Power optimization scheduling variable reduction method and system based on data driving
By building virtual machine groups and modular design, the computational burden problem of traditional power optimization scheduling after large-scale new energy access is solved, and efficient and flexible power system optimization is achieved.
Patent Information
- Application Number
- CN202510845462.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-23
- Publication Date
- 2025-10-03
AI Technical Summary
Traditional power optimization and scheduling methods face the complex and dynamic environment after the access of large-scale new energy and energy storage facilities, which results in excessive computational burden and makes it difficult to find high-precision optimization solutions within an acceptable time.
By building a virtual machine group, multiple units are aggregated into a virtual unit, and the actual operation of the power system is trained using simulation parameter optimization, decision variables are reduced, a large-scale integration model is built, and modular design and collaborative optimization technology are adopted.
It effectively reduces the number of decision variables, improves the efficiency and flexibility of power system optimization, ensures the real-time and reliability of optimization results, and supports continuous improvement.
Smart Images
Figure CN120749894A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power systems, and in particular to a data-driven power optimization scheduling variable reduction method and system. Background Art
[0002] As the construction of new power systems deepens, a vast number of market players are gradually entering the power trading market, including new electricity consumers and producers such as distributed energy, microgrids, and electric vehicles. The entry of these new market players has profoundly changed the structure and operation of the power market, posing new challenges to power system dispatch and optimization. In particular, the integration of large-scale renewable energy (such as wind power and photovoltaics) and energy storage facilities has further increased the complexity of power system dispatch decisions. These renewable energy and energy storage facilities are volatile and uncertain, and their output is affected by various factors such as weather and seasons, requiring more sophisticated and flexible dispatch strategies. Therefore, traditional power optimization dispatch methods are struggling in such a complex and dynamic environment, and new technologies and methods are urgently needed to improve dispatch efficiency and accuracy.
[0003] In order to meet the above challenges, improving the computational accuracy and speed of mathematical programming solvers has become an inevitable choice. Mathematical programming solvers are the core tools for power optimization and scheduling, and they achieve optimal scheduling of power systems by solving large-scale linear programming, integer programming, or mixed integer programming problems. However, with the increase in the number of market players and the access to new energy and energy storage facilities, the scale of problems to be solved has grown exponentially, resulting in huge pressure on existing solvers in terms of computing time and computing resources. In addition, the need for high-precision solutions also places higher demands on the algorithm, requiring the global optimal solution or high-quality approximate solution to be found within an acceptable time. Therefore, how to improve the computational performance of the solver so that it can efficiently handle large-scale and complex power optimization and scheduling problems has become one of the important directions of current research. Summary of the Invention
[0004] In response to the shortcomings of the existing technology, the present invention provides a data-driven power optimization scheduling variable reduction method and system, which has the advantages of reducing power optimization scheduling variables, reducing problem scale and accelerating solution, and solves the problem of the sudden increase in solution difficulty caused by the increasing scale of the problem in power optimization scheduling.
[0005] To achieve the above object, the present invention provides the following technical solutions: A data-driven power optimization scheduling variable reduction method includes the following steps: Sort out the basic parameters of the units, and combine multiple units to form a virtual unit based on the basic parameters of the units; The basic parameters of the unit are characterized, and the virtual machine group is randomly integrated according to the basic parameter information of the unit to obtain the parameter set of the virtual machine group; Taking the virtual machine group as the smallest unit, a scale integration model is constructed based on the parameter set of the virtual machine group and the functional requirements of the power system; The simulation parameters are used to optimize the scale integration model and train it to simulate the actual operation of the power plant, thus obtaining an integrated model for power optimization scheduling with a small number of variables.
[0006] Preferably, the basic parameters of the unit include the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
[0007] Preferably, the plurality of units are aggregated to form a virtual unit according to the basic parameters of the units, and the units are divided: setting is the set of all units, According to the collection of units, it is divided into mutually disjoint subsets, i.e. , ,use Indicates the index of a virtual machine group.
[0008] Preferably, the specific process of constructing a virtual machine group is: assigning different groups to different subsets The above subsets are represented as the constructed virtual machine group.
