid operation basic data and the multi-operation generalized scenario data, and obtaining the forward-looking dispatching benchmark decision results based on the power grid forward-looking dispatching benchmark model; obtaining a comprehensive evaluation index value of the forward-looking dispatching agent decision results based on the forward-looking dispatching agent decision results and the forward-looking dispatching benchmark decision results, and visualizing the comprehensive evaluation index value to obtain a visualization result. This invention can effectively improve the automation and intelligence level of forward-looking dispatching of power grids.

Description

Technical Field

[0001] This invention relates to the field of power system optimization scheduling technology, and in particular to an evaluation method and apparatus for forward scheduling of power grids under multiple operating scenarios. Background Technology

[0002] With the gradual advancement of my country's new power system construction, the number of entities involved in power system optimization and dispatch has increased, and power source volatility has intensified, significantly increasing the difficulty of forward-looking power grid dispatch. To address this challenge, various data-driven or partially data-driven intelligent algorithms have been applied to assist decision-making in forward-looking power grid dispatch, aiming to improve the automation and intelligence level of forward-looking power grid dispatch across multiple operating scenarios. Against this backdrop, considering the strong adaptability of forward-looking dispatch agents across multiple scenarios, a comprehensive evaluation of the decision-making performance of these agents is of practical significance in guiding the construction of intelligent forward-looking dispatch algorithms.

[0003] In the field of power grid economic dispatch assessment, existing research mainly treats it as a step in the design of economic dispatch agents, comparing it with traditional dispatch methods to verify the effectiveness of the agent design. The overall process is relatively simple, and there is a lack of research specifically targeting economic dispatch assessment. Some studies have proposed a comprehensive power grid economic dispatch assessment method that considers source-load uncertainty by constructing uncertainty calculation models on the source and load sides to simulate uncertainty in the assessment, thus improving the accuracy of power grid economic dispatch assessment. Other studies have proposed a power grid economic dispatch pre-assessment system for day-ahead generation dispatch plans, constructing a basic data acquisition subsystem, a standardized benchmark modeling subsystem, and a deviation rate calculation subsystem, improving the automation level of power grid economic dispatch pre-assessment. Still other studies have evaluated intelligent algorithms for power grid optimal dispatch from the dimensions of decision optimality and computational complexity, fully demonstrating the computational efficiency advantages of data-driven methods, using various data-driven approaches. Finally, some studies have investigated the impact of bad data in power grid forward dispatch, proposing a sensitivity matrix for bad data in forward dispatch to quickly assess the impact of missing or erroneous data during line congestion on forward dispatch decisions.

[0004] Existing research lacks a comprehensive evaluation method for the decision-making performance of forward-looking dispatching agents for power grids, and there are no relevant research and practical results on the evaluation of forward-looking dispatching agents for power grids in multiple operating scenarios. Summary of the Invention

[0005] The present invention aims to at least partially solve one of the technical problems in the related art.

[0006] Therefore, the purpose of this invention is to propose an evaluation method for forward scheduling of power grids under multiple operating scenarios. This method combines multiple topological scenarios and multiple net load scenarios of the power grid to construct multiple operating scenarios for the intelligent agent. The safety and economy of the intelligent agent's decision-making are evaluated based on the safe operating boundary and the forward scheduling benchmark decision, thereby achieving a comprehensive evaluation of the decision-making effect of the forward scheduling intelligent agent.

[0007] To achieve the above objectives, this invention proposes an evaluation method for forward-looking power grid dispatch under multiple operating scenarios, comprising:

[0008] The topology classification result of the current agent is obtained based on the classification results of all topology changes in the power grid, and the net load scenario data is evaluated based on the net load scenario data during the training of the current agent.

[0009] Based on the current agent's topology classification results and the net load scenario data, multi-run generalization scenario data is obtained;

[0010] Based on the power grid operation basic data and the multi-operation generalization scenario data, the decision results of the forward scheduling agent under the multi-operation generalization scenario are obtained, and the forward scheduling benchmark decision results are obtained based on the power grid forward scheduling benchmark model.

[0011] Based on the decision results of the forward scheduling agent and the decision results of the forward scheduling baseline, a comprehensive evaluation index value of the decision results of the forward scheduling agent is obtained, and the comprehensive evaluation index value is visualized to obtain a visualization result.

