Power system double-layer optimization planning method, device, equipment and medium

Abstract

The invention discloses a power system double-layer optimization planning method, device, equipment and medium, and relates to the field of power grid collaborative planning, and the method comprises the steps: constructing a power side operation model and a load side operation model according to the historical operation data of a to-be-planned power system, scene clustering is carried out on renewable energy output data in the historical operation data, and a weather-driven scene set is constructed; according to the power side operation model and the load side operation model, an upper-layer planning model is constructed by taking the equipment full life cycle cost minimization as a target, and an upper-layer planning constraint is determined based on a planning scale; according to the weather-driven scene set, taking periodic operation cost minimization as a target, constructing a lower-layer scheduling model, and determining a lower-layer scheduling constraint based on a scheduling scale; and combining the upper-layer planning model and the lower-layer scheduling model into a double-layer optimization model, performing iterative solution based on a Benders decomposition algorithm to obtain a planning scheme, and performing planning optimization on the power system according to the planning scheme.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power grid collaborative planning, and in particular to a power system double-layer optimization planning method, device, equipment and medium. BACKGROUND

[0002] New energy power generation represented by wind power, photovoltaic and other renewable energy sources has gradually been widely used in power systems due to its flexible controllability, economic and environmental protection characteristics to address global issues such as climate change and environmental protection. However, the inherent intermittency, randomness and volatility of renewable energy sources have brought security and stability impacts to the power system dominated by traditional energy sources. Under the environment of gradual application of renewable energy sources, how to reasonably plan the power system to improve the security and stability of the power system operation is one of the main research topics.

[0003] The existing planning for the power system mainly uses a single-layer planning model to solve the planning scheme, which is insufficient and does not verify the scheme, resulting in inaccurate planning schemes and low practical feasibility. At the same time, the existing technology usually only studies the influence of new energy operation on the power system under a single time scale, which cannot reflect the different influences of new energy operation under multiple time scales, resulting in planning schemes that cannot fully reflect the dynamic characteristics of new energy and are not accurate and reliable in actual application. Therefore, how to reasonably plan the power system to fully reflect the characteristics of new energy and improve the accuracy of the planning scheme is still a technical problem to be solved in the existing technology. SUMMARY

[0004] The present application provides a power system double-layer optimization planning method, device, equipment and medium to solve the technical problem that the existing planning for the power system is not accurate.

[0005] According to a first aspect of the embodiments of the present application, a power system double-layer optimization planning method is provided, comprising: According to the historical operation data of the power system to be planned, a power source side operation model and a load side operation model are constructed, and renewable energy output data in the historical operation data is clustered to construct a weather-driven scenario set; According to the power source side operation model and the load side operation model, an upper-layer planning model is constructed with the objective of minimizing the device full life cycle cost, and upper-layer planning constraints of the upper-layer planning model are determined based on the planning scale; According to the weather-driven scenario set, a lower-layer scheduling model is constructed with the objective of minimizing the periodic operation cost of the power system, and lower-layer scheduling constraints of the lower-layer scheduling model are determined based on the scheduling scale; The upper-layer planning model and the lower-layer scheduling model are combined into a bi-level optimization model, the bi-level optimization model is iteratively solved based on a Benders decomposition algorithm, a planning scheme of the power system is obtained, and the power system is planned and optimized according to the planning scheme.

[0006] According to historical operation data of a power system to be planned, a power source side operation model, a load side operation model and a weather-driven scenario set are constructed, an upper-layer planning model and a lower-layer scheduling model are constructed respectively with minimization of a device full life cycle cost as a target and minimization of a periodic operation cost as a target, then a bi-level optimization model is combined and iteratively solved based on a Benders decomposition algorithm, compared with existing planning and solving by using a single-layer planning model, the bi-level planning and solving model can fully consider different aspects of optimization in the power system, thereby improving the accuracy of planning and solving, that is, improving the accuracy of planning of the power system; meanwhile, the upper-layer planning model determines upper-layer planning constraints based on a planning scale, the lower-layer scheduling model determines lower-layer scheduling constraints based on a scheduling scale, influences in different time scales of planning and scheduling are respectively studied, and the lower-layer scheduling model is constructed by using a weather-driven scenario set obtained by clustering of renewable energy output data, which can fully reflect dynamic characteristics of new energy, thereby improving the accuracy of solving and planning, that is, improving the accuracy of planning of the power system.

