Micro-grid layered collaborative scheduling method and system for power system
By constructing a three-layer optimization model and a distributed parallel ADMM algorithm, the problems of multi-level coordination and low computational efficiency in the scheduling of virtual power plants and microgrids in power systems are solved. This achieves efficient and reliable hierarchical collaborative scheduling, which is suitable for large-scale microgrid scenarios and protects the privacy of microgrids.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-29
- Publication Date
- 2026-03-31
AI Technical Summary
Existing power systems lack multi-level coordination mechanisms in virtual power plant-microgrid dispatching, resulting in low computational efficiency, difficulty in coping with complex coupling constraints in large-scale microgrid scenarios, and a tendency to encounter computational bottlenecks and single-point failures.
A hierarchical collaborative scheduling method is adopted to construct a three-layer optimization model consisting of a market scheduling layer, a virtual power plant layer, and a microgrid layer. Consistency variables and augmented Lagrangian functions are introduced to couple the model, and the distributed parallel ADMM algorithm is used for solving. By combining deviation penalties and ancillary service reward settlement, hierarchical collaborative scheduling is achieved.
It improves the reliability, accuracy and efficiency of power system microgrid dispatch, supports real-time or near-real-time dispatch in large-scale microgrid scenarios, protects the privacy of microgrid operation, and enhances user data security.
Smart Images
Figure CN121769895A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electrical automation, and specifically relates to a microgrid hierarchical collaborative scheduling method and system for power systems. Background Technology
[0002] With economic and technological development and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and daily life, bringing endless convenience. Therefore, ensuring a stable and reliable supply of electricity has become one of the most important tasks of the power system.
[0003] At present, more and more new energy power generation systems are being integrated into the power system and generating electricity. The output fluctuation and uncertainty of new energy power generation systems, as well as the multi-level aggregation of virtual power plants and microgrids, make it difficult for traditional centralized scheduling and optimization schemes to meet the requirements in terms of computational efficiency, robustness and scalability. Existing schemes have the following main problems when scheduling virtual power plants-microgrids in the power system: (1) Lack of multi-level coordination mechanism: Existing schemes either stay at the overall level of virtual power plants or only focus on local optimization of microgrids; (2) Limited computational efficiency: Existing schemes rely on centralized optimization, which is difficult to cope with the complex coupling constraints in large-scale microgrid scenarios, and is prone to computational bottlenecks and single-point failures. Summary of the Invention
[0004] One of the objectives of this invention is to provide a highly reliable, accurate, and efficient microgrid hierarchical collaborative scheduling method for power systems.
[0005] The second objective of this invention is to provide a system for implementing the aforementioned microgrid hierarchical collaborative scheduling method for power systems.
[0006] The microgrid hierarchical collaborative scheduling method for power systems provided by this invention includes the following steps:
[0007] S1. Acquire data information of the target power system and preprocess it to obtain an optimized scheduling dataset;
[0008] S2. Based on the data obtained in step S1, construct a three-layer optimization model consisting of a market dispatch layer, a virtual power plant layer, and a microgrid layer, and introduce consistency variables and augmented Lagrangian functions to couple the model.
[0009] S3. Based on the distributed parallel ADMM algorithm, the model constructed in step S2 is solved to obtain the virtual power plant-microgrid optimized scheduling scheme;
[0010] S4. The dispatch center publishes the plan obtained in step S3, makes corrections based on the actual operation of the target power system, and settles the corresponding deviation penalties and ancillary service rewards.
[0011] S5. Perform encrypted data interaction to achieve hierarchical collaborative scheduling of the microgrid in the target power system.
[0012] Step S1, which involves acquiring data information from the target power system and preprocessing it to obtain an optimized scheduling dataset, specifically includes the following steps:
[0013] Acquire data information from the target power system; the data information includes:
[0014] The reported quantity and bid data of coal-fired unit i in coal-fired unit set G, the reported quantity and bid data of virtual power plant j in virtual power plant set V, and the microgrid set under virtual power plant j. The data includes the reported load and price quotation data of microgrid k, the load forecast curve of the target power system, and the new energy output forecast curve and upper and lower limit data of coal-fired power unit output of microgrid k under virtual power plant j.
[0015] The acquired data is preprocessed, including noise reduction and correction; the acquired load forecast curve is smoothed.
