Real-time rolling scheduling method based on day-ahead scheduling and integrated primary and secondary microgrid

Abstract

The application discloses a kind of based on day-ahead scheduling main and micro integrated real-time rolling scheduling method, including the following steps: according to output historical data to construct output prediction model, utilizes output prediction model to determine the day-ahead baseline plan of main and micro system day-ahead scheduling P DA ( t );Build target function including predicted correction and operation risk, optimize target function to obtain optimal plan value in rolling window, calculate the difference vector of optimal plan and day-ahead baseline plan P DA ( t ) to obtain correction power vector, embed optimal plan in rolling window into full-day baseline plan to obtain updated scheduling plan;Through cloud-edge collaborative system, the scheduling plan of each rolling period is issued Instruction, realize the real-time rolling scheduling of main and micro system.The application can effectively improve the continuity and stability of scheduling scheme, reduce real-time operation risk, realize the adaptive coordination control of main grid, distribution network and microgrid level, improve system robustness and new energy consumption capacity.

Description

Technical Field

[0001] This invention relates to power system operation and intelligent dispatching technology, specifically to a real-time rolling dispatching method based on day-ahead dispatching that integrates primary, secondary, and micro-level systems. Background Technology

[0002] With the advancement of "dual carbon" targets and the gradual depletion of fossil fuels, developing a high proportion of renewable energy has become an inevitable trend in the global energy transition. Distributed photovoltaic and wind power, among other clean energy sources, are widely integrated into distribution networks and microgrids due to their low-carbon and high-efficiency characteristics. However, their output is significantly affected by weather and the environment, exhibiting substantial volatility and uncertainty, posing challenges to the stability and economic efficiency of real-time power system operation. Simultaneously, with the maturity of energy storage technology, enhanced flexible load response capabilities, and the implementation of time-of-use pricing mechanisms, the dispatch and control mode of main and distribution microgrids is evolving from the traditional centralized approach to a multi-level collaborative and dynamic optimization model.

[0003] Existing scheduling methods mostly focus on static optimization during the day-ahead phase, neglecting the dynamic impact of renewable energy output and load forecasting errors during operation. This leads to deviations from the plan in actual operation and makes it difficult to identify risks in advance. In addition, the main grid and distribution network scheduling often operate independently, resulting in poor information transmission and significant response delays. This makes it difficult to achieve rolling corrections and hierarchical linkage control, and fails to meet the real-time security requirements under high-proportion distributed resource access. Summary of the Invention

[0004] The technical problem this invention aims to solve is to provide a real-time rolling scheduling method based on day-ahead scheduling that integrates main grid, distribution grid, and microgrid. Building upon day-ahead scheduling results, a sliding time window and a real-time rolling optimization mechanism are introduced to construct a real-time optimization model with the dual objectives of minimizing prediction error correction and minimizing overall operational risk. This effectively improves the continuity and stability of scheduling schemes, reduces real-time operational risk, achieves adaptive coordinated control at the main grid, distribution grid, and microgrid levels, enhances system robustness and renewable energy absorption capacity, and demonstrates good real-time performance and promising engineering application prospects.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A real-time rolling scheduling method based on day-ahead scheduling and integrating master, dispatch, and micro scheduling includes the following steps: A power output prediction model is constructed based on historical power output data. This model is then used for day-ahead scheduling to determine the day-ahead baseline plan for the main distribution microsystems. P DA ( t ); Construct an objective function that includes the predicted correction amount and operational risk, optimize the objective function, and obtain the optimal plan value within the current rolling window. PG ( t ), P tie ( t ), Pch ES ( t ), Pdis ES ( t ), Δ P DR ( t Correcting the power vector It is the optimal plan and the current baseline plan. P DA ( t The updated scheduling plan is obtained based on the difference vector of the power vector and the corrected power vector. Updated scheduling plan P new ( t This is a global plan object that embeds the optimal plan within the rolling window into the daily baseline plan, and issues the scheduling plan for each rolling period. The instructions enable real-time rolling scheduling of the main and auxiliary microsystems.