[0007] In some embodiments of the present application, the historical operation data includes renewable energy output data and traditional energy output data; and the power source side operation model and the load side operation model are constructed according to the historical operation data of the power system to be planned, specifically including: Renewable energy expected output data is fitted according to the renewable energy output data, and renewable energy output model is obtained by scene disturbance correction; The power source side operation model is constructed according to the renewable energy output model and a traditional energy output model; wherein the traditional energy output model is constructed based on the traditional energy output data; A plurality of power consumption behavior types are determined according to the historical operation data, and each power consumption behavior type is modeled to obtain a corresponding power demand model; wherein the power consumption behavior types include industrial power consumption, interruptible load and transferable load; The load side operation model is constructed according to the power demand model corresponding to each power consumption behavior type.

[0008] The renewable energy output data is first fitted to determine expected renewable energy output data, and a renewable energy output model is constructed by combining scene disturbance correction. Through data fitting and scene correction, the accuracy of the renewable energy output model constructed can be improved, and then when a power supply side operation model is constructed in combination with a traditional energy output model constructed by traditional energy output data, the accuracy of the power supply side operation model obtained can be improved. At the same time, various types of electricity consumption behaviors including industrial electricity, interruptible load and transferable load are determined according to historical operation data, and electricity demand models are respectively constructed to construct a load side operation model. By modeling different electricity consumption behaviors, a more detailed load side operation model can be constructed, thereby improving the accuracy of the load side operation model.

[0009] In some embodiments of the present application, the renewable energy output data in the historical operation data is clustered according to scenes, and a weather-driven scene set is constructed, specifically including: Based on each data column in the renewable energy output data, feature extraction is performed on the renewable energy output data to construct a plurality of joint feature vectors. The data columns include wind speed, irradiance, air temperature, load rate, wind power output rate, photovoltaic output rate and fault indication state. According to the plurality of joint feature vectors, scenes are iteratively clustered based on unsupervised learning. When clustering, the similarity between each joint feature vector is represented based on Mahalanobis distance, so as to cluster according to the similarity to obtain each scene cluster, and the iteration is stopped until the number of scene clusters meets a preset threshold. According to each scene cluster obtained by the last iteration, a weather-driven scene set is constructed.

[0010] In the present application, feature extraction is first performed on each data column in the renewable energy output data to construct a plurality of joint feature vectors, and then scene iterative clustering of the plurality of joint feature vectors is performed by unsupervised learning. When clustering, the similarity between each joint feature vector is represented based on Mahalanobis distance, and the iteration is stopped when the number of scene clusters meets the condition. According to each scene cluster obtained by the last iteration, a weather-driven scene set is constructed. By clustering, the features of different scenes in the renewable energy output data can be fully extracted, and then each scene cluster can better reflect different situations in the new energy operation process, so that the weather-driven scene set constructed can better fully reflect the dynamic characteristics of new energy, and the accuracy of the weather-driven scene set constructed can be improved.

[0011] In some embodiments of the present application, the upper planning model is constructed with the minimum device full life cycle cost as the target based on the power supply side operation model and the load side operation model, and the upper planning constraints of the upper planning model are determined based on the planning scale, specifically including: determine a device coordination degree index according to the power supply side operation model and the load side operation model; decide on power supply device commissioning and power grid device expansion as decision variables, construct various decision costs of the devices of the power system, and integrate to obtain device full life cycle costs; wherein the decision costs include commissioning costs, operation and maintenance costs, fuel costs, and device residual costs; minimize device full life cycle costs as an objective, combine the device full life cycle costs and the device coordination degree index to construct an upper layer planning model; predict annual maximum predicted loads and flexible regulation margins of the power system according to the historical operation data, and determine upper layer planning constraints of the upper layer planning model based on planning scales, in combination with the annual maximum predicted loads and the flexible regulation margins.

