Resource aggregation scheduling method considering time-space characteristic uncertainty analysis and flexibility
By constructing a flexible resource aggregation and scheduling method that considers the uncertainties of spatiotemporal characteristics, and using the Copula model and Minkowski sum for photovoltaic power output prediction and resource aggregation, the spatiotemporal correlation of distributed photovoltaic power in the power system is resolved, thereby improving the scheduling accuracy and flexibility of the power system.
Patent Information
- Application Number
- CN202511662235.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-02-10
AI Technical Summary
Existing technologies lack flexible resource aggregation schemes and resource scheduling models, and fail to effectively address the spatiotemporal correlation effects of distributed photovoltaics, resulting in high uncertainty in power system operation and insufficient regulation capacity of traditional units.
By constructing a flexible resource aggregation scheduling method that considers spatiotemporal uncertainty analysis, photovoltaic output is decoupled and modeled using the Copula transfer kernel-CSMC model and the hybrid Copula model. Combined with Minkowski algorithm for flexible resource aggregation, a scheduling model is constructed to optimize energy system scheduling.
It enables accurate prediction of distributed photovoltaic power output and effective aggregation of flexible resources, improving the dispatch accuracy and flexibility of the power system and reducing operational uncertainty.
Smart Images
Figure CN121507967A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to the technical field of energy configuration, and particularly relates to a method for analyzing spatial and temporal characteristics uncertainty and aggregating and scheduling flexible resources. BACKGROUND
[0002] With the acceleration of the global energy structure transformation process, the power system is facing the dual challenges of a sharp increase in power supply pressure and a continuous rise in new energy penetration. According to statistics of the International Energy Agency (IEA), the global distributed photovoltaic newly installed capacity reached 268 GW in 2022, with a year-on-year growth of 55%. The fluctuation characteristics lead to a significant "duck curve" shape of the net load curve of the power grid. This uncertainty brings a severe test to the safe operation of the power system, and the traditional unit regulation capacity has been difficult to cope with the minute-level power fluctuation, so it is urgent to build a new prediction and scheduling system. SUMMARY
[0003] The application aims at the deficiency that the existing technology lacks a model combining the flexible resource aggregation scheme and the resource scheduling, and proposes a method for analyzing spatial and temporal characteristics uncertainty and aggregating and scheduling flexible resources. After the flexible resource aggregation based on Minkowski sum, the flexible resource aggregation model and the predicted photovoltaic are used to optimize the scheduling of the energy system.
[0004] The application is implemented by the following technical scheme:
[0005] The application relates to a method for analyzing spatial and temporal characteristics uncertainty and aggregating and scheduling flexible resources. The model parameters of the energy system and the historical data of the photovoltaic are collected, the photovoltaic output prediction considering the spatial and temporal correlation is performed, the flexible resource aggregation model is constructed, the Minkowski sum is used to sum the multivariate array to form an overall flexible resource feasible region, and after the flexible resource aggregation, the scheduling model is constructed. BRIEF DESCRIPTION OF DRAWINGS
[0006] Figure 1 The flowchart of the application is shown in the figure;
[0007] Figure 2 The flowchart of the photovoltaic output scene generation of the embodiment is shown in the figure;
[0008] Figure 3 The schematic diagram of the energy cluster structure of the embodiment is shown in the figure. DETAILED DESCRIPTION
[0009] As shown in the figure, the application relates to a method for analyzing spatial and temporal characteristics uncertainty and aggregating and scheduling flexible resources, which comprises the following steps: Figure 1
[0010] Step 1, collecting the model parameters of the energy system;
[0011] Step 2: Photovoltaic output prediction considering spatiotemporal correlation: The Copula transfer kernel-CSMC model and the hybrid Copula model are used to decouple the temporal and spatial correlations, respectively. These correlations are then coupled during sampling to ensure the realism of the generated scenario. Finally, the photovoltaic output is predicted, specifically including:
[0012] 2.1 Constructing a time-related model, specifically including:
[0013] For any dimension N, there exists a Copula that satisfies the requirement that i has a joint probability density of N, specifically: ,in: Representing variables The marginal cumulative probability distribution function; taking the derivative with respect to both sides yields: ,in: For a multi-dimensional joint probability density function, for Marginal probability density function, Let Copula be the joint probability density function, obtained by finding... , , The marginal cumulative probability distribution function, and can be expressed by the marginal cumulative probability density function alone. , , .