[0006] Furthermore, the day-ahead scheduling of the main and distribution microsystems is determined using the output prediction model. P DA ( t Specifically, the power output baseline is generated using a Monte Carlo simulation model based on a power output prediction model. This involves the following steps: Monte Carlo simulation was used to randomly generate a set of output sampling scenarios for the output prediction model; Calculate the geometric distance between each pair of scenes s and s′ in the scene set; Select the scene d that minimizes the sum of the probabilities of the remaining scenes; Replace scene d with scene r that is geometrically closest to scene d in the scene set, add the probability of scene d to scene r, eliminate scene d, and form a new scene set; Determine if the number of remaining scenes in the new scene set meets the requirement; if not, repeat the step of calculating the geometric distance between each pair of scenes s and s′ in the scene set until the number of remaining scenes in the new scene set meets the requirement. A collection of typical scenarios recently; A day-ahead scheduling optimization model is constructed based on a set of typical day-ahead scenarios. The day-ahead scheduling optimization model takes the dual objectives of minimizing network loss and minimizing comprehensive risk as its optimization objectives. The constraints include power balance, tie-line switching power constraints, energy storage SOC constraints, upper and lower limits of unit / new energy output and ramping constraints, distribution network voltage / line power flow constraints, and interruptible load constraints. Solve the day-ahead scheduling optimization model to obtain the day-ahead initial scheduling plan, which is then used as the baseline input for real-time rolling scheduling.

[0007] Furthermore, the mathematical expression of the objective function is as follows:

[0008] in: The objective function for predicting corrections is expressed mathematically as follows:

[0009] In the above formula, It is the real-time corrected power of the load. It is the real-time corrected power of the i-th new energy source; It is the objective function for running the risk, and its mathematical expression is as follows:

[0010] In the above formula, Let be the voltage of the i-th node. For reference voltage, Let j be the current in the j-th branch. The upper limit of the allowable current for the branch. , Risk weighting factor; , The dynamic weights are expressed mathematically as follows:

[0011] in, As a risk indicator, As the warning threshold, k This is the sensitivity coefficient (gain coefficient) for dynamic weight adjustment, used to adjust the dynamic weight. Follow - The speed of response to changes.

[0012] Furthermore, the mathematical expression for the real-time corrected power is as follows:

[0013] in, This represents the real-time corrected power of output i. The day-ahead baseline plan for output i, This represents the prediction error rate of output i. To correct the gain coefficient; If the output is a load, the mathematical expression for the prediction error rate is as follows:

[0014] in, Let be the predicted load power at time t. The measured load power at time t; If the power output is from new energy sources, the mathematical expression for the prediction error rate is as follows:

[0015] in, N G The number of new energy units. For the first i The predicted value of time t for Taiwan New Energy. For the first i The measured value of time t of the new energy vehicle.

[0016] Furthermore, before optimizing the objective function in each rolling cycle, the process also includes: obtaining the measured and predicted output values ​​at the beginning of each rolling cycle, and calculating risk indicators based on the measured and predicted output values. And the prediction error rate of each output; The scheduling plan for each rolling cycle is issued through the cloud-edge collaborative system. Before the instructions, it also includes: If risk indicators Exceeding the warning threshold Under the premise of meeting the upper limit of energy storage power and state of charge constraints, the energy storage will be switched to discharge operation and the corresponding discharge power command will be executed; if the energy storage discharge alone is still insufficient... If the load falls back below the threshold, flexible load adjustment will be further activated, and interruptible loads will be reduced according to the load reduction amount obtained from rolling optimization.

[0017] Furthermore, when optimizing the objective function in each rolling cycle, the constraints include: The power balance constraint is mathematically expressed as follows:

[0018] In the above formula, To provide power to the main network units. For distribution network interconnection power, For energy storage charging power, For energy storage discharge power, For load power, For a moment t Active power loss in the system network; The energy storage state of charge constraint is mathematically expressed as follows:

[0019] In the above formula t For rolling scheduling t In each time period, Δ T The duration of a single scheduling period / sampling interval. E ES ( t For energy storage devices during time periods t The energy state at the initial moment, E ES ( t +1) represents the energy storage device during the time period. t The energy state at the end (i.e., the start of the next period). Pch ES ( t (This refers to the energy storage period) t The charging active power, Pdis ES ( t (This refers to the energy storage period) t The active power of discharge, η ch Charging efficiency represents the effective conversion ratio of a unit of charging power on the energy side. η dis For efficient discharge, due to losses between the energy side and the power side, the energy reduction during discharge is... ; The mathematical expression for flexible load constraints is as follows:

[0020] In the above formula, Interruptible load response quantity ; The mathematical expression for node voltage constraints is as follows:

[0021] In the above formula, Let be the voltage at time t of the i-th node; The risk threshold constraint is expressed mathematically as follows:

[0022] in, As a risk indicator, This is the warning threshold.