[0016] Step S2 involves constructing a three-layer optimization model—comprising a market dispatch layer, a virtual power plant layer, and a microgrid layer—based on the data obtained in step S1, and introducing consistency variables and augmented Lagrangian functions for model coupling. Specifically, this includes the following steps:
[0017] Market scheduling layer:
[0018] The market scheduling layer is used for overall economic and safety scheduling; decision variables include the output of each coal-fired unit i at time t. and the output of each virtual power plant j at time t ;
[0019] The following formula is used as the objective function of the market scheduling layer:
[0020] In the formula The objective function of the market scheduling layer; The scheduling period; Let be the equivalent bid-price cost function for coal-fired unit i, and , The first equivalent fuel cost coefficient, This is the second equivalent fuel cost coefficient. The first equivalent fuel cost coefficient; Let be the cost function of the virtual power plant j, and , The adjustable output range of virtual power plant j The total number of consecutive segments. Let be the maximum adjustable output of the virtual power plant j. Let J be the output value of the virtual power plant j at time t in segment s. and , This represents the upper limit of output for segment s. This represents the lower limit of the output of segment s. The unit output cost of virtual power plant j in section s;
[0021] The constraints include system power balance constraints, upper and lower limits of thermal power unit output constraints, thermal power unit output ramping constraints, upper and lower limits of virtual power plant output constraints, and virtual power plant output ramping constraints; among them, the system power balance constraint is expressed as... , Let be the load demand at time t;
[0022] Virtual power plant layer:
[0023] The virtual power plant layer satisfies the output requirements of each virtual power plant j at time t issued by the market dispatch layer. Under the guidance of instructions, optimize the various microgrids under its jurisdiction;
[0024] The following formula is used as the objective function for the virtual power plant layer:
[0025] In the formula The objective function for the virtual power plant layer; Let be the output of the k-th microgrid under the virtual power plant j at time t. It is a cost function, and , The adjustable output range of k microgrids under virtual power plant j The total number of consecutive segments. Let be the maximum adjustable output of the k microgrids under the virtual power plant j. Let be the output value of the k microgrids under the virtual power plant j in segment s at time t. and , This represents the upper limit of output for segment s. This represents the lower limit of the output of segment s. The unit output cost of the k microgrids under the virtual power plant j in section s; This is the deviation penalty coefficient; The upward execution bias slack variable is set. The downward execution bias slack variable is set, and ; This is the output smoothing penalty coefficient; Rewards for virtual power plants to submit unit standby capacity to higher levels; The ancillary service capacity reported by the virtual power plant to the market dispatch layer, wherein the ancillary services include reserve and peak shaving; To measure the penalty coefficient for unfairness caused by uneven internal distribution; Let J be the average output of all microgrids under the virtual power plant j, and , The total number of microgrids under virtual power plant j;
[0026] The constraints include a consistency constraint on aggregate output; the consistency constraint on aggregate output is expressed as follows: ;
[0027] Microgrid layer:
[0028] The microgrid layer meets the output targets set by the virtual power plant. Under the premise of optimization, the microgrid layer performs its own optimization; the decision variables of the microgrid layer include the l-th wind power output of the j-th virtual power plant and the k-th microgrid. The output of the m-th photovoltaic cell The charging power of the s-th energy storage Discharge power of energy storage Micro gas turbine power and start / stop state variables ;
[0029] The following formula is used as the objective function for the microgrid layer:
[0030] In the formula The objective function of the microgrid layer; The cost of penalizing the abandonment of wind and solar power; Energy storage losses and cycle costs; Costs of operating and starting / shutting micro gas turbines; Cost per unit of wind curtailment; This refers to the available power output of wind power. Cost per unit of abandoned light; The photovoltaic power output is available. Cost per unit of energy storage loss; The startup cost of a single gas turbine; The shutdown cost for a single gas turbine;
[0031] The constraints include power balance constraints, energy storage SOC state balance constraints, wind turbine output constraints, photovoltaic output constraints, energy storage output constraints, and gas turbine output constraints;
[0032] The power balance constraint is expressed as:
[0033] In the formula Let be the load demand of the k-th microgrid of virtual power plant j;
[0034] The energy storage SOC state equilibrium constraint is expressed as:
[0035] In the formula Let be the state of charge of the s-th energy storage in the k-th microgrid of virtual power plant j at time t; For charging efficiency; It is discrete time; For discharge efficiency;
[0036] Constructing Consistent Variables and Constraints:
[0037] Variables of the market scheduling layer Introducing global variables Indicates variables for the market scheduling layer. Introducing local variables The system power balance constraint of the market scheduling layer is then expressed as: At the same time, add constraints. ;
[0038] By introducing Lagrange multipliers and a quadratic penalty term, we construct the augmented Lagrange function. , is represented as:
[0039] In the formula In order to be in For vectors , For vectors The objective function of the market scheduling layer at that time; The Lagrange multiplier corresponding to the system power balance constraint; The coefficient of the second-order penalty term;
[0040] The virtual power plant layer and the microgrid layer are merged into a local subproblem of virtual power plant, and each virtual power plant subproblem internally handles the microgrid; the local Lagrangian function of the j-th virtual power plant is constructed. , is represented as:
[0041] In the formula Let be the local cost of the j-th virtual power plant, and .