[0014] ii. In the multidimensional time distribution of photovoltaics, adjacent moments can be treated as a binary joint distribution. Therefore, the discretized Markov state transition characteristics can be analyzed. For a first-order Markov chain, the Copula function is used to replace its state transition matrix, forming a continuous state-space Markov chain model (CSMC). Its advantage lies in maintaining high modeling accuracy without increasing complexity. Specifically: ,in: and They are respectively The values at times t+1 and t. For the continuous variable being referred to, Let be the conditional probability density function, representing that in under conditions Probability.
[0015] iii. The expression for the Copula function is derived mathematically: ,in: They represent The marginal cumulative probability distribution function at times t+1 and t , .
[0016] 2.2 Constructing a Spatial Correlation Model: A hybrid Copula function is used to describe distributed photovoltaic (PV) power output scenarios and predict PV output. The hybrid Copula function, by adjusting the weights of different components, can simulate a richer and more diverse correlation structure, thus better adapting to the statistical characteristics of actual data. Specifically: ,in: The marginal distribution of the random variable; These are the parameters related to the Copula function; The weight parameters are set. The D-Vine Copula function is used for decomposition, breaking it down from a multivariate distribution function into multiple binary distribution functions; ,in: Let represent the conditional Copula probability density function between the j-th and j+i-th variables, given that all variables from the (j+1)-th to the (j+i-1)-th variables are known.
[0017] 2.3 Scenarios for generating spatiotemporal correlations of photovoltaics, specifically including:
[0018] i. Initialize the cumulative probability distribution function matrix of the photovoltaic spatiotemporal correlation scenario;
[0019] ii. At time t, select an N-dimensional vector of photovoltaic random sampling from the spatial correlation D-Vine Copula;
[0020] iii. Select the Copula model for time correlation under different seasons;
[0021] iv. For any photovoltaic unit, the edge cumulative distribution function at time t-1 is used as a condition, and the spatial correlation sample value is used as a conditional probability distribution. These are substituted into the Copula function correlation model to obtain the edge cumulative distribution function at time t.
[0022] v repeats step ii-iv at time t+1;
[0023] vi transforms the marginal cumulative distribution function matrix into an inverse function through probability integral inverse transformation to obtain the photovoltaic scenario.
[0024] Step 3: Construct a flexible resource aggregation model: Analyze the scheduling potential of distributed energy storage, electric vehicles, and temperature-controlled loads by performing flexibility modeling and aggregation. This includes:
[0025] 3.1 Electric Vehicle Model: , , , , , ,in: Let i be the energy of electric vehicle i at time t; and The charging and discharging efficiencies of electric vehicle i respectively; The charging power of electric vehicle i at time t; The charging power of electric vehicle i at time t; and These refer to the entry and exit times of the electric vehicle; , These are the upper and lower limits of the discharge power of electric vehicle i; , The upper and lower limits of charging power for electric vehicle i; , Let represent the upper and lower limits of the battery capacity of electric vehicle i at time t.
[0026] The electric vehicle model is transformed into a unified form as follows: ,in: Let i be the constraint space for electric vehicle i; , Let be the coefficient matrix and vector of electric vehicle i, respectively; Let be the decision variables for electric vehicle i.
[0027] 3.2 Distributed Energy Storage Model: Due to its excellent regulation capabilities, distributed energy storage can be flexibly adjusted in four dimensions, specifically: ,in: Let i be the energy storage discharge power of distributed energy storage at time t. Let i be the energy storage charging power of distributed energy storage at time t. The rated power of distributed energy storage i; Let the state variable of distributed energy storage i take the values {0,1}. The charging and discharging efficiency of distributed energy storage i. , Let i be the upper and lower capacity limits of distributed energy storage. Let be the initial capacity of distributed energy storage i; the upper and lower limits of the capacity as a function of time are: ,in: , Distributed energy storage i in time period Upper and lower limits on content volume.
[0028] The distributed energy storage model is transformed into a unified form as follows: ,in: Let i be the constrained space of distributed energy storage. , These are the coefficient matrix and vector of distributed energy storage i, respectively; Let i be the decision variable for distributed energy storage.