[0023] definition for:

[0024]

[0025]

[0026] In the above formula,t For rolling scheduling discrete time period / time index, , The set of nodes and the number of nodes in the network under consideration; i For node indexing, , M The set of branches and the number of branches in the network under consideration; j For branch indexes. V i ( t ) period t node i voltage amplitude, Vmin i , Vmax i For nodes i Lower and upper limits of permissible voltage. Δ V i for Vmax i Vmin i , I j ( t (Time period) t branch road j Current amplitude. Imax j branch road j Maximum allowable current. α , β It serves as a risk weighting factor, used to balance the relative importance of "voltage risk" and "current risk".

[0027] Furthermore, when optimizing the objective function in each rolling cycle, specifically in each rolling cycle, the scheduling schemes in different scheduling cycles are treated as individual particles in the population, and the multi-objective particle swarm optimization algorithm is used to optimize the objective function to obtain the Pareto optimal solution.

[0028] Furthermore, it also includes: after each rolling cycle, collecting operational feedback data from the main and auxiliary microsystems, and updating the parameters and weight coefficients of the prediction model based on the operational feedback data, specifically including: The actual quantity collected within this rolling cycle is recorded as... The predicted value given by the prediction model in the previous cycle is denoted as Calculate the prediction error:

[0029] And with a scrolling window The mean square error within is the correction target:

[0030] in To predict model parameters, online gradient correction is used for parameter updates:

[0031] in The learning rate / step size controls the magnitude of parameter updates; the updated... Used for output prediction in the next rolling cycle; Calculate voltage risk components based on feedback data With current risk component In the scrolling window Internal statistical mean The weights are then smoothly updated based on the proportion of each risk source:

[0032] in For smoothing coefficients, To avoid constants with a denominator of 0, and ; According to the updated Calculate the comprehensive risk index:

[0033] And Substituting into the dynamic weight mapping formula, we obtain the weights of the objective function in the next cycle:

[0034] in As the warning threshold, This is the sensitivity coefficient. , These are the dynamic weights for the predicted correction amount and the operational risk, respectively.

[0035] Furthermore, the scheduling plan for each rolling cycle is issued. When issuing instructions, the scheduling plan is specifically distributed through the cloud-edge collaborative system. The instructions are executed globally by the main network, and locally by the distribution network and microgrid.

[0036] Compared with the prior art, the advantages of the present invention are as follows: This invention implements a rolling correction mechanism based on day-ahead planning to improve the continuity and stability of scheduling schemes; introduces a sliding time window and error correction penalty mechanism to enhance robustness to new energy fluctuations and load disturbances; balances risk constraints and flexible resource response to achieve main-distribution-micro coordination optimization and real-time risk mitigation; and forms an adaptive real-time scheduling closed loop to improve system operation safety and new energy absorption capacity, thus having good engineering application value. Attached Figure Description

[0037] Figure 1 This is a simplified flowchart of the method according to an embodiment of the present invention.

[0038] Figure 2 This is a detailed flowchart of the method according to an embodiment of the present invention.

[0039] Figure 3 This is a topology diagram of the master-supplier-micro cooperative scheduling structure in an embodiment of the present invention.

[0040] Figure 4 This is a detailed flowchart of the objective function optimization solution in an embodiment of the present invention. Detailed Implementation

[0041] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0042] This embodiment proposes a real-time rolling scheduling method integrating the main grid, distribution network, and microgrid based on day-ahead scheduling. First, the initial power plan for the main grid, distribution network, and microgrid system is determined through day-ahead scheduling. Second, a sliding time window and correction model are introduced in the real-time phase to achieve dynamic power coordination and risk constraint control among the main grid, distribution network, and microgrid. Under the condition of fully considering system power balance and operational safety, a real-time rolling scheduling model integrating the main grid, distribution network, and microgrid is established with the dual objectives of "minimizing the amount of prediction correction" and "minimizing operational risk." In each rolling cycle, the day-ahead plan is corrected based on the latest prediction information and measured operational data, generating coordinated control commands. This method can effectively improve the continuity and stability of the scheduling scheme, reduce real-time operational risks, achieve adaptive coordinated control at the main grid, distribution network, and microgrid levels, enhance system robustness and renewable energy absorption capacity, and has good real-time performance and engineering application prospects.