[0042] Step S3, based on the distributed parallel ADMM algorithm, solves the model constructed in step S2 to obtain the virtual power plant-microgrid optimized scheduling scheme, specifically including the following steps:
[0043] The distributed parallel ADMM algorithm is used to solve the model constructed in step S2:
[0044] A. Initialize variables , and ;
[0045] B. The market scheduling layer uses the ADMM algorithm to update vector z: in the (r+1)th iteration, the value is fixed. and The market scheduling layer solves the following model: st
[0046] C. Distributed parallel local update vectors for each virtual power plant and :fixed and For each virtual power plant, the following local optimization subproblem is solved in parallel: st
[0047] D. Update the Lagrange multipliers: ;
[0048] E. Convergence criterion of ADMM algorithm: when the following conditions are met Or satisfy When the iteration stops, the iteration continues; where, The step size for updating the dual variable is used to adjust the penalty strength of the consistency constraint. Its value reflects the coordination strength between the system layer and the virtual power plant-microgrid, and can achieve a balance between the algorithm's convergence speed and stability. The required accuracy for the initial residuals. The accuracy requirements that the dual residuals need to meet.
[0049] Step S4 involves the dispatch center publishing the scheme obtained in step S3, making corrections based on the actual operating conditions of the target power system, and settling corresponding deviation penalties and ancillary service rewards. Specifically, this includes the following steps:
[0050] a. Results Release: The dispatch center will send the output values of each virtual power plant to the corresponding virtual power plant, and each virtual power plant will then allocate and execute the output tasks of its own microgrid.
[0051] b. Operational calibration: Obtain the actual output value of each microgrid. And calculate the deviation value. for ; This is the sum of the output values of all virtual power plants obtained in step S3;
[0052] like Greater than the set threshold Then the virtual power plant or microgrid will perform its own correction;
[0053] c. The dispatch center rewards virtual power plants: the incentive reward is calculated using the following formula. :
[0054] In the formula Rewards for ancillary services provided to virtual power plants; A positive deviation penalty coefficient is set.
[0055] d. Distribute the incentive rewards obtained in step c among the various microgrids to which the virtual power plant belongs;
[0056] e. The dispatch center records and stores the results of each microgrid.
[0057] Step S5, which involves encrypted data interaction to achieve hierarchical collaborative scheduling of the microgrid in the target power system, specifically includes the following steps:
[0058] A distributed computing architecture is constructed, including a central dispatcher and several virtual power plant nodes, each virtual power plant node having several microgrids under it; the central dispatcher is used for global coordination and market dispatch layer calculations, while the virtual power plant nodes are used for calculations at the virtual power plant layer and the microgrid layer; the status data and output data of the microgrid itself are stored locally only.
[0059] A lightweight communication protocol is used for data transmission: each virtual power plant uploads data to the central dispatcher during each iteration of the ADMM algorithm. and After the central dispatcher updates the data at the market dispatch layer, it distributes the variables. To each virtual power plant;
[0060] Each virtual power plant receives variables Then, the data updates for the virtual power plant layer and the microgrid layer are performed in parallel, and the updated data is released after each iteration.
[0061] After the ADMM algorithm converges, it is corrected based on the actual operating conditions of the target power system, and corresponding deviation penalties and ancillary service rewards are settled.
[0062] Repeat the above steps to perform encrypted data interaction and achieve hierarchical collaborative scheduling of the microgrid in the target power system.
[0063] This invention also provides a system for implementing the aforementioned microgrid hierarchical collaborative scheduling method for power systems, comprising a data acquisition module, a model building module, a model solving module, a result correction module, and a collaborative scheduling module; the data acquisition module, model building module, model solving module, result correction module, and collaborative scheduling module are connected in series; the data acquisition module is used to acquire data information of the target power system, preprocess it to obtain an optimized scheduling dataset, and upload the data information to the model building module; the model building module is used to construct a three-layer optimization model of market scheduling layer, virtual power plant layer, and microgrid layer based on the received data information and the obtained data information, and introduces consistency variables and augmented Lagrange multipliers. The daily function couples the model and uploads the data to the model solving module. The model solving module solves the constructed model based on the distributed parallel ADMM algorithm according to the received data to obtain the virtual power plant-microgrid optimized scheduling scheme, and uploads the data to the result correction module. The result correction module corrects the scheme issued by the dispatch center according to the received data and combines it with the actual operation of the target power system, and performs corresponding deviation penalties and ancillary service reward settlements, and uploads the data to the collaborative scheduling module. The collaborative scheduling module performs encrypted data interaction according to the received data to realize the hierarchical collaborative scheduling of the microgrid of the target power system.
[0064] The microgrid hierarchical collaborative scheduling method and system for power systems provided by this invention, through the construction and coupling of a three-layer optimization model of market scheduling layer, virtual power plant layer and microgrid layer, as well as parallel computing solution and deviation penalty and ancillary service reward settlement, not only realizes the hierarchical collaborative scheduling of microgrids in power systems, but also has higher reliability, better accuracy and higher efficiency. Attached Figure Description
[0065] Figure 1 This is a schematic diagram of the method flow of the present invention.