[0009] Preferably, the characteristic representation of the basic parameters of the unit is specifically: setting each unit The eigenvector of ,in is each dimensional element of the eigenvector, and each dimensional element of the eigenvector corresponds to the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
[0010] Preferably, according to the random integration of unit parameter information:
[0011] Where: Indicates the Subsets, and Represent different feature vectors in the subset respectively; The element representing the dimension of the initial start and stop state of the unit. Here, the units with the same initial start and stop state need to be integrated together, so as to use This penalty factor is used to indicate that the units with the same initial start-up and shutdown status must be in the same group. 08; express The distance threshold is ≤100.
[0012] Preferably, the virtual machine group is used as the smallest unit, and a scale integration model is constructed according to the parameter set of the virtual machine group and the functional requirements of the power system: the scale integration model includes a load forecasting module, a generator group scheduling module and an energy storage management module.
[0013] A data-driven power optimization scheduling variable reduction system includes a virtual machine group module, a parameter collection module, an integration module and a reduction optimization module; The virtual machine group module is used to sort out the basic parameters of the units and aggregate multiple units to form a virtual machine group based on the basic parameters of the units; The parameter set module is used to characterize the basic parameters of the unit and randomly integrate the virtual machine group according to the basic parameter information of the unit to obtain the parameter set of the virtual machine group; The integration module is used to build a scale integration model based on the virtual machine group as the smallest unit and the parameter set of the virtual machine group and the functional requirements of the power system; The reduction optimization module uses simulation parameters to optimize the scale integration model and train it to simulate the actual operation of the power plant, thus obtaining an integrated power optimization scheduling model with a small number of variables.
[0014] Preferably, the basic parameters of the unit include the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
[0015] Preferably, the plurality of units are aggregated to form a virtual unit according to the basic parameters of the units, and the units are first divided: is the set of all units, According to the collection of units, it is divided into mutually disjoint subsets, i.e. , ,use Indicates the index of a virtual machine group.
[0016] Preferably, the specific process of constructing a virtual machine group is: assigning different groups to different subsets The above subsets are represented as the constructed virtual machine group.
[0017] Preferably, the characteristic representation of the basic parameters of the unit is specifically: setting each unit The eigenvector of ,in is each dimensional element of the eigenvector, and each dimensional element of the eigenvector corresponds to the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
[0018] Preferably, according to the random integration of unit parameter information:
[0019] Where: Indicates the Subsets, and Represent different feature vectors in the subset respectively; The element representing the dimension of the initial start and stop state of the unit. Here, the units with the same initial start and stop state need to be integrated together, so as to use This penalty factor is used to indicate that the units with the same initial start-up and shutdown status must be in the same group. 08; express The distance threshold is ≤100.
[0020] Preferably, the virtual machine group is used as the smallest unit, and a scale integration model is constructed according to the parameter set of the virtual machine group and the functional requirements of the power system: the scale integration model includes a load forecasting module, a generator group scheduling module and an energy storage management module.
[0021] Compared with the existing technology, the present invention provides a data-driven power optimization scheduling variable reduction method, which has the following beneficial effects: The present invention is based on a data-driven power optimization scheduling variable reduction method. It constructs a virtual machine group by aggregating multiple units according to the basic parameters of the units; the basic parameters of the units are characterized, and the virtual machine groups are randomly integrated according to the basic parameter information of the units to obtain a parameter set strategy for the virtual machine group. The complex power system is decomposed into multiple virtual machine groups with similar characteristics, which can effectively reduce the number of decision variables and thus reduce the complexity of the problem. This simplification not only makes the problem easier to manage and solve, but also improves the overall optimization efficiency.
[0022] Preferably, each virtual machine group is taken as the smallest unit, and a scale integration model is constructed according to the parameter set of each virtual machine group and the functional requirements of the power system. Corresponding optimization technologies are used for different modules in the scale integration model (such as load forecasting module, generator group scheduling module and energy storage management module), which helps to further improve processing efficiency. Each module focuses on its specific function, so that the algorithm can operate more accurately and efficiently, and the collaborative optimization and feedback mechanism between modules can enhance the flexibility of the system. This design allows the system to quickly adjust and respond when facing changing conditions and demands to ensure the real-time and effectiveness of the optimization results by constructing a virtual power plant simulation model.