[0012] The evaluation method for forward-looking power grid dispatch under multiple operating scenarios according to embodiments of the present invention may also have the following additional technical features:

[0013] Furthermore, in one embodiment of the present invention, the method further includes: obtaining a power grid flow transfer distribution factor matrix corresponding to the topology structure based on the topology structure classification result of the current agent; and,

[0014] The net load scenario data during the current agent training and the node net load fluctuation value for each time period are used to obtain the evaluation net load scenario data, and the power grid node net load matrix corresponding to the net load scenario is obtained based on the evaluation net load scenario data.

[0015] Furthermore, in one embodiment of the present invention, the basic data for power grid operation includes multiple parameters such as: upper and lower limits of unit output, upper and lower limits of unit ramp-up and deceleration rates, total operating cost function of units, and line transmission capacity.

[0016] Furthermore, in one embodiment of the present invention, obtaining the forward scheduling benchmark decision result based on the power grid forward scheduling benchmark model includes: constructing a power grid forward scheduling benchmark model, and obtaining the objective function and constraints of the power grid forward scheduling benchmark model based on the power grid operation basic data; solving the power grid forward scheduling benchmark model based on the objective function and constraints, and calculating the power grid forward scheduling benchmark active power output in all forward windows to obtain the forward scheduling benchmark decision result under multiple operating scenarios.

[0017] Furthermore, in one embodiment of the present invention, obtaining the comprehensive evaluation index value of the forward scheduling agent's decision-making results based on the forward scheduling agent's decision-making results and the forward scheduling benchmark decision-making results includes: constructing a forward scheduling agent decision-making evaluation index system; obtaining all single-scenario evaluation index values ​​respectively using the forward scheduling agent's decision-making evaluation index system based on the forward scheduling agent's decision-making results and the forward scheduling benchmark decision-making results, and obtaining the comprehensive evaluation index value of the forward scheduling agent's decision-making results using the forward scheduling agent's decision-making evaluation index system.

[0018] To achieve the above objectives, another aspect of the present invention proposes an evaluation device for forward-looking power grid scheduling under multiple operating scenarios, comprising:

[0019] The classification acquisition module is used to obtain the topology classification result of the current agent based on the classification results of all topology changes in the power grid, and to obtain the evaluation net load scenario data based on the net load scenario data during the training of the current agent.

[0020] The scenario construction module is used to obtain multi-run generalized scenario data based on the topology classification results of the current agent and the evaluation net load scenario data;

[0021] The decision calculation module is used to obtain the decision results of the forward scheduling agent under the multi-operation generalization scenario based on the power grid operation basic data and the multi-operation generalization scenario data, and to obtain the forward scheduling benchmark decision results based on the power grid forward scheduling benchmark model.

[0022] The indicator evaluation module is used to obtain a comprehensive evaluation indicator value of the decision result of the forward scheduling agent based on the decision result of the forward scheduling agent and the decision result of the forward scheduling baseline, and to perform a visualization operation on the comprehensive evaluation indicator value to obtain a visualization result.

[0023] The evaluation method and apparatus for forward-looking dispatching of power grids under multiple operating scenarios in this invention construct multiple operating scenarios for intelligent agents by combining multiple topological scenarios and multiple net load scenarios of the power grid. The safety and economy of the intelligent agent's decision-making are evaluated based on the safe operating boundary and the forward-looking dispatching benchmark decision, respectively. This enables a comprehensive evaluation of the decision-making effect of the forward-looking dispatching intelligent agent, provides a reference for the construction and optimization of data-driven forward-looking dispatching intelligent agents, and can effectively improve the automation and intelligence level of forward-looking dispatching of power grids.

[0024] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0025] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0026] Figure 1 A flowchart illustrating the evaluation method for forward-looking power grid dispatching under multiple operating scenarios according to an embodiment of the present invention;

[0027] Figure 2 This is an architecture diagram of the evaluation method for forward-looking power grid scheduling under multiple operating scenarios according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the structure of an evaluation device for forward-looking power grid scheduling under multiple operating scenarios according to an embodiment of the present invention. Detailed Implementation

[0029] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0031] The following description, with reference to the accompanying drawings, describes the evaluation method and apparatus for forward-looking power grid scheduling under multiple operating scenarios according to embodiments of the present invention.