[0066] Figure 2This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation
[0067] like Figure 1 The diagram shown is a flowchart of the method of the present invention: The microgrid hierarchical collaborative scheduling method for power systems disclosed in this invention includes the following steps:
[0068] S1. Obtain data information from the target power system and preprocess it to obtain an optimized scheduling dataset; specifically, this includes the following steps:
[0069] Acquire data information from the target power system; the data information includes:
[0070] The reported quantity and bid data of coal-fired unit i in coal-fired unit set G, the reported quantity and bid data of virtual power plant j in virtual power plant set V, and the microgrid set under virtual power plant j. The data includes the reported load and price quotation data of microgrid k, the load forecast curve of the target power system, and the new energy output forecast curve and upper and lower limit data of coal-fired power unit output of microgrid k under virtual power plant j.
[0071] The acquired data is preprocessed; the preprocessing includes denoising (denoising schemes may include moving average, weighted filtering or wavelet transform, etc.) and correction; the acquired load forecast curve is smoothed.
[0072] S2. Based on the data obtained in step S1, construct a three-layer optimization model consisting of a market dispatch layer, a virtual power plant layer, and a microgrid layer, and introduce consistency variables and augmented Lagrangian functions to couple the model; specifically, the following steps are included:
[0073] Market scheduling layer:
[0074] The market scheduling layer is used for overall economic and safety scheduling; decision variables include the output of each coal-fired unit i at time t. and the output of each virtual power plant j at time t ;
[0075] The following formula is used as the objective function of the market scheduling layer:
[0076] In the formula The objective function of the market scheduling layer; The scheduling cycle is used for the thermal power units. Market quotations for these units typically exhibit a piecewise linear structure. To avoid the complexity of solving non-differentiable functions, this invention smooths the quotation data and forms a continuously differentiable equivalent quotation cost function for coal-fired units by fitting the piecewise quotation curves. Let be the equivalent bid-price cost function for coal-fired unit i, and , The first equivalent fuel cost coefficient, This is the second equivalent fuel cost coefficient. The first equivalent fuel cost coefficient is a function that exhibits the monotonicity of the price quote and the marginal cost structure, thereby reflecting the characteristics of market pricing while ensuring the stability of the model solution. Let be the cost function of the virtual power plant j, and , The adjustable output range of virtual power plant j The total number of consecutive segments. Let be the maximum adjustable output of the virtual power plant j. Let J be the output value of the virtual power plant j at time t in segment s. and , This represents the upper limit of output for segment s. This represents the lower limit of the output of segment s. The unit output cost of virtual power plant j in section s;
[0077] The constraints include system power balance constraints, upper and lower limits of thermal power unit output constraints, thermal power unit output ramping constraints, upper and lower limits of virtual power plant output constraints, and virtual power plant output ramping constraints; among them, the system power balance constraint is expressed as... , Let be the load demand at time t;
[0078] Virtual power plant layer:
[0079] The virtual power plant layer satisfies the output requirements of each virtual power plant j at time t issued by the market dispatch layer. Under the guidance of instructions, optimize the various microgrids under its jurisdiction;
[0080] The following formula is used as the objective function for the virtual power plant layer:
[0081] In the formula The objective function for the virtual power plant layer; Let be the output of the k-th microgrid under the virtual power plant j at time t. It is a cost function, and , The adjustable output range of k microgrids under virtual power plant j The total number of consecutive segments. Let be the maximum adjustable output of the k microgrids under the virtual power plant j. Let be the output value of the k microgrids under the virtual power plant j in segment s at time t. and , This represents the upper limit of output for segment s. This represents the lower limit of the output of segment s. The unit output cost of the k microgrids under the virtual power plant j in section s; This is the deviation penalty coefficient; The upward execution bias slack variable is set. The downward execution bias slack variable is set, and ; This is the output smoothing penalty coefficient; Rewards for virtual power plants to submit unit standby capacity to higher levels; The ancillary service capacity reported by the virtual power plant to the market dispatch layer, wherein the ancillary services include reserve and peak shaving; To measure the penalty coefficient for unfairness caused by uneven internal distribution; Let J be the average output of all microgrids under the virtual power plant j, and , The total number of microgrids under virtual power plant j;
[0082] The constraints include a consistency constraint on aggregate output; the consistency constraint on aggregate output is expressed as follows: ;
[0083] Microgrid layer:
[0084] The microgrid layer meets the output targets set by the virtual power plant. Under the premise of optimization, the microgrid layer performs its own optimization; the decision variables of the microgrid layer include the l-th wind power output of the j-th virtual power plant and the k-th microgrid. The output of the m-th photovoltaic cell The charging power of the s-th energy storage Discharge power of energy storage Micro gas turbine power and start / stop state variables ;
[0085] The following formula is used as the objective function for the microgrid layer:
[0086] In the formula The objective function of the microgrid layer; The cost of penalizing the abandonment of wind and solar power; Energy storage losses and cycle costs; Costs of operating and starting / shutting micro gas turbines; Cost per unit of wind curtailment; This refers to the available power output of wind power. Cost per unit of abandoned light; The photovoltaic power output is available. Cost per unit of energy storage loss; The startup cost of a single gas turbine; The shutdown cost for a single gas turbine;