[0029] 3.3 Temperature-Controlled Load Model: The operating mechanism of temperature-controlled loads is greatly influenced by external factors such as air temperature, building structure, and heat source. However, due to its certain heat storage characteristics, it is considered as a quasi-energy storage model. This invention uses a first-order ordinary differential equation to represent the operating characteristics of the cold storage air conditioner, specifically: ,in: The indoor temperature at time t; R is the outdoor temperature; C is the building thermal resistance; and C is the specific heat of the air. The thermal effect coefficient of air conditioning The on / off state of the air conditioner at time t is 1 for on and 0 for off;
[0030] In constructing temperature-controlled load models, the indoor temperature benchmark setting serves as a crucial reference for describing temperature-controlled load energy storage models. , , ,in: and These represent the power and energy of the temperature-controlled load i at time t, respectively. , These represent the temperature-controlled load i during the time period. Upper and lower limits of internal power; , These represent the temperature-controlled load i during the time period. Upper and lower limits of content volume.
[0031] The temperature-controlled load model is transformed into a unified form as follows: ,in: The constraint space for temperature-controlled load i; , These are the coefficient matrix and vector of the temperature-controlled load i, respectively; Let i be the decision variable for the temperature-controlled load.
[0032] Step 4: Use Minkowski summation to sum the multivariate arrays to form a holistic flexible resource feasible region, thus achieving flexible resource aggregation. This specifically includes:
[0033] 4.1 The model constructed in step 3 unifies the various flexibility resources as follows: ,in: Let i be the constraint space for the i-th flexible resource of the j-th type of flexible resource; , These are the coefficient matrix and vector for the i-th resource of the j-th type of flexibility, respectively. Let be the decision variable for the i-th type of flexibility resource.
[0034] 4.2 The constraint intervals for various flexible resources form a polyhedron constrained by linear inequality equations. To represent multiple flexible resources using a single polyhedron, this invention employs an averaging method to set a benchmark. , ,in: The mean of the flexibility resource parameters for the j-th type; Let j represent the number of flexible resources of type j.
[0035] 4.3 By performing affine transformations, translations, and scaling on the reference set, an internal approximation solution model for individual flexible resources is constructed. ,in: Let i be the scaling factor for the j-th type of flexibility resource. For the j-th type of flexibility resource, the i-th translation coefficient is used to aggregate the flexibility resource cluster. Diverse and flexible resource clusters .
[0036] Step 5, construct as follows Figure 3 The scheduling model shown: Distributed photovoltaic, energy storage, gas turbine, electric vehicle cluster, distributed energy storage cluster, and temperature-controlled load cluster are selected for scheduling analysis, specifically including:
[0037] 5.1 Gas turbine model: , , ,in: The power output of the gas turbine at time t; For gas turbine conversion efficiency; The power of natural gas consumed by the gas turbine at time t; , These are the upper and lower limits of gas turbine power; , These are the upper and lower limits of the gas turbine's ramp power;
[0038] 5.2 Source-side large-scale energy storage model: ,in: In order to be in The storage capacity of the time storage device; To store its own loss rate; These refer to the energy storage charging and discharging efficiency, respectively. They are respectively in Storing and discharging power at all times; They are respectively in The state variables of the time storage device during charging and discharging take values of: ; These are the minimum and maximum values of the energy storage capacity, respectively. These are the upper limits for storing and charging / discharging energy, respectively.
[0039] 5.3 Distributed Photovoltaic Model ,in: Let be the photovoltaic power used at time t; Let be the predicted photovoltaic power at time t.
[0040] 5.4 Model Constraints: , ,in: Let be the interactive electric power at time t; This represents the maximum electrical power delivered. The power limit for purchasing natural gas, ,in: , , These are the dispatch power for temperature-controlled loads, electric vehicles, and energy storage clusters, respectively. Let t be the electrical load of the energy system at time t.
[0041] 5.5 Model Objective Function: Considering the system scheduling operation cost and the flexibility resource scheduling cost, the objective function is to minimize the total cost, specifically: , , , ,in: For the maintenance costs of the energy system; For energy system interaction costs; To reduce the cost of flexible resource scheduling, and The unit maintenance costs for photovoltaic and energy storage are respectively. and These are the electricity purchase price and the gas purchase price, , , These are the dispatch prices for temperature-controlled loads, electric vehicles, and energy storage clusters, respectively.
[0042] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.
Claims
1. A method for resource aggregation and scheduling considering spatiotemporal uncertainty analysis and flexibility, characterized in that, By collecting energy system model parameters and historical photovoltaic data, a flexible resource aggregation model is constructed after predicting photovoltaic output considering spatiotemporal correlation. Then, the Minkowski sum is used to sum the multivariate array to form a global flexible resource feasible region. After realizing flexible resource aggregation, a scheduling model is constructed.