[0043] like Figure 1 As shown, the method includes the following steps: S101) Construct an output prediction model based on historical output data, and use the output prediction model to determine the day-ahead baseline plan for the main distribution microsystems through day-ahead scheduling. P DA ( t ); S102) Construct an objective function that includes predicted corrections and operational risks, solve the objective function to obtain the optimal planned value within the current rolling window, including the main grid unit output. P G ( t Distribution network interconnection power P tie ( t Energy storage charging power Pch ES ( t ), energy storage and discharge power Pdis ES ( t Interruptible load response Δ P DR (t ), calculate the optimal plan and the current-day baseline plan P DA ( t The difference vector yields the corrected power vector. The optimal plan within the rolling window is embedded into the all-day baseline plan. Based on the modified power vector, operations such as physical constraint verification, command trigger threshold judgment, and transition curve generation are performed. The resulting global plan object serves as the updated scheduling plan. P new ( t ); Issue the scheduling plan for each rolling cycle. The instructions enable real-time rolling scheduling of the main and auxiliary microsystems.

[0044] The following is in conjunction with the appendix Figure 2 Each step will be explained in detail.

[0045] Step S101 of this embodiment includes the following steps: S1) Establish day-ahead scheduling baselines and rolling optimization scenarios. Specifically, during the day-ahead phase, historical data is used to construct output prediction models for photovoltaic, wind power, and load. These models are then used to generate typical scenarios via Monte Carlo simulation as day-ahead power output baselines, providing a reference for real-time scheduling. The output prediction models are then used to determine the day-ahead baseline plan for the main and distribution microsystems. P DA ( t When ), the following steps are included: S11) The output prediction model uses Monte Carlo simulation to randomly generate a photovoltaic and load output sampling scenario set S; S12) Calculate the geometric distance between each pair of scenes s and s′ in the scene set S; S13) Select the scene d that minimizes the sum of the probabilities of the remaining scenes; S14) Replace scene d with scene r, which is the closest to scene d in geometric distance from scene d in scene set S, add the probability of scene d to scene r, eliminate scene d, and form a new scene set S′; S15) Determine whether the number of remaining scenes in the new scene set meets the requirements; if not, execute step S12 again to calculate the geometric distance between each pair of scenes s and s′ in the scene set until the number of remaining scenes in the new scene set meets the requirements, thereby forming the current typical scene set; S16) Construct a day-ahead scheduling optimization model based on a set of typical day-ahead scenarios: with the dual objectives of minimizing network loss and minimizing comprehensive risk as the optimization objectives, the constraints include power balance, tie-line switching power constraints, energy storage SOC constraints, upper and lower limits of unit / new energy output and ramping constraints, distribution network voltage / line power flow constraints, interruptible load constraints, etc. S17) Solve the day-ahead scheduling optimization model to obtain the day-ahead initial scheduling plan for the primary and secondary microsystems. P DA ( t ), and serve as the baseline input for real-time rolling scheduling.

[0046] Step S102 in this embodiment includes the following steps: S2) Construct a source-load prediction error model, and calculate the new energy output prediction error rate and load prediction error rate respectively, including: S21) Establish a load forecasting error model: definition To predict load power, To measure the actual load power, then at time... t The load forecasting error rate is: (1) S22) Establish a new energy prediction error model: definition N G The number of new energy units. For the first i Taiwan New Energy's forecast value, For the first i The measured values ​​of Taiwan New Energy are constantly... t The error rate for new energy prediction is: (2) S23) Calculate the real-time corrected power: definition To contribute to i's current plan, The prediction error rate of output i, To correct the gain coefficient, the real-time corrected power of output i is: (3) If the output is a load, then the prediction error rate is... Calculated using formula (1), if the power output is from a new energy source, the prediction error rate is... It is calculated using formula (2).