[0023] This invention can verify the effectiveness of the algorithm. By comparing model simulation results with actual data, it can ensure the reliability of the solution and help identify and correct potential problems. Performance evaluation and adjustment strategies provide the possibility for continuous improvement. By setting clear evaluation indicators, the performance of the algorithm can be quantified, and the algorithm can be adjusted and optimized based on these indicators to improve the practicality of the method. By reducing complexity, improving efficiency, enhancing flexibility and adaptability, ensuring verifiability, and supporting continuous improvement, this invention provides a comprehensive and effective solution for power system optimization.
[0024] By comparing the model simulation results with the actual data, the present invention can ensure the reliability of the solution and help identify and correct potential problems. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 This is a flow chart of the data-driven power optimization scheduling variable reduction method of the present invention.
[0026] Figure 2 This is a structural diagram of the data-driven power optimization scheduling variable reduction system of the present invention. DETAILED DESCRIPTION The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0027] See also Figure 1 , a data-driven power optimization scheduling variable reduction method, comprising the following steps: Based on the basic parameters of the units, multiple units are aggregated to form a virtual machine group. The basic parameters of the units are characterized and randomly integrated based on the basic parameter information of the units to obtain the parameter set of the virtual machine group, thereby reducing the number of decision variables (units with similar characteristics are aggregated into a virtual machine group based on the unit's location, type, or other relevant characteristics). Record and analyze the unit parameters within each virtual machine group to obtain the parameter set of each virtual machine group (reducing the number of variables to be optimized and simplifying the optimization model); Taking the virtual machine group as the new minimum unit, a scale integration model is constructed based on the parameter set of each virtual machine group and the functional requirements of the power system. The scale integration model retains the main characteristics of the original problem, but is smaller in scale and easier to solve. In the specific implementation of the present application, the algorithm optimization process is divided into a load forecasting module, a generator group scheduling module, and an energy storage management module in the scale integration model; different functional optimization algorithms are set within each module to improve the overall optimization efficiency; information transmission between modules is established through the network to ensure coordination and consistency between the modules; The scale integration model is optimized and trained using simulation parameters to simulate the actual operation of the power plant. The scale integration model receives module parameters through the network, and the effectiveness of the scale integration model is verified by comparing the model simulation results with the actual data. According to the simulation operation parameters, the units are combined into virtual units, and an integrated model of power optimization scheduling with a small number of variables is obtained to improve the practicality and adaptability of the model.
[0028] The data-driven power optimization scheduling variable reduction method adopted by the present invention combines multiple physical units into a virtual machine group, records and analyzes the unit parameters in each virtual machine group, obtains a parameter set strategy for each virtual machine group, and decomposes the complex power system into multiple virtual machine groups with similar characteristics. It can effectively reduce the number of decision variables and thus reduce the complexity of the problem. This simplification not only makes the problem easier to manage and solve, but also improves the overall optimization efficiency.
[0029] Taking each virtual machine group as the smallest unit, a scaled integration model is constructed based on the parameter set of each virtual machine group and the functional requirements of the power system. Using corresponding optimization techniques for different modules in the scaled integration model (such as the load forecasting module, the generator group scheduling module, and the energy storage management module) helps further improve processing efficiency. Each module focuses on its specific function, enabling the algorithm to operate more accurately and efficiently. The collaborative optimization and feedback mechanism between modules enhances the flexibility of the system. This design allows the system to quickly adjust and respond to changing conditions and requirements, ensuring the real-time and effectiveness of the optimization results. By constructing a virtual power plant simulation model, the present invention can verify the effectiveness of the algorithm. By comparing the model simulation results with actual data, it can ensure the reliability of the solution and help identify and correct potential problems. Performance evaluation and adjustment strategies provide the possibility of continuous improvement. By setting clear evaluation indicators, the performance of the algorithm can be quantified, and the algorithm can be adjusted and optimized based on these indicators to improve the practicality of the method. The present invention provides a comprehensive and effective solution for power system optimization by reducing complexity, improving efficiency, enhancing flexibility and adaptability, ensuring verifiability, and supporting continuous improvement.
[0030] The basic parameters of the unit include the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the startup cost, the shutdown cost, the ramp-up rate and the initial start and stop status of the unit (because each unit has certain physical properties and parameters for economic dispatch operation, these parameters can be used as characteristics of each unit to distinguish them).
[0031] Divide the units: Set is the set of all units, According to the collection of units, it is divided into mutually disjoint subsets, i.e. , ,use Indicates the index of a virtual machine group.