[0032] Figure 1 This is a flowchart of an evaluation method for forward-looking power grid scheduling under multiple operating scenarios, according to an embodiment of the present invention.

[0033] like Figure 1 As shown, the method includes, but is not limited to, the following steps:

[0034] S1. Obtain the topology classification result of the current agent based on the classification results of all topology changes in the power grid, and obtain the evaluation net load scenario data based on the net load scenario data during the training of the current agent.

[0035] S2, based on the current topology classification results of the intelligent agent and the net load scenario data, obtain multi-run generalization scenario data;

[0036] S3. Based on the basic data of power grid operation and the data of multiple operation generalization scenarios, the decision results of the forward scheduling agent under the multiple operation generalization scenarios are obtained, and the forward scheduling benchmark decision results are obtained based on the power grid forward scheduling benchmark model.

[0037] S4. Based on the decision results of the forward scheduling agent and the decision results of the forward scheduling baseline, a comprehensive evaluation index value of the decision results of the forward scheduling agent is obtained, and the comprehensive evaluation index value is visualized to obtain the visualization result.

[0038] The evaluation method for forward-looking power grid scheduling under multiple operating scenarios according to embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0039] like Figure 2 As shown, the evaluation method for forward-looking dispatching of the power grid under multiple operating scenarios specifically includes the following steps:

[0040] Step S100: Construct multiple operating scenarios for the intelligent agent, including three steps: S110: Obtain the power grid topology classification results; S120: Generate the net load assessment scenario; S130: Combine and construct the multi-operation generalization scenario.

[0041] Step S110, obtain the power grid topology classification results:

[0042] Based on the classification results of all power grid topology changes, the classification of the topology applicable to the current agent is selected, and all topologies in this category are numbered 1, 2, ..., S. T And record These are topologies 1, 2, ..., S. T The corresponding power flow transfer distribution factor matrix;

[0043] Step S120, generate the net load assessment scenario:

[0044] Based on the current payload scenario used in agent training, the node payload for each time period is set to fluctuate within a range of ±E%, and S is randomly generated. D Assess the net load scenario and record it. These are net load scenarios 1, 2, ..., S. DThe corresponding net load matrix of power grid nodes;

[0045] Step S130, combine and construct multi-run generalized scenarios:

[0046] Combining the topology obtained from S110 and S120 with the net load scenario, S is generated. T S D There are multiple generalized operating scenarios, each corresponding to a topology and a payload scenario, denoted as (s T ,s D (1≤s) T ≤S T ,1≤s D ≤S D );

[0047] Step S200 calculates forward scheduling decisions under multiple operating scenarios, including three steps: S210 acquires basic power grid operation data, S220 calculates forward scheduling agent decisions under multiple operating scenarios, and S230 calculates forward scheduling baseline decisions under multiple operating scenarios.

[0048] Step S210: Obtain basic power grid operation data:

[0049] Basic data for power grid operation include upper and lower limits of unit output, upper and lower limits of unit ramping acceleration and deceleration rates, total operating cost function of units, and line transmission capacity;

[0050] Step S220: Calculate the decision of the look-ahead scheduling agent under multiple running scenarios, as shown in the following expression:

[0051]

[0052] in For running scenarios (s) T ,s D Lookahead window t, t+1, ..., t+N τ -1 shows the active power output vector of the generator unit during the time period t+τ, where 0≤τ≤N. τ -1, N τ This refers to the number of forward-looking window periods; For net load scenario s D The node net load vector of the power grid in time period t; F(·) is the forward scheduling agent, whose input is the unit output vector of the previous time period and the net load vector of each time period in the forward window, and whose output is the unit output vector of all time periods in the current forward window.

[0053] In equation (1), For running scenarios (s) T ,s D The initial active power output vector of the unit; All represent All test periods, N T Number of test periods All represent All operating scenarios;

[0054] Step S230: Calculate the look-ahead scheduling baseline decision under multiple running scenarios:

[0055] Step S231, construct the power grid forward scheduling benchmark model:

[0056] The objective function is as follows:

[0057]

[0058] in For running scenarios (s) T ,s D Lookahead window t, t+1, ..., t+N τ -1 is the active power output vector of the unit in the forward scheduling baseline during the time period t+τ; C(·) is the total operating cost function of the unit; the objective of forward scheduling is to minimize the total operating cost of each unit within the forward window;

[0059] The constraints are as follows:

[0060]

[0061]

[0062]

[0063]

[0064] Where (3) represents the system power balance constraint; in the formula For running scenarios (s) T ,s D Lookahead window t, t+1, ..., t+N τ In -1, the active power output of unit i during the forward-looking scheduling baseline in time period t+τ is N. G The number of generating units; For net load scenario s D The net load of node j in time period t+τ, N D This represents the number of nodes in the power grid.