[0047] S3) Construct a dual-objective real-time rolling optimization model, with the optimization objectives of minimizing source load prediction error and minimizing system operation risk, including: S31) Construct a dual-objective real-time rolling optimization model: definition , These are dynamic weighting coefficients; definition ; Define the objective function as power: (4) in, It is the objective function for predicting corrections. It is the objective function for running risk; S32) Construct the objective function for the prediction correction: (5) In the above formula, It is the real-time corrected power of the load. It is the real-time corrected power of the i-th new energy source; S33) Construct the operational risk objective function; definition Let be the voltage of the i-th node; definition Reference voltage; definition Let j be the current in the j-th branch; definition The maximum allowable current for the branch; definition , Risk weighting factor; Define the objective function for running the risk: (6) S34) Adjust weights dynamically based on risk indicators; (7) in, As a risk indicator, As the warning threshold, k This is the sensitivity coefficient (gain coefficient) for dynamic weight adjustment, used to adjust the dynamic weight. Follow - The speed of response to changes.

[0048] S4) Define decision variables and constraints, including: S41: The main decision variables for building the model; definition To provide power to the main network units; definition For distribution network interconnection power; definition Energy storage charging power ; definition Energy storage discharge power ; definition Interruptible load response quantity ; S42) The constraints include the following five aspects: The power balance constraint is mathematically expressed as follows: (8) In the above formula, To provide power to the main network units. For distribution network interconnection power, For energy storage charging power, For energy storage discharge power, For load power, For a moment t Active power loss in the system network; The energy storage state of charge constraint is mathematically expressed as follows: (9) In the above formula t For rolling scheduling t In each time period, Δ T The duration of a single scheduling period / sampling interval. E ES ( t For energy storage devices during time periods t The energy state at the initial moment, E ES ( t +1) represents the energy storage device during the time period. t The energy state at the end (i.e., the start of the next period). Pch ES ( t (This refers to the energy storage period) t The charging active power, Pdis ES ( t (This refers to the energy storage period) t The active power of discharge, η ch Charging efficiency represents the effective conversion ratio of a unit of charging power on the energy side. η dis For efficient discharge, due to losses between the energy side and the power side, the energy reduction during discharge is... ; The mathematical expression for flexible load constraints is as follows: (10) In the above formula, Interruptible load response quantity ; The mathematical expression for node voltage constraints is as follows: (11) In the above formula, Let be the voltage at time t of the i-th node; The risk threshold constraint is expressed mathematically as follows: (12) in, As a risk indicator, This is the warning threshold.

[0049] definition for: (13) (14) (15) In the above formula, t For rolling scheduling discrete time period / time index, , The set of nodes and the number of nodes in the network under consideration; i For node indexing, , M The set of branches and the number of branches in the network under consideration; j For branch indexes. V i ( t ) period t node i voltage amplitude, Vmin i , Vmax i For nodes i Lower and upper limits of permissible voltage. Δ V i for Vmax i Vmin i , I j ( t (Time period) t branch road j Current amplitude. Imax j branch road j Maximum allowable current. α , β It serves as a risk weighting factor, used to balance the relative importance of "voltage risk" and "current risk".

[0050] S5) Introducing energy storage devices and interruptible load response mechanisms forms a flexible adjustment means for operational constraints, thereby addressing risk indicators. Exceeding the warning threshold When the energy storage rapid response and flexible load adjustment strategy is triggered, the risk is reduced by discharging the energy storage or reducing the load. The flexible load adjustment strategy is implemented through an interruptible load response mechanism. S6) The proposed model is optimized using the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm, such as... Figure 3 As shown, the steps are as follows: S61: Data initialization: Input the composition and structural parameters of each unit model, model parameters, MOPSO algorithm parameters, etc.; at the same time, initialize the particle population, and each particle in the population corresponds to a scheduling scheme within a scheduling cycle. S62: Input individual particles as system variables into the simulation model to correct variables that violate constraints; S63: Using individual fitness as input to the optimization model, the offspring population is obtained through equation (13); (16) S64: Determine the individual extreme value pbest: Take pbest as the initial individual extreme value of the particle. If the current particle dominates pbest, then take the current particle as the individual extreme value pbest. If the two cannot be compared, calculate the number of other particles that each particle dominates in the group. If the particle dominates more, then take the one with more dominance as the individual extreme value pbest. S65: Rank the population into strata and sort the populations, and select the optimal non-dominated solutions. Pareto Store in an external archive collection and clear non-existent files. Pareto Solve the problem and determine whether the external archive set exceeds the specified capacity. If so, select m particles according to the crowding distance. S66: Saved using an external archive set Pareto Optimal solution; S67: Return to step S63 until the termination condition is met.