[0032] The specific process of combining multiple physical groups into a virtual machine group is: assigning different groups to different sub-groups The above subsets are represented as the constructed virtual machine group.
[0033] The basic parameters of the unit are characterized by: setting each unit The eigenvector of ,in is each dimensional element of the characteristic vector, and each dimensional element of the characteristic vector corresponds to the unit's maximum output value, unit minimum output value, minimum continuous start time, minimum continuous shutdown time, start cost, shutdown cost, ramp rate and unit initial start and stop state parameters.
[0034] According to the random integration of basic parameter information of the unit:
[0035] Where: Indicates the Subsets, and Represent different feature vectors in the subset respectively; The element representing the dimension of the initial start and stop state of the unit. Here, the units with the same initial start and stop state need to be integrated together, so as to use This penalty factor is used to indicate that the units with the same initial start-up and shutdown status must be in the same group. 08; express The distance threshold is ≤100, the units with the same basic parameters are integrated together to obtain the parameter set of the virtual machine group.
[0036] In another embodiment of the present invention, Figure 2 As shown, a data-driven power optimization scheduling variable reduction system includes a virtual machine group module, a parameter collection module, an integration module and a reduction optimization module; The virtual machine group module is used to sort out the basic parameters of the units and aggregate multiple units to form a virtual machine group based on the basic parameters of the units; The parameter set module is used to characterize the basic parameters of the unit and randomly integrate the virtual machine group according to the basic parameter information of the unit to obtain the parameter set of the virtual machine group; The integration module is used to build a scale integration model based on the virtual machine group as the smallest unit and the parameter set of the virtual machine group and the functional requirements of the power system; The reduction optimization module uses simulation parameters to optimize the scale integration model and train it to simulate the actual operation of the power plant, thus obtaining an integrated power optimization scheduling model with a small number of variables.
[0037] In a specific embodiment of the present application, the basic parameters of the unit include the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the startup cost, the shutdown cost, the ramp-up rate and the initial start and stop status of the unit (since each unit has certain physical properties and parameters for economic scheduling operation, these parameters can be used as characteristics of each unit to distinguish them).
[0038] According to the basic parameters of the units, multiple units are combined to form virtual units, and the units are divided: set is the set of all units, According to the collection of units, it is divided into mutually disjoint subsets, i.e. , ,use Indicates the index of a virtual machine group.
[0039] The specific process of constructing a virtual machine group is: assign different groups to different subsets The above subsets are represented as the constructed virtual machine group.
[0040] The basic parameters of the unit are characterized by: setting each unit The eigenvector of ,in is each dimensional element of the eigenvector, and each dimensional element of the eigenvector corresponds to the parameters of the maximum output value and the minimum output value of the unit in turn.
[0041] According to the random integration of basic parameter information of the unit:
[0042] Where: Indicates the Subsets, and Represent different feature vectors in the subset respectively; The element representing the dimension of the initial start and stop state of the unit. Here, the units with the same initial start and stop state need to be integrated together, so as to use This penalty factor is used to indicate that the units with the same initial start-up and shutdown status must be in the same group. 08; express The distance threshold is ≤100, the units with the same basic parameters are integrated together to obtain the parameter set of the virtual machine group.
[0043] After the machine group is combined into a virtual machine group according to the setting of random parameters, the corresponding constraints and objective functions are also adjusted and transformed to obtain a large-scale integration model with a small number of variables.
[0044] In a specific embodiment of the present application, the following data-driven power optimization scheduling variable reduction system is provided, and the implementation process of the embodiment is as follows: System Settings Unit set: A collection of multiple generating units, each with its own unique physical properties and economic dispatch operating parameters.
[0045] Unit parameters: The basic parameters of each unit include maximum output value, minimum output value, minimum continuous on time, minimum continuous off time, start-up cost, shutdown cost, ramp-up rate and initial start-up and shutdown status of the unit.
[0046] Eigenvector: The eigenvector of each unit is composed of its basic parameters, including maximum output value and minimum output value.
[0047] Set partitioning: Divide all generator sets into mutually disjoint subsets, each of which represents a virtual machine group.