[0065] Equation (4) represents the output constraint of the generator set; where With P i These are the upper and lower limits of the output of unit i, respectively;

[0066] Equation (5) represents the constraint on the rate of increase or decrease in output of each unit; where RU i With RD iThese represent the increase and decrease in the upper limit of unit i's output in adjacent time periods, respectively; [Definition] For running scenarios (s) T ,s D The initial active power output of unit i;

[0067] Equation (6) represents the power flow constraint of the line; where F l N represents the transmission capacity of line l. L This refers to the number of lines; For the power grid topology s T The power flow transfer distribution factor of line l corresponding to node j; J(i) is the node number of unit i;

[0068] In the above objective function and constraints, All represent All generator sets, All represent All test periods All represent All time periods within the forward window, All represent All operating scenarios;

[0069] Step S232: Under each operating scenario, the linear programming solver is used to solve the above model in each forward window, and the active power output of the power grid forward scheduling benchmark within all forward windows is calculated, which is the forward scheduling benchmark decision under multiple operating scenarios.

[0070] Step S300 involves a comprehensive evaluation of the forward scheduling agent's decisions, which includes two steps: S310 statistical evaluation of the agent's decisions and S311 visual evaluation of the forward scheduling decision results.

[0071] Step S310: Calculate the comprehensive evaluation indicators for the agent's decision-making.

[0072] Step S311: Establish a decision evaluation index system for the forward-looking scheduling agent, expressed as follows:

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084] Equations (7)-(12) are comprehensive evaluation indicators, including one comprehensive economic indicator RCE and four comprehensive safety indicators RVM, RVS, NVC, NVT and decision availability η; Equations (13)-(17) are single-scenario evaluation indicators corresponding to the above indicators.

[0085] Equation (7) is the definition of the relative cost deviation of the comprehensive economic index, where RCE is the comprehensive relative cost deviation, which is determined by the relative cost deviation of each single scenario under each operating scenario. The average is obtained; For running scenarios (s) T ,s D The relative cost deviation of a single scenario under the scenario is calculated by the total operating cost of the forward scheduling agent decision and the total operating cost of the forward scheduling benchmark decision in all test periods and all forward windows under the scenario. The calculation formula is (13).

[0086] Equation (8) is the definition of the maximum relative limit exceedance of the comprehensive safety index, where RVM is the comprehensive maximum relative limit exceedance, which is determined by the maximum relative limit exceedance of a single scenario under each operating scenario. The maximum value is obtained by finding the maximum value. For running scenarios (s) T ,s D The maximum relative limit exceedance in a single scenario under the given conditions is the maximum relative limit exceedance of the forward scheduling agent's decision relative to each constraint boundary within all test periods and all forward windows under that scenario, and is calculated using formula (14); in formula (14) These are the maximum relative limit exceedance matrices for each test period and each look-ahead window corresponding to the generator output constraints, generator output rate constraints, and line power flow constraints in this scenario, respectively, calculated as (18)-(20); Define

[0087]

[0088]

[0089]

[0090] Equation (9) is the definition of the total relative over-limit of the comprehensive safety index, where RVS is the total relative over-limit of the comprehensive index, which is determined by the total relative over-limit of each single scenario under each operating scenario. Summation yields the result; For running scenarios (s) T ,s D The total relative limit exceedance in a single scenario under this scenario is the sum of the relative limit exceedances of the forward scheduling agent's decisions relative to each constraint boundary within all test periods and all forward windows under this scenario, and is calculated by formula (15); in formula (15) These are the relative over-limit total matrixes for each test period and each look-ahead window corresponding to the generator output constraint, generator output rate constraint and line power flow constraint in this scenario, and the calculation formulas are (21)-(23).