[0051] S7) Collect operational feedback data, correct prediction model parameters and optimize weights in real time, and form an adaptive rolling scheduling closed loop.

[0052] Based on the above sub-steps, a complete rolling optimization process is obtained as follows: S201) Data Initialization: Obtain the output plan for the day-ahead scheduling phase. P DA ( t This serves as the initial baseline for real-time rolling scheduling, and acquires energy storage status, prediction model parameters, and risk thresholds. A rolling period Δ is also set. T and window length N; (S202) At the beginning of each rolling cycle, real-time load, photovoltaic and wind power data are collected, and corresponding forecast data are obtained through a prediction model. Risk indicators are then calculated based on the measured output and the predicted output. And the prediction error rate of each output; S203) Following the method in step S6, solve the optimization objective function based on formulas (4) to (16) to obtain the optimal plan value within the current scrolling window, and take the value. P G (t ), P tie ( t ), Pch ES ( t ), Pdis ES ( t ), Δ P DR ( t Correcting the power vector It is the optimal plan and the current baseline plan. P DA ( t The difference vector of ) and the updated scheduling plan P new ( t This is a global plan object formed by embedding the optimal plan within the scrolling window into the all-day baseline plan; S204 Risk Indicators Exceeding the warning threshold When this occurs, a fast response control process is triggered: First, the energy storage fast response is invoked, meaning that, provided the energy storage power limit and state of charge constraints are met, the energy storage is switched from charging / standby to discharging operation and the corresponding discharge power command is executed; if discharging the energy storage alone is insufficient... If the load falls back below the threshold, flexible load adjustment will be further activated, and interruptible loads will be reduced according to the load reduction amount obtained by rolling optimization. The above control commands are sent to the energy storage and load control terminal for execution through cloud-edge collaboration to achieve risk suppression.

[0053] S205) Distributes the scheduling plan for each rolling cycle through the cloud-edge collaborative system. P new ( t Instructions such as Figure 4 As shown, the cloud is based on P new ( t The system generates global control commands on the main grid side and local adjustment commands on the distribution network and microgrid side. The main grid side commands include reference values ​​for the output of controllable units on the main grid and power setting values ​​for the tie lines (or exchanges) with the distribution network / microgrid. These are used to maintain global power balance at the main grid level and to uniformly control the boundary power across levels. After receiving the corresponding commands, the edge control systems on the distribution network and microgrid side, under the premise of meeting the power targets and related constraints of the tie lines given by the main grid, combine local real-time measurement information to perform charging and discharging, reduction / restoration and other adjustment operations on controllable resources such as energy storage devices and flexible loads in their respective jurisdictions. This achieves local power balance, voltage / power flow constraint satisfaction and deviation elimination in the distribution network and microgrid. The execution results are then transmitted back to the cloud for status updates and re-optimization in the next rolling cycle, thereby realizing "the main grid performs global control, and the distribution network and microgrid perform local adjustment".

[0054] (S206) After each rolling cycle, the operation feedback data of the main distribution microsystem is collected (including the actual output / consumption of photovoltaic, wind power, and load, as well as node voltage, branch current, energy storage SOC, flexible load execution, etc.), and the online updating of prediction model parameters and weighting coefficients is realized based on this data, specifically including: 1) Update prediction model parameters: Record the actual quantities collected within this rolling period as... (Corresponding to photovoltaic / wind power / load, etc.), the predicted value given by the prediction model in the previous cycle is denoted as... Calculate the prediction error

[0055] And with a scrolling window The mean square error within is the correction target:

[0056] in These are the parameters for the prediction model. Parameter updates can be performed using online gradient correction.

[0057] in The learning rate / step size controls the magnitude of parameter updates; the updated... Used for output prediction in the next rolling cycle.

[0058] 2) Update risk weights Calculate voltage risk component based on feedback data With current risk component (Its calculation method is consistent with the definition of risk indicators), in the scrolling window Internal statistical mean The weights are then smoothly updated based on the proportion of each risk source:

[0059] in For smoothing coefficients, To avoid constants with a denominator of 0, and .

[0060] 3) Update dynamic weights According to the updated Calculate the comprehensive risk index

[0061] And Substituting the dynamic weight mapping formula given in the document, we obtain the weights for the next cycle:

[0062] in As the warning threshold, This is the sensitivity coefficient. This achieves a rolling closed loop of "feedback data → risk assessment → weight update → further optimization in the next cycle".