[0048] Scale integration model establishment: Taking the virtual machine group as the smallest unit, a scale integration model is constructed based on the parameter set of the virtual machine group and the functional requirements of the power system. The constructed scale integration model includes: Load forecasting module: responsible for forecasting power load and providing basic data for dispatching; Generator group scheduling module: schedules generator groups based on load forecast results and virtual group parameters.
[0049] Energy storage management module: manages energy storage devices and optimizes the charging and discharging strategies of energy storage devices.
[0050] Randomly ensemble penalty factors in medium-scale ensemble models: Penalty factor: Use the penalty factor to ensure that units with the same initial start-up and shutdown status are integrated together; Distance threshold: Set a distance threshold to integrate units with similar basic parameters.
[0051] Optimization and adjustment of scale integration model: According to the generation of virtual machine groups, the corresponding constraints and objective functions are adjusted and transformed to obtain an integrated model of power optimization scheduling with fewer variables.
[0052] Implementation steps S1.1. Data collection: Collect basic parameter data of all generator sets; S1.2, Feature vector construction: construct a feature vector for each unit; S1.3, Subset division: Divide the units into different subsets based on the eigenvectors and penalty factors; S1.4, treat the subset as a virtual machine group; S1.5. Model construction: Construct a small-scale integrated model, including the objective function and constraints; S1.6, Module Verification: Verify the model through the load forecasting module, generator scheduling module and energy storage management module; S1.7. Model adjustment: Adjust model parameters based on verification results to optimize model performance.
[0053] Comparison with existing data-driven power optimization scheduling methods: System Settings Unit set: A collection of multiple generating units, each with its own unique physical properties and economic dispatch operating parameters.
[0054] Unit parameters: The basic parameters of each unit include maximum output value, minimum output value, minimum continuous on time, minimum continuous off time, start-up cost, shutdown cost, ramp-up rate and initial start-up and shutdown status of the unit.
[0055] Eigenvector: There is no concept of eigenvector, and the basic parameters of the unit are directly used for classification.
[0056] Set partitioning: Simple clustering analysis is performed directly based on basic parameters, without the concept of explicit penalty factors and distance thresholds.
[0057] Module creation: A large-scale optimization scheduling model is directly constructed based on unit parameters without variable reduction or model simplification.
[0058] Comparison Summary
[0059] Example 2 System settings: When there are 10 generator sets, the basic parameters of each unit are as follows:
[0060] Implementation steps S2.1. Data preprocessing: Organize unit data into feature vectors; S2.2, Feature vector construction: Construct the feature vector of each unit. For example, for unit 1, the feature vector is [100, 20]; S2.3. Subset division: Use clustering algorithm to divide units with similar feature vectors into a subset; S2.4. Consider the subset as a virtual machine group. For example, the subset {1, 2, 3} is a virtual machine group. S2.5. Model construction: Construct a small-scale integrated model, including the objective function and constraints; S2.6, Module Verification: Verify the model through the load forecasting module, generator scheduling module and energy storage management module; S2.7. Model adjustment: Adjust model parameters based on verification results to optimize model performance.
[0061] Comparative Example 2: Power Optimization Scheduling Method Based on Existing Data When there are 10 generator sets, the basic parameters of each unit are as follows:
[0062] By comparing Example 1 with Comparative Example 1, and Example 2 with Comparative Example 2, it can be clearly seen that the data-driven power optimization scheduling variable reduction system of the embodiment of the present invention has advantages in reducing the number of decision variables, simplifying the optimization model, and improving scheduling efficiency and accuracy. The embodiment overcomes the shortcomings of traditional methods through modular design, dynamic adjustment and simulation verification, and provides strong support for the efficient operation of modern power systems.
[0063] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.
Claims
1. A data-driven power optimization scheduling variable reduction method, characterized in that: The following steps are involved: Sort out the basic parameters of the units, and combine multiple units to form a virtual unit based on the basic parameters of the units; The basic parameters of the unit are characterized, and the virtual machine group is randomly integrated according to the basic parameter information of the unit to obtain the parameter set of the virtual machine group; Taking the virtual machine group as the smallest unit, a scale integration model is constructed based on the parameter set of the virtual machine group and the functional requirements of the power system; The simulation parameters are used to optimize the scale integration model and train it to simulate the actual operation of the power plant, thus obtaining an integrated model for power optimization scheduling with a small number of variables.