[0087] The constraints include power balance constraints, energy storage SOC state balance constraints, wind turbine output constraints, photovoltaic output constraints, energy storage output constraints, and gas turbine output constraints;
[0088] The power balance constraint is expressed as:
[0089] In the formula Let be the load demand of the k-th microgrid of virtual power plant j;
[0090] The energy storage SOC state equilibrium constraint is expressed as:
[0091] In the formula Let be the state of charge of the s-th energy storage in the k-th microgrid of virtual power plant j at time t; For charging efficiency; It is discrete time; For discharge efficiency;
[0092] Constructing Consistent Variables and Constraints:
[0093] Variables of the market scheduling layer Introducing global variables Indicates variables for the market scheduling layer. Introducing local variables The system power balance constraint of the market scheduling layer is then expressed as: At the same time, add constraints. ;
[0094] By introducing Lagrange multipliers and a quadratic penalty term, we construct the augmented Lagrange function. , is represented as:
[0095] In the formula In order to be in For vectors , For vectors The objective function of the market scheduling layer at that time; The Lagrange multiplier corresponding to the system power balance constraint; The coefficient of the second-order penalty term;
[0096] The virtual power plant layer and the microgrid layer are merged into a local subproblem of virtual power plant, and each virtual power plant subproblem internally handles the microgrid; the local Lagrangian function of the j-th virtual power plant is constructed. , is represented as:
[0097] In the formula Let be the local cost of the j-th virtual power plant, and ;
[0098] S3. Based on the distributed parallel ADMM algorithm, solve the model constructed in step S2 to obtain the virtual power plant-microgrid optimal scheduling scheme; specifically including the following steps:
[0099] The distributed parallel ADMM algorithm is used to solve the model constructed in step S2:
[0100] A. Initialize variables , and ;
[0101] B. The market scheduling layer uses the ADMM algorithm to update vector z: in the (r+1)th iteration, the value is fixed. and The market scheduling layer solves the following model: st
[0102] C. Distributed parallel local update vectors for each virtual power plant and :fixed and For each virtual power plant, the following local optimization subproblem is solved in parallel: st
[0103] D. Update the Lagrange multipliers: ;
[0104] E. Convergence criterion of ADMM algorithm: when the following conditions are met... Or satisfy When the iteration stops, the iteration continues; where, The step size for updating the dual variable is used to adjust the penalty strength of the consistency constraint. Its value reflects the coordination strength between the system layer and the virtual power plant-microgrid, and can achieve a balance between the algorithm's convergence speed and stability. The required accuracy for the initial residuals. The required accuracy for the dual residuals to be set;
[0105] S4. The dispatch center publishes the plan obtained in step S3, adjusts it based on the actual operating conditions of the target power system, and settles the corresponding deviation penalties and ancillary service rewards; specifically including the following steps:
[0106] a. Results Release: The dispatch center will send the output values of each virtual power plant to the corresponding virtual power plant, and each virtual power plant will then allocate and execute the output tasks of its own microgrid.
[0107] b. Operational calibration: Obtain the actual output value of each microgrid. And calculate the deviation value. for ; This is the sum of the output values of all virtual power plants obtained in step S3;
[0108] like Greater than the set threshold Then the virtual power plant or microgrid will perform its own correction, such as adjusting the charging and discharging power of energy storage or dispatching flexible loads, to achieve local balance; this correction ensures the coordination of system supply and demand and reduces operational risks.
[0109] c. The dispatch center rewards virtual power plants: the incentive reward is calculated using the following formula. :
[0110] In the formula Rewards for ancillary services provided to virtual power plants; A positive deviation penalty coefficient is set.
[0111] d. Distribute the incentive rewards obtained in step c among the various microgrids belonging to the virtual power plant (e.g., using methods such as proportional allocation, marginal cost allocation, or optimized allocation); through this mechanism, a balance between rewards and penalties is achieved to ensure member motivation;
[0112] e. The dispatch center records and stores the results for each microgrid;
[0113] This step, through release, correction, deviation penalty, reward settlement and fair distribution, forms a closed-loop mechanism from optimized scheduling to actual execution, which ensures both system security and economic efficiency and the enthusiasm of participating members.
[0114] S5. Perform encrypted data exchange to achieve hierarchical collaborative scheduling of the microgrid in the target power system; specifically including the following steps:
[0115] A distributed computing architecture is constructed, including a central dispatcher and several virtual power plant nodes, each virtual power plant node having several microgrids under it. The central dispatcher is used for global coordination and market dispatch layer calculations, while the virtual power plant nodes are used for calculations at the virtual power plant layer and the microgrid layer. The status data and output data of each microgrid are stored locally only, including renewable energy output, energy storage charging and discharging power, and unit operating status. Through the distributed architecture, detailed data from each microgrid can be locally aggregated into virtual power plant aggregated output, and only this aggregated value is transmitted to the central dispatcher, thus protecting the privacy of the microgrids.
[0116] To reduce communication overhead and ensure real-time performance, a lightweight communication protocol is used for data transmission: each virtual power plant uploads data to the central dispatcher during each iteration of the ADMM algorithm. and After the central dispatcher updates the data at the market dispatch layer, it distributes the variables. This data transmission method, which distributes data to each virtual power plant, ensures that local data from each microgrid is not directly transmitted to the central node. Instead, it participates in global optimization only through aggregated information, thus protecting privacy.