[0063] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A real-time rolling scheduling method based on day-ahead scheduling integrating master-distribution-micro scheduling, characterized in that, Includes the following steps: A power output prediction model is constructed based on historical power output data. This model is then used for day-ahead scheduling to determine the day-ahead baseline plan for the main distribution microsystems. P DA ( t ); Construct an objective function that includes forecast adjustments and operational risks, optimize the objective function to obtain the optimal plan value within the rolling window, and calculate the optimal plan and the day-ahead baseline plan. P DA ( t The difference vector yields the corrected power vector. The optimal plan within the rolling window is embedded into the global plan object formed by the all-day baseline plan, which is then used as the updated scheduling plan. Distribute the scheduling plan for each rolling cycle. The instructions enable real-time rolling scheduling of the main and auxiliary microsystems.

2. The real-time rolling scheduling method based on day-ahead scheduling and integrated master-distributor-micro scheduling according to claim 1, characterized in that, Day-ahead scheduling using output prediction models to determine the day-ahead baseline plan for primary and secondary microsystems. P DA ( t Specifically, the power output baseline is generated using a Monte Carlo simulation model based on a power output prediction model. This involves the following steps: Monte Carlo simulation was used to randomly generate a set of output sampling scenarios for the output prediction model; Calculate the geometric distance between each pair of scenes s and s′ in the scene set; Select the scene d that minimizes the sum of the probabilities of the remaining scenes; Replace scene d with scene r that is geometrically closest to scene d in the scene set, add the probability of scene d to scene r, eliminate scene d, and form a new scene set; Determine if the number of remaining scenes in the new scene set meets the requirement; if not, repeat the step of calculating the geometric distance between each pair of scenes s and s′ in the scene set until the number of remaining scenes in the new scene set meets the requirement, thus forming the current typical scene set; A day-ahead scheduling optimization model is constructed based on a set of typical day-ahead scenarios. The day-ahead scheduling optimization model takes the dual objectives of minimizing network loss and minimizing comprehensive risk as its optimization objectives. The constraints include power balance, tie-line switching power constraints, energy storage SOC constraints, upper and lower limits of unit / new energy output and ramping constraints, distribution network voltage / line power flow constraints, and interruptible load constraints. Solve the day-ahead scheduling optimization model to obtain the day-ahead initial scheduling plan, which is then used as the baseline input for real-time rolling scheduling.

3. The real-time rolling scheduling method based on day-ahead scheduling and integrated master-distributor-micro scheduling according to claim 1, characterized in that, The mathematical expression of the objective function is as follows: in: The objective function for predicting corrections is expressed mathematically as follows: In the above formula, It is the real-time corrected power of the load. It is the real-time corrected power of the i-th new energy source; It is the objective function for running the risk, and its mathematical expression is as follows: In the above formula, Let be the voltage of the i-th node. For reference voltage, Let j be the current in the j-th branch. The upper limit of the allowable current for the branch. , Risk weighting factor; , The dynamic weights are expressed mathematically as follows: in, As a risk indicator, As the warning threshold, k This is the sensitivity coefficient for dynamic weight adjustment, used to adjust the dynamic weights. Follow - The speed of response to changes.

4. The real-time rolling scheduling method based on day-ahead scheduling and integrated master-distributor-micro scheduling according to claim 3, characterized in that, The mathematical expression for the real-time correction power is as follows: in, This represents the real-time corrected power of output i. The day-ahead baseline plan for output i, This represents the prediction error rate of output i. To correct the gain coefficient; If the output is a load, the mathematical expression for the prediction error rate is as follows: in, Let be the predicted load power at time t. Let be the measured load power at time t; If the power output is from new energy sources, the mathematical expression for the prediction error rate is as follows: in, N G The number of new energy units. For the first i The predicted value of time t for Taiwan New Energy. For the first i The measured value of time t of the new energy vehicle.

5. The real-time rolling scheduling method based on day-ahead scheduling and integrated master-distributor-micro scheduling according to claim 4, characterized in that, Before optimizing the objective function in each rolling cycle, the process also includes: obtaining the measured and predicted output values ​​at the beginning of each rolling cycle, and calculating risk indicators based on the measured and predicted output values. And the prediction error rate of each output; The scheduling plan for each rolling cycle is issued through the cloud-edge collaborative system. Before the instructions, it also includes: If risk indicators Exceeding the warning threshold Under the premise of meeting the upper limit of energy storage power and state of charge constraints, the energy storage will be switched to discharge operation and the corresponding discharge power command will be executed; if the energy storage discharge alone is still insufficient... If the load falls back below the threshold, flexible load adjustment will be further activated, and interruptible loads will be reduced according to the load reduction amount obtained from rolling optimization.