2. The method for considering spatiotemporal uncertainty analysis and flexible resource aggregation scheduling according to claim 1, characterized in that, The aforementioned photovoltaic power output prediction considering spatiotemporal correlation refers to: using the Copula transfer kernel-CSMC model and the hybrid Copula model to decouple and model the temporal and spatial correlations respectively, then coupling them during sampling to make the generated scenario realistic, and finally predicting the photovoltaic power output, specifically including: 2.1 Constructing a time-related model, specifically including: For any dimension N, there exists a Copula that satisfies the requirement that i has a joint probability density of N, specifically: ,in: Representing variables The marginal cumulative probability distribution function; taking the derivative with respect to both sides yields: , For a multi-dimensional joint probability density function, for Marginal probability density function, Let Copula be the joint probability density function, obtained by finding... , , The marginal cumulative probability distribution function, and can be expressed by the marginal cumulative probability density function alone. , , ; ii. In the multidimensional time distribution of photovoltaics, adjacent moments can be treated as a binary joint distribution. Therefore, the discretized Markov state transition characteristics can be analyzed. For a first-order Markov chain, the Copula function is used to replace its state transition matrix, forming a continuous state-space Markov chain model (CSMC), specifically: ,in: and They are respectively The values at times t+1 and t, For the continuous variable being referred to, Let be the conditional probability density function, representing that in under conditions probability; iii. The expression for the Copula function is derived mathematically: ,in: They represent The marginal cumulative probability distribution function at times t+1 and t , ; 2.2 Constructing a Spatial Correlation Model: A hybrid Copula function is used to describe distributed photovoltaic (PV) power output scenarios and predict PV output. By adjusting the weights of different components, the hybrid Copula function simulates a richer and more diverse correlation structure, thus better adapting to the statistical characteristics of actual data. Specifically: ,in: The marginal distribution of the random variable; These are the parameters related to the Copula function; The weight parameters are decomposed using the D-VineCopula function, which decomposes the multivariate distribution function into multiple binary distribution functions. , Let represent the conditional Copula probability density function between the j-th and j+i-th variables, given that all variables from the (j+1)-th to the (j+i-1)-th variables are known. 2.3 Scenarios for generating spatiotemporal correlations of photovoltaics.
3. The method for considering spatiotemporal uncertainty analysis and flexible resource aggregation scheduling according to claim 2, characterized in that, Step 2.3 specifically includes: i. Initialize the cumulative probability distribution function matrix of the photovoltaic spatiotemporal correlation scenario; ii. At time t, select the N-dimensional vector of photovoltaic random sampling from the spatial correlation D-VineCopula; iii. Select the Copula model for time correlation under different seasons; iv. For any photovoltaic unit, the edge cumulative distribution function at time t-1 is used as a condition, and the spatial correlation sample value is used as a conditional probability distribution. These are substituted into the Copula function correlation model to obtain the edge cumulative distribution function at time t. v repeats step ii-iv at time t+1; vi transforms the marginal cumulative distribution function matrix into an inverse function through probability integral inverse transformation to obtain the photovoltaic scenario.
4. The method for considering spatiotemporal uncertainty analysis and flexible resource aggregation scheduling according to claim 1, characterized in that, The aforementioned flexible resource aggregation model is obtained through the following method: 3.1 Electric Vehicle Model: , , , , , ,in: Let i be the energy of electric vehicle i at time t; and The charging and discharging efficiencies of electric vehicle i respectively; The charging power of electric vehicle i at time t; The charging power of electric vehicle i at time t; and These refer to the entry and exit times of the electric vehicle; , These are the upper and lower limits of the discharge power of electric vehicle i; , The upper and lower limits of charging power for electric vehicle i; , Let be the upper and lower limits of the battery capacity of electric vehicle i at time t; The electric vehicle model is transformed into a unified form as follows: ,in: Let i be the constraint space for electric vehicle i; , Let be the coefficient matrix and vector of electric vehicle i, respectively; Let i be the decision variable for electric vehicle i; 3.2 Distributed Energy Storage Model: Due to its excellent regulation capabilities, distributed energy storage can be flexibly adjusted in four dimensions, specifically: ,in: Let i be the energy storage discharge power of distributed energy storage at time t. Let i be the energy storage charging power of distributed energy storage at time t. The rated power of distributed energy storage i; Let the state variable of distributed energy storage i take the values {0,1}. The charging and discharging efficiency of distributed energy storage i. , Let i be the upper and lower capacity limits of distributed energy storage. Let be the initial capacity of distributed energy storage i; the upper and lower limits of the capacity as a function of time are: , , Distributed energy storage i in time period Content volume upper and lower limits; The distributed energy storage model is transformed into a unified form as follows: ,in: Let i be the constrained space of distributed energy storage. , These are the coefficient matrix and vector of distributed energy storage i, respectively; Let i be the decision variable for distributed energy storage. 