[0117] Each virtual power plant receives variables Then, the data of the virtual power plant layer and the microgrid layer are updated in parallel, and the updated data is published after each iteration. This step can process a large amount of microgrid data at the same time, improve computing efficiency, and upload new aggregate variables after each iteration to realize a closed loop of prediction-optimization-publishing.
[0118] After the ADMM algorithm converges, it is corrected based on the actual operation of the target power system, and corresponding deviation penalties and ancillary service rewards are settled. By transmitting only aggregate and dual variables, the system protects the sensitive operating data of each microgrid, and realizes a complete closed-loop process from prediction to optimization to release to settlement, ensuring the system's economy, robustness and microgrid privacy protection.
[0119] Repeat the above steps to perform encrypted data interaction and achieve hierarchical collaborative scheduling of the microgrid in the target power system.
[0120] The present invention proposes a three-layer optimized scheduling framework consisting of a market scheduling layer (upper layer), a virtual power plant layer (middle layer), and a microgrid layer (lower layer). For the first time, it consistently couples the aggregated output of virtual power plants with the system load clearing demand, enabling the dispatch center to schedule virtual power plants as efficiently as traditional coal-fired units, achieving the goal of balancing macro-level security and micro-level flexibility.
[0121] This invention addresses the challenges of large-scale microgrids, diverse constraints, and high model complexity by introducing a distributed parallel ADMM solution method. Each virtual power plant independently and in parallel solves local optimization subproblems, while the central dispatcher updates upper-level decision variables and dual variables in batches to achieve global coordination. This method significantly improves computational efficiency, supports concurrent dispatching of large-scale microgrids, and is suitable for real-time or near-real-time power dispatching scenarios.
[0122] The method of this invention introduces flexible constraints such as execution deviation penalty, wind and solar curtailment penalty, energy storage loss and ancillary service reward into the optimization model, which can improve the microgrid response capability while ensuring that the system output matches the load demand, and incentivize the microgrid to actively participate in dispatch through the settlement mechanism;
[0123] This invention uses a lightweight communication protocol to transmit only the aggregate and dual variables of the virtual power plant, enabling the exchange of scheduling information without uploading detailed output, energy storage status, or unit operation data within the microgrid. This effectively protects the operational privacy of the microgrid and enhances the security of user data.
[0124] This invention constructs a closed-loop process from load forecasting and wind and solar power output forecasting to distributed parallel optimization calculation, and then to scheduling release and deviation settlement. It realizes dynamic feedback of forecast information, optimization results and actual operation data, and provides guarantees for the system's economy, robustness and fairness.
[0125] like Figure 2The diagram shows the functional modules of the system of this invention: The system disclosed in this invention for implementing the microgrid hierarchical collaborative scheduling method for power systems includes a data acquisition module, a model building module, a model solving module, a result correction module, and a collaborative scheduling module; the data acquisition module, model building module, model solving module, result correction module, and collaborative scheduling module are connected in series; the data acquisition module is used to acquire data information of the target power system, preprocess it to obtain an optimized scheduling dataset, and upload the data information to the model building module; the model building module is used to construct a three-layer optimization model of market scheduling layer, virtual power plant layer, and microgrid layer based on the received data information and the obtained data information, and introduce consistency. The variables are coupled with the augmented Lagrangian function in the model, and the data information is uploaded to the model solving module. The model solving module solves the constructed model based on the received data information using the distributed parallel ADMM algorithm to obtain the virtual power plant-microgrid optimized scheduling scheme, and uploads the data information to the result correction module. The result correction module corrects the scheme issued by the dispatch center based on the received data information and combines it with the actual operation of the target power system, and performs corresponding deviation penalties and ancillary service reward settlements, and uploads the data information to the collaborative scheduling module. The collaborative scheduling module performs encrypted data interaction based on the received data information to realize hierarchical collaborative scheduling of the microgrid of the target power system.
Claims
1. A hierarchical collaborative scheduling method for microgrids in power systems, comprising the following steps: S1. Acquire data information of the target power system and preprocess it to obtain an optimized scheduling dataset; S2. Based on the data obtained in step S1, construct a three-layer optimization model consisting of a market dispatch layer, a virtual power plant layer, and a microgrid layer, and introduce consistency variables and augmented Lagrangian functions to couple the model. S3. Based on the distributed parallel ADMM algorithm, the model constructed in step S2 is solved to obtain the virtual power plant-microgrid optimized scheduling scheme; S4. The dispatch center publishes the plan obtained in step S3, makes corrections based on the actual operation of the target power system, and settles the corresponding deviation penalties and ancillary service rewards. S5. Perform encrypted data interaction to achieve hierarchical collaborative scheduling of the microgrid in the target power system.