[0091]

[0092]

[0093]

[0094] Equation (10) is the definition of the average percentage of out-of-limit constraints in the comprehensive safety index, where NVC is the average percentage of out-of-limit constraints in the comprehensive index, which is determined by the percentage of out-of-limit constraints in each operating scenario. The average is obtained; For running scenarios (s) T ,s D The percentage of out-of-bounds constraints in a single scenario under the given conditions is the percentage of out-of-bounds constraints caused by the decisions of the look-ahead scheduling agent in all test periods and all look-ahead windows in that scenario, calculated as formula (16); in formula (16) These are the number matrices of the number of over-limit constraints for each test period and each look-ahead window corresponding to the generator output constraints, generator output rate constraints, and line power flow constraints in this scenario, respectively, and the calculation formulas are (24)-(26).

[0095]

[0096]

[0097]

[0098] Equation (11) is the definition of the average percentage of overdue periods for the comprehensive safety index, where NVT is the average percentage of overdue periods for the comprehensive safety index, which is determined by the percentage of overdue periods for each single scenario under each operating scenario. The average is obtained; For running scenarios (s) T ,s DThe percentage of time periods exceeding the limit in a single scenario under the scenario is the percentage of the number of time periods exceeding the limit due to the decision of the forward scheduling agent in all test periods and all forward windows in the scenario, and the calculation formula is (17).

[0099] Equation (12) is the definition of the availability of forward scheduling decision, where η is the availability of forward scheduling agent decision, which is defined as the proportion of the number of running scenarios in which the forward scheduling agent decision does not exceed the limit in the total number of running scenarios.

[0100] Step S312: Based on the forward scheduling decision results of the agent in multiple operating scenarios calculated in S220 and the forward scheduling benchmark decision results in multiple operating scenarios calculated in S230, the values ​​of all the above single-scenario evaluation indicators are calculated using formulas (13)-(17) for each operating scenario, and the values ​​of all comprehensive evaluation indicators are calculated using formulas (7)-(12).

[0101] Step S320: Visualize and evaluate the results of the forward scheduling decision.

[0102] Radar charts were used to plot the comprehensive evaluation index of the forward scheduling agent's decision-making calculated by equations (7)-(12), in order to evaluate and analyze the comprehensive effect of the forward scheduling agent's decision-making.

[0103] The evaluation method for forward-looking dispatching of power grids under multiple operating scenarios according to embodiments of the present invention constructs multiple operating scenarios of intelligent agents by combining multiple topological scenarios and multiple net load scenarios of power grids, and evaluates the safety and economy of intelligent agent decision-making based on safe operating boundaries and forward-looking dispatching benchmark decisions, thereby realizing a comprehensive evaluation of the decision-making effect of forward-looking dispatching intelligent agents, providing a reference for the construction and optimization of data-driven forward-looking dispatching intelligent agents, and effectively improving the automation and intelligence level of forward-looking dispatching of power grids.

[0104] To achieve the above embodiments, such as Figure 3 As shown, this embodiment also provides an evaluation device 10 for forward-looking scheduling of power grids under multiple operating scenarios. The device 10 includes: a classification acquisition module 100, a scenario construction module 200, a decision calculation module 300, and an indicator evaluation module 400.

[0105] The classification acquisition module 100 is used to obtain the topology classification result of the current agent based on the classification results of all topology changes in the power grid, and to obtain the evaluation net load scenario data based on the net load scenario data during the training of the current agent.

[0106] The scenario construction module 200 is used to obtain multi-run generalized scenario data based on the current topology classification results of the intelligent agent and the net load scenario data.

[0107] The decision calculation module 300 is used to obtain the decision results of the forward scheduling agent under the multi-operation generalization scenario based on the power grid operation basic data and multi-operation generalization scenario data, and to obtain the forward scheduling benchmark decision results based on the power grid forward scheduling benchmark model.

[0108] The indicator evaluation module 400 is used to obtain a comprehensive evaluation indicator value of the decision result of the forward scheduling agent based on the decision result of the forward scheduling agent and the decision result of the forward scheduling baseline, and to perform visualization operation on the comprehensive evaluation indicator value to obtain the visualization result.