3.3 Temperature-Controlled Load Model: The operating mechanism of temperature-controlled loads is greatly influenced by external factors such as air temperature, building structure, and heat source. However, due to its certain heat storage characteristics, it is considered as a quasi-energy storage model. This invention uses a first-order ordinary differential equation to represent the operating characteristics of the cold storage air conditioner, specifically: ,in: The indoor temperature at time t; R is the outdoor temperature; C is the building thermal resistance; and C is the specific heat of the air. The thermal effect coefficient of air conditioning The on / off state of the air conditioner at time t is 1 for on and 0 for off; In constructing temperature-controlled load models, the indoor temperature benchmark setting serves as a crucial reference for describing temperature-controlled load energy storage models. , , ,in: and These represent the power and energy of the temperature-controlled load i at time t, respectively. , These represent the temperature-controlled load i during the time period. Upper and lower limits of internal power; , These represent the temperature-controlled load i during the time period. Upper and lower limits of content volume; The temperature-controlled load model is transformed into a unified form as follows: ,in: The constraint space for temperature-controlled load i; , These are the coefficient matrix and vector of the temperature-controlled load i, respectively; Let i be the decision variable for the temperature-controlled load.
5. The method for considering spatiotemporal uncertainty analysis and flexible resource aggregation scheduling according to claim 1, characterized in that, The overall flexibility resource feasible domain is obtained in the following way: 4.1 The flexible resource aggregation model unifies all flexible resources into: ,in: Let i be the constraint space for the j-th type of flexibility resource and the ith flexibility resource; , These are the coefficient matrix and vector for the i-th resource of the j-th type of flexibility, respectively. Let be the decision variable for the i-th resource of the j-th type of flexibility; 4.2 The constraint intervals for various flexible resources form a polyhedron constrained by linear inequality equations. To represent multiple flexible resources using a single polyhedron, this invention employs an averaging method to set a benchmark. , ,in: The mean of the flexibility resource parameters for the j-th type; Let j be the number of flexible resources of type j; 4.3 By performing affine transformations, translations, and scaling on the reference set, an internal approximation solution model for individual flexible resources is constructed. ,in: Let i be the scaling factor for the j-th type of flexibility resource. For the j-th type of flexibility resource, the i-th translation coefficient is used to aggregate the flexibility resource cluster. Diverse and flexible resource clusters .
6. The method for considering spatiotemporal uncertainty analysis and flexible resource aggregation scheduling according to claim 1, characterized in that, The aforementioned scheduling model specifically includes: 5.1 Gas turbine model: , , ,in: The power output of the gas turbine at time t; For gas turbine conversion efficiency; The power of natural gas consumed by the gas turbine at time t; , These are the upper and lower limits of gas turbine power; , These are the upper and lower limits of the gas turbine's ramp power; 5.2 Source-side large-scale energy storage model: ,in: In order to be in The storage capacity of the time storage device; To store its own loss rate; These refer to the energy storage charging and discharging efficiency, respectively. They are respectively in Storing and discharging power at all times; They are respectively in The state variables of the time storage device during charging and discharging take values of: ; These are the minimum and maximum values of the energy storage capacity, respectively. These are the upper limits for energy storage charging and discharging, respectively. 5.3 Distributed Photovoltaic Model ,in: Let be the photovoltaic power used at time t; The photovoltaic power is predicted at time t; 5.4 Model Constraints: , ,in: Let be the interactive electric power at time t; This represents the maximum electrical power delivered. The power limit for purchasing natural gas, ,in: , , These are the dispatch power for temperature-controlled loads, electric vehicles, and energy storage clusters, respectively. Let be the electrical load of the energy system at time t; 5.5 Model Objective Function: Considering the system scheduling operation cost and the flexibility resource scheduling cost, the objective function is to minimize the total cost, specifically: , , , ,in: For the maintenance costs of the energy system; For energy system interaction costs; To reduce the cost of flexible resource scheduling, and The unit maintenance costs for photovoltaic and energy storage are respectively. and These are the electricity purchase price and the gas purchase price, , , These are the dispatch prices for temperature-controlled loads, electric vehicles, and energy storage clusters, respectively.