2. The microgrid hierarchical collaborative dispatch method for power systems according to claim 1, characterized in that... Step S1, which involves acquiring data information from the target power system and preprocessing it to obtain an optimized scheduling dataset, specifically includes the following steps: Acquire data information from the target power system; the data information includes: The reported quantity and bid data of coal-fired unit i in coal-fired unit set G, the reported quantity and bid data of virtual power plant j in virtual power plant set V, and the microgrid set under virtual power plant j. The data includes the reported load and price quotation data of microgrid k, the load forecast curve of the target power system, and the new energy output forecast curve and upper and lower limit data of coal-fired power unit output of microgrid k under virtual power plant j. The acquired data is preprocessed, including noise reduction and correction; the acquired load forecast curve is smoothed.
3. The microgrid hierarchical collaborative dispatch method for power systems according to claim 2, characterized in that... Step S2 involves constructing a three-layer optimization model—comprising a market dispatch layer, a virtual power plant layer, and a microgrid layer—based on the data obtained in step S1, and introducing consistency variables and augmented Lagrangian functions for model coupling. Specifically, this includes the following steps: Market scheduling layer: The market scheduling layer is used for overall economic and safety scheduling; decision variables include the output of each coal-fired unit i at time t. and the output of each virtual power plant j at time t ; The following formula is used as the objective function of the market scheduling layer: In the formula The objective function of the market scheduling layer; The scheduling period; Let be the equivalent bid-price cost function for coal-fired unit i, and , The first equivalent fuel cost coefficient, This is the second equivalent fuel cost coefficient. The first equivalent fuel cost coefficient; Let be the cost function of the virtual power plant j, and , The adjustable output range of virtual power plant j The total number of consecutive segments. Let be the maximum adjustable output of the virtual power plant j. Let J be the output value of the virtual power plant j at time t in segment s. and , This represents the upper limit of output for segment s. This represents the lower limit of the output of segment s. The unit output cost of virtual power plant j in section s; The constraints include system power balance constraints, upper and lower limits of thermal power unit output constraints, thermal power unit output ramping constraints, upper and lower limits of virtual power plant output constraints, and virtual power plant output ramping constraints; among them, the system power balance constraint is expressed as... , Let be the load demand at time t; Virtual power plant layer: The virtual power plant layer satisfies the output requirements of each virtual power plant j at time t issued by the market dispatch layer. Under the guidance of instructions, optimize the various microgrids under its jurisdiction; The following formula is used as the objective function for the virtual power plant layer: In the formula The objective function for the virtual power plant layer; Let be the output of the k-th microgrid under the virtual power plant j at time t. It is a cost function, and , The adjustable output range of k microgrids under virtual power plant j The total number of consecutive segments. Let be the maximum adjustable output of the k microgrids under the virtual power plant j. Let be the output value of the k microgrids under the virtual power plant j in segment s at time t. and , This represents the upper limit of output for segment s. This represents the lower limit of the output of segment s. The unit output cost of the k microgrids under the virtual power plant j in section s; This is the deviation penalty coefficient; The upward execution bias slack variable is set. The downward execution bias slack variable is set, and ; This is the output smoothing penalty coefficient; Rewards for virtual power plants to submit unit standby capacity to higher levels; The ancillary service capacity reported by the virtual power plant to the market dispatch layer, wherein the ancillary services include reserve and peak shaving; To measure the penalty coefficient for unfairness caused by uneven internal distribution; Let J be the average output of all microgrids under the virtual power plant j, and , The total number of microgrids under virtual power plant j; The constraints include a consistency constraint on aggregate output; the consistency constraint on aggregate output is expressed as follows: ; Microgrid layer: The microgrid layer meets the output targets set by the virtual power plant. Under the premise of optimization, the microgrid layer performs its own optimization; the decision variables of the microgrid layer include the l-th wind power output of the j-th virtual power plant and the k-th microgrid. The output of the m-th photovoltaic cell The charging power of the s-th energy storage Discharge power of energy storage Micro gas turbine power and start / stop state variables ; The following formula is used as the objective function for the microgrid layer: In the formula The objective function of the microgrid layer; The cost of penalizing the abandonment of wind and solar power; Energy storage losses and cycle costs; Costs of operating and starting / shutting micro gas turbines; Cost per unit of wind curtailment; This refers to the available power output of wind power. Cost per unit of abandoned light; The photovoltaic power output is available. Cost per unit of energy storage loss; The startup cost of a single gas turbine; The shutdown cost for a single gas turbine; The constraints include power balance constraints, energy storage SOC state balance constraints, wind turbine output constraints, photovoltaic output constraints, energy storage output constraints, and gas turbine output constraints; The power balance constraint is expressed as: In the formula Let be the load demand of the k-th microgrid of virtual power plant j; The energy storage SOC state equilibrium constraint is expressed as: In the formula Let be the state of charge of the s-th energy storage in the k-th microgrid of virtual power plant j at time t; For charging efficiency; It is discrete time; For discharge efficiency; Constructing Consistent Variables and Constraints: Variables of the market scheduling layer Introducing global variables Indicates variables for the market scheduling layer. Introducing local variables The system power balance constraint of the market scheduling layer is then expressed as: At the same time, add constraints. ; By introducing Lagrange multipliers and a quadratic penalty term, we construct the augmented Lagrange function. , is represented as: In the formula In order to be in For vectors , For vectors The objective function of the market scheduling layer at that time; The Lagrange multiplier corresponding to the system power balance constraint; The coefficient of the second-order penalty term; The virtual power plant layer and the microgrid layer are merged into a local subproblem of virtual power plant, and each virtual power plant subproblem internally handles the microgrid; the local Lagrangian function of the j-th virtual power plant is constructed. , is represented as: In the formula Let be the local cost of the j-th virtual power plant, and .