[0109] Furthermore, the aforementioned device 10 also includes:

[0110] The first matrix acquisition module is used to obtain the power grid flow transfer distribution factor matrix corresponding to the current topology structure based on the topology structure classification result of the current agent; and,

[0111] The second matrix acquisition module is used to obtain the evaluation net load scenario data based on the net load scenario data during the current training of the intelligent agent and the node net load fluctuation value for each time period, and to obtain the power grid node net load matrix corresponding to the net load scenario data based on the evaluation net load scenario data.

[0112] Furthermore, basic power grid operation data includes: upper and lower limits of unit output, upper and lower limits of unit ramp-up and deceleration rates, total operating cost function of units, and various other data such as line transmission capacity.

[0113] Furthermore, the aforementioned decision calculation module 300 is also used for:

[0114] A benchmark model for forward-looking power grid scheduling is constructed, and the objective function and constraints of the benchmark model are obtained based on basic power grid operation data.

[0115] Based on the objective function and constraints, the forward scheduling benchmark model of the power grid is solved, and the active power output of the forward scheduling benchmark of the power grid within all forward windows is calculated to obtain the forward scheduling benchmark decision results under multiple operating scenarios.

[0116] Furthermore, the aforementioned indicator evaluation module 400 is also used for:

[0117] Construct a decision-making evaluation index system for forward-looking scheduling intelligent agents;

[0118] Based on the decision results of the forward scheduling agent and the decision results of the forward scheduling baseline, the evaluation index values ​​of all single scenarios are obtained by using the forward scheduling agent decision evaluation index system, and the comprehensive evaluation index value of the decision results of the forward scheduling agent is obtained by using the forward scheduling agent decision evaluation index system.

[0119] The evaluation device for forward-looking dispatching of power grids under multiple operating scenarios according to embodiments of the present invention constructs multiple operating scenarios of intelligent agents by combining multiple topological scenarios and multiple net load scenarios of power grids, and evaluates the safety and economy of intelligent agent decision-making based on safe operating boundaries and forward-looking dispatching benchmark decisions, thereby realizing a comprehensive evaluation of the decision-making effect of forward-looking dispatching intelligent agents, providing a reference for the construction and optimization of data-driven forward-looking dispatching intelligent agents, and effectively improving the automation and intelligence level of forward-looking dispatching of power grids.

[0120] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0121] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0122] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An evaluation method for forward-looking dispatching of power grids under multiple operating scenarios, characterized in that, Includes the following steps: Constructing multiple operational scenarios for the intelligent agent includes step S110, obtaining the power grid topology classification results: based on the classification results of all power grid topology changes, selecting the classification to which the current intelligent agent belongs to the applicable topology, and numbering all topologies in that category. And record Topology The corresponding power flow transfer distribution factor matrix; Step S120, Generate Evaluation Net Load Scenario: Based on the net load scenario used for the current agent training, set an upper and lower fluctuation range for the node net load in each time period. Randomly generated Assess the net load scenario and record it. Net load scenario The corresponding net load matrix of power grid nodes; Step S130, Combine and construct multi-run generalized scenarios: Combine the topology and payload scenarios obtained in S110 and S120 to generate There are multiple generalized operating scenarios, each corresponding to a topology and a payload scenario, denoted as . ( ); Step S200, calculate the look-ahead scheduling decision under multiple running scenarios, including: Step S210: Obtain basic power grid operation data: Basic power grid operation data includes upper and lower limits of unit output, upper and lower limits of unit ramping acceleration and deceleration rates, total operating cost function of units, and line transmission capacity; Step S220: Calculate the decision-making of the forward scheduling agent under multiple operating scenarios; Step S230, calculate the forward scheduling benchmark decision under multiple operating scenarios: Step S231, construct the power grid forward scheduling benchmark model; Step S232, under each operating scenario, use a linear programming solver to solve the above model in each forward window, calculate the power grid forward scheduling benchmark active power output in all forward windows, which is the forward scheduling benchmark decision under multiple operating scenarios. Based on the decision results of the forward scheduling agent and the decision results of the forward scheduling baseline, a comprehensive evaluation index value of the decision results of the forward scheduling agent is obtained, and the comprehensive evaluation index value is visualized to obtain a visualization result.