6. The real-time rolling scheduling method based on day-ahead scheduling and integrated master-distributor-micro scheduling according to claim 1, characterized in that, When optimizing the objective function in each rolling cycle, the constraints include: The power balance constraint is mathematically expressed as follows: In the above formula, To provide power to the main network units. For distribution network interconnection power, For energy storage charging power, For energy storage discharge power, For load power, For a moment t Active power loss in the system network; The energy storage state of charge constraint is mathematically expressed as follows: In the above formula t For rolling scheduling t In each time period, Δ T The duration of a single scheduling period / sampling interval. E ES ( t For energy storage devices during time periods t The energy state at the initial moment, E ES ( t +1) represents the energy storage device during the time period. t Energy state at the end; Pch ES ( t (This refers to the energy storage period) t The charging active power, Pdis ES ( t (This refers to the energy storage period) t The active power of discharge, η ch Charging efficiency represents the effective conversion ratio of a unit of charging power on the energy side. η dis For efficient discharge, due to losses between the energy side and the power side, the energy reduction during discharge is... ; The mathematical expression for flexible load constraints is as follows: In the above formula, Interruptible load response quantity ; The node voltage constraint is expressed mathematically as follows: In the above formula, For the first i Each node time t The voltage; The risk threshold constraint is expressed mathematically as follows: in, As a risk indicator, This is the warning threshold.

7. The real-time rolling scheduling method based on day-ahead scheduling and integrated master-distributor-micro scheduling according to claim 6, characterized in that, The mathematical expression for the risk indicator is as follows: In the above formula, t For rolling scheduling discrete time period / time index, , The set of nodes and the number of nodes in the network under consideration; i For node indexing, , M This represents the set of branches and the number of branches in the network under consideration. j For branch index; V i ( t ( ) represents a time period t node i voltage amplitude, Vmin i , Vmax i For nodes i Lower and upper limits of permissible voltage; Δ V i = Vmax i Vmin i , I j ( t (Time period) t branch road j Current amplitude; Imax j branch road j Maximum allowable current; α , β This is a risk weighting factor.

8. The integrated real-time rolling scheduling method for master, distribution, and microcomputer systems according to claim 1, characterized in that, When optimizing the objective function in each rolling cycle, specifically in each rolling cycle, the scheduling schemes in different scheduling cycles are treated as individual particles in the population, and the multi-objective particle swarm optimization algorithm is used to optimize the objective function to obtain the Pareto optimal solution.

9. The integrated real-time rolling scheduling method for master, distribution, and microcomputer systems according to claim 1, characterized in that, Also includes: At the end of each rolling cycle, operational feedback data of the main and auxiliary microsystems is collected. The parameters and weight coefficients of the prediction model are updated based on the operational feedback data, specifically including: The actual quantity collected within this rolling cycle is recorded as... The predicted value given by the prediction model in the previous cycle is denoted as Calculate the prediction error: And with scroll window The mean square error within is the correction target: in To predict model parameters, online gradient correction is used for parameter updates: in The learning rate / step size controls the magnitude of parameter updates; the updated... Used for output prediction in the next rolling cycle; Calculate voltage risk components based on feedback data With current risk component In the scrolling window Internal statistical mean The weights are then smoothly updated based on the proportion of risk sources. in For smoothing coefficients, To avoid constants with a denominator of 0, and ; According to the updated Calculate the comprehensive risk index: And Substituting into the dynamic weight mapping formula, we obtain the weights of the objective function in the next cycle: in As the warning threshold, This is the sensitivity coefficient. , These are the dynamic weights for the predicted correction amount and the operational risk, respectively.

10. The integrated real-time rolling scheduling method for master, slave, and microcomputer systems according to claim 1, characterized in that, Distribute the scheduling plan for each rolling cycle When issuing instructions, the scheduling plan is specifically distributed through the cloud-edge collaborative system. The instructions are executed globally by the main network, and locally by the distribution network and microgrid.

Images