4. The microgrid hierarchical collaborative dispatch method for power systems according to claim 3, characterized in that... Step S3, based on the distributed parallel ADMM algorithm, solves the model constructed in step S2 to obtain the virtual power plant-microgrid optimized scheduling scheme, specifically including the following steps: The distributed parallel ADMM algorithm is used to solve the model constructed in step S2: A. Initialize variables , and ; B. The market scheduling layer uses the ADMM algorithm to update vector z: in the (r+1)th iteration, the value is fixed. and The market scheduling layer solves the following model: s.t. C. Distributed parallel local update vectors for each virtual power plant and :fixed and For each virtual power plant, the following local optimization subproblem is solved in parallel: s.t. D. Update the Lagrange multipliers: ; E. Convergence criterion of ADMM algorithm: when the following conditions are met Or satisfy When the iteration stops, the iteration continues; where, The step size for updating the dual variable. The required accuracy for the initial residuals. The accuracy requirements that the dual residuals need to meet.
5. The microgrid hierarchical collaborative dispatch method for power systems according to claim 4, characterized in that... Step S4 involves the dispatch center publishing the scheme obtained in step S3, making corrections based on the actual operating conditions of the target power system, and settling corresponding deviation penalties and ancillary service rewards. Specifically, this includes the following steps: a. Results Release: The dispatch center will send the output values of each virtual power plant to the corresponding virtual power plant, and each virtual power plant will then allocate and execute the output tasks of its own microgrid. b. Operational calibration: Obtain the actual output value of each microgrid. And calculate the deviation value. for ; This is the sum of the output values of all virtual power plants obtained in step S3; like Greater than the set threshold Then the virtual power plant or microgrid will perform its own correction; c. The dispatch center rewards virtual power plants: the incentive reward is calculated using the following formula. : In the formula Rewards for ancillary services provided to virtual power plants; A positive deviation penalty coefficient is set. d. Distribute the incentive rewards obtained in step c among the various microgrids to which the virtual power plant belongs; e. The dispatch center records and stores the results of each microgrid.
6. The microgrid hierarchical collaborative dispatch method for power systems according to claim 5, characterized in that... Step S5, which involves encrypted data interaction to achieve hierarchical collaborative scheduling of the microgrid in the target power system, specifically includes the following steps: A distributed computing architecture is constructed, including a central dispatcher and several virtual power plant nodes, each virtual power plant node having several microgrids under it; the central dispatcher is used for global coordination and market dispatch layer calculations, while the virtual power plant nodes are used for calculations at the virtual power plant layer and the microgrid layer; the status data and output data of the microgrid itself are stored locally only. A lightweight communication protocol is used for data transmission: each virtual power plant uploads data to the central dispatcher during each iteration of the ADMM algorithm. and After the central dispatcher updates the data at the market dispatch layer, it distributes the variables. To each virtual power plant; Each virtual power plant receives variables Then, the data updates for the virtual power plant layer and the microgrid layer are performed in parallel, and the updated data is released after each iteration. After the ADMM algorithm converges, it is corrected based on the actual operating conditions of the target power system, and corresponding deviation penalties and ancillary service rewards are settled. Repeat the above steps to perform encrypted data interaction and achieve hierarchical collaborative scheduling of the microgrid in the target power system.
7. A system for implementing the microgrid hierarchical collaborative scheduling method for power systems according to any one of claims 1 to 6, characterized in that... The system comprises a data acquisition module, a model building module, a model solving module, a result correction module, and a collaborative scheduling module, which are connected in series. The data acquisition module acquires data from the target power system, preprocesses it to obtain an optimized scheduling dataset, and uploads the data to the model building module. The model building module constructs a three-layer optimization model (market scheduling layer, virtual power plant layer, and microgrid layer) based on the received data, introduces consistency variables and augmented Lagrangian functions for model coupling, and uploads the data to the model solving module. The model solving module solves the constructed model using the distributed parallel ADMM algorithm based on the received data to obtain the virtual power plant-microgrid optimized scheduling scheme, and uploads the data to the result correction module. The result correction module corrects the scheme obtained from the dispatch center based on the received data and the actual operating conditions of the target power system, performs corresponding deviation penalties and ancillary service reward settlements, and uploads the data to the collaborative scheduling module. The collaborative scheduling module is used to perform encrypted data interaction based on the received data information to realize hierarchical collaborative scheduling of the microgrid of the target power system.