2. The method according to claim 1, characterized in that, The method further includes: Based on the topology classification result of the current agent, the power flow transfer distribution factor matrix corresponding to the topology is obtained; and, The net load scenario data during the current agent training and the node net load fluctuation value for each time period are used to obtain the evaluation net load scenario data, and the power grid node net load matrix corresponding to the net load scenario is obtained based on the evaluation net load scenario data.

3. The method according to claim 1, characterized in that, The basic data for power grid operation includes: upper and lower limits of unit output, upper and lower limits of unit ramp-up and deceleration rates, total operating cost function of units, and multiple data points including line transmission capacity.

4. The method according to claim 1, characterized in that, The process of obtaining the forward scheduling benchmark decision results based on the power grid forward scheduling benchmark model includes: A power grid forward scheduling benchmark model is constructed, and the objective function and constraints of the power grid forward scheduling benchmark model are obtained based on the power grid operation data. Based on the objective function and constraints, the power grid forward scheduling benchmark model is solved, and the active power output of the power grid forward scheduling benchmark within all forward windows is calculated to obtain the forward scheduling benchmark decision results under multiple operating scenarios.

5. The method according to claim 1, characterized in that, The comprehensive evaluation index value of the decision results of the look-ahead scheduling agent, obtained based on the decision results of the look-ahead scheduling agent and the look-ahead scheduling baseline decision results, includes: Construct a decision-making evaluation index system for forward-looking scheduling intelligent agents; Based on the decision results of the forward scheduling agent and the decision results of the forward scheduling baseline, the evaluation index values ​​of all single scenarios are obtained using the decision evaluation index system of the forward scheduling agent, and the comprehensive evaluation index value of the decision results of the forward scheduling agent is obtained using the decision evaluation index system of the forward scheduling agent.

6. An evaluation device for forward-looking power grid scheduling under multiple operating scenarios using the method described in claim 1, characterized in that, include: The classification acquisition module is used to obtain the topology classification result of the current agent based on the classification results of all topology changes in the power grid, and to obtain the evaluation net load scenario data based on the net load scenario data during the training of the current agent. The scenario construction module is used to obtain multi-run generalized scenario data based on the topology classification results of the current agent and the evaluation net load scenario data; The decision calculation module is used to obtain the decision results of the forward scheduling agent under the multi-operation generalization scenario based on the power grid operation basic data and the multi-operation generalization scenario data, and to obtain the forward scheduling benchmark decision results based on the power grid forward scheduling benchmark model. The indicator evaluation module is used to obtain a comprehensive evaluation indicator value of the decision result of the forward scheduling agent based on the decision result of the forward scheduling agent and the decision result of the forward scheduling baseline, and to perform a visualization operation on the comprehensive evaluation indicator value to obtain a visualization result.

7. The apparatus according to claim 6, characterized in that, The device further includes: The first matrix acquisition module is used to obtain the power grid flow transfer distribution factor matrix corresponding to the topology structure based on the topology structure classification result of the current agent; and, The second matrix acquisition module is used to obtain the evaluation net load scenario data based on the net load scenario data during the current training of the intelligent agent and the node net load fluctuation value for each time period, and to obtain the power grid node net load matrix corresponding to the net load scenario data based on the evaluation net load scenario data.

8. The apparatus according to claim 6, characterized in that, The basic data for power grid operation includes: upper and lower limits of unit output, upper and lower limits of unit ramp-up and deceleration rates, total operating cost function of units, and multiple data points including line transmission capacity.

9. The apparatus according to claim 6, characterized in that, The decision calculation module is also used for: A power grid forward scheduling benchmark model is constructed, and the objective function and constraints of the power grid forward scheduling benchmark model are obtained based on the power grid operation data. Based on the objective function and constraints, the power grid forward scheduling benchmark model is solved, and the active power output of the power grid forward scheduling benchmark within all forward windows is calculated to obtain the forward scheduling benchmark decision results under multiple operating scenarios.

10. The apparatus according to claim 6, characterized in that, The indicator evaluation module is also used for: Construct a decision-making evaluation index system for forward-looking scheduling intelligent agents; Based on the decision results of the forward scheduling agent and the decision results of the forward scheduling baseline, the evaluation index values ​​of all single scenarios are obtained using the decision evaluation index system of the forward scheduling agent, and the comprehensive evaluation index value of the decision results of the forward scheduling agent is obtained using the decision evaluation index system of the forward scheduling agent.

Images