2. The data-driven power optimization scheduling variable reduction method according to claim 1 is characterized in that: The basic parameters of the unit include the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
3. The data-driven power optimization scheduling variable reduction method according to claim 1 is characterized in that: According to the basic parameters of the units, multiple units are aggregated to form a virtual unit, and the units are divided: is the set of all units, According to the collection of units, it is divided into mutually disjoint subsets, i.e. , ,use Indicates the index of a virtual machine group.
4. The data-driven power optimization scheduling variable reduction method according to claim 3 is characterized in that: The specific process of constructing a virtual machine group is: assign different groups to different subsets The above subsets are represented as the constructed virtual machine group.
5. The data-driven power optimization scheduling variable reduction method according to claim 4 is characterized in that: The characteristic representation of the basic parameters of the unit is specifically as follows: The eigenvector of ,in is each dimensional element of the eigenvector, and each dimensional element of the eigenvector corresponds to the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
6. The data-driven power optimization scheduling variable reduction method according to claim 1 is characterized in that: According to the random integration of unit parameter information: Where: Indicates the Subsets, and Represent different feature vectors in the subset respectively; The element representing the dimension of the initial start and stop state of the unit. Here, the units with the same initial start and stop state need to be integrated together, so as to use This penalty factor is used to indicate that the units with the same initial start-up and shutdown status must be in the same group. 08; express The distance threshold is ≤100.
7. The data-driven power optimization scheduling variable reduction method according to claim 1 is characterized in that: The virtual machine group is used as the smallest unit, and a scale integration model is constructed according to the parameter set of the virtual machine group and the functional requirements of the power system: the scale integration model includes a load forecasting module, a generator group scheduling module and an energy storage management module.
8. A data-driven power optimization scheduling variable reduction system, characterized in that: It includes virtual machine group module, parameter collection module, integration module and reduction optimization module; The virtual machine group module is used to sort out the basic parameters of the units and aggregate multiple units to form a virtual machine group based on the basic parameters of the units; The parameter set module is used to characterize the basic parameters of the unit and randomly integrate the virtual machine group according to the basic parameter information of the unit to obtain the parameter set of the virtual machine group; The integration module is used to build a scale integration model based on the virtual machine group as the smallest unit and the parameter set of the virtual machine group and the functional requirements of the power system; The reduction optimization module uses simulation parameters to optimize the scale integration model and train it to simulate the actual operation of the power plant, thus obtaining an integrated power optimization scheduling model with a small number of variables.
9. The data-driven power optimization scheduling variable reduction system according to claim 8, characterized in that: The basic parameters of the unit include the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
10. The data-driven power optimization scheduling variable reduction system according to claim 8, characterized in that: The virtual machine group is constructed by combining multiple units according to the basic parameters of the units. First, the units are divided: is the set of all units, According to the collection of units, it is divided into mutually disjoint subsets, i.e. , ,use Indicates the index of a virtual machine group.
11. The data-driven power optimization scheduling variable reduction system according to claim 10, characterized in that: The specific process of constructing a virtual machine group is: assign different groups to different subsets The above subsets are represented as the constructed virtual machine group.
12. The data-driven power optimization scheduling variable reduction system according to claim 11, characterized in that: The characteristic representation of the basic parameters of the unit is specifically as follows: The eigenvector of ,in is each dimensional element of the eigenvector, and each dimensional element of the eigenvector corresponds to the maximum output value of the unit, the minimum output value of the unit, the minimum continuous start time, the minimum continuous shutdown time, the start cost, the shutdown cost, the ramp rate and the initial start and stop status of the unit.
13. The data-driven power optimization scheduling variable reduction system according to claim 8, characterized in that: According to the random integration of unit parameter information: Where: Indicates the Subsets, and Represent different feature vectors in the subset respectively; The element representing the dimension of the initial start and stop state of the unit. Here, the units with the same initial start and stop state need to be integrated together, so as to use This penalty factor is used to indicate that the units with the same initial start-up and shutdown status must be in the same group. 08; express The distance threshold is ≤100.
14. The data-driven power optimization scheduling variable reduction system according to claim 8, characterized in that: The virtual machine group is used as the smallest unit, and a scale integration model is constructed according to the parameter set of the virtual machine group and the functional requirements of the power system: the scale integration model includes a load forecasting module, a generator group scheduling module and an energy storage management module.