[0012] The present application first determines a device coordination degree index according to a power supply side operation model and a load side operation model, then decides on power supply device commissioning and power grid device expansion as decision variables, constructs various decision costs of the devices including commissioning costs, operation and maintenance costs, fuel costs, and device residual costs, integrates to obtain device full life cycle costs, then minimizes device full life cycle costs as an objective, combines the device full life cycle costs and the device coordination degree index to construct an upper layer planning model, and determines upper layer planning constraints based on planning scales, in combination with annual maximum predicted loads and flexible regulation margins predicted from historical operation data. The construction of device full life cycle costs from various decision costs and the construction of an upper layer planning model in combination with a device coordination degree index can fully and comprehensively construct an upper layer objective, improve the matching degree of the upper layer planning model and actual demands, and simultaneously study upper layer planning constraints in a planning time scale, which matches the actual demands of the upper layer planning model, thereby improving the accuracy of the upper layer planning model and its constraints.

[0013] In certain embodiments of the present application, the lower layer scheduling model is constructed to minimize the periodic operation cost of the power system according to the weather-driven scenario set, and lower layer scheduling constraints of the lower layer scheduling model are determined based on scheduling scales, specifically including: determine various periodic operation sub-costs of the power system according to the historical operation data, and integrate to obtain the periodic operation cost of the power system; wherein the periodic operation sub-costs include generation costs, network loss costs, and scenario penalties; wherein the network loss cost is calculated based on a node unit price in the power system and power flow losses of each node; and the scenario penalty is obtained in combination with the weather-driven scenario set; construct a lower layer scheduling model to minimize the periodic operation cost of the power system; Determine the price scheduling constraint of the node unit price based on the scheduling scale to obtain the lower scheduling constraint of the lower scheduling model.

[0014] The application first determines each cycle operation sub-cost including generation cost, network loss cost and scenario penalty based on historical operation data, and then integrates to obtain cycle operation cost. Then, the lower scheduling model is constructed with the goal of minimizing the cycle operation cost, and the price scheduling constraint of the node unit price in the network loss cost is determined based on the scheduling scale, and then the lower scheduling constraint is determined. The cycle operation cost is constructed by multiple cycle operation sub-costs, and then the lower scheduling model is constructed. The lower target can be fully and comprehensively constructed, the matching degree of the lower scheduling model and the actual demand is improved, and the accuracy of the lower scheduling model and its constraints is improved at the same time. At the same time, the weather-driven scenario set constructed by the renewable energy output data is combined to determine the scenario penalty, which can integrate the dynamic characteristics of new energy into the lower scheduling model, and then the lower scheduling model is more in line with the actual demand when solving at the scheduling scale, improving the accuracy of the lower scheduling model.

[0015] In some embodiments of the application, the upper planning model and the lower scheduling model are combined into a bi-level optimization model, and the bi-level optimization model is iteratively solved based on the Benders decomposition algorithm to obtain the planning scheme of the power system, specifically including: combining the upper planning model and the lower scheduling model into a bi-level optimization model; iteratively solving the bi-level optimization model based on the Benders decomposition algorithm. When iterating, the initial planning scheme is solved by the upper planning model and passed to the lower scheduling model. The lower scheduling model simulates the initial planning scheme to update the cycle operation cost and passes it to the upper planning model for optimization of the initial planning scheme until the bi-level optimization model converges to a preset threshold to complete iteration. the initial planning scheme obtained by the last iteration is taken as the planning scheme of the power system.

[0016] The application first combines the upper planning model and the lower scheduling model into a bi-level optimization model, and then iteratively solves based on the Benders decomposition algorithm to obtain the planning scheme of the power system. Compared with the existing single planning model, the bi-level planning model can fully consider the different aspects of optimization in the power system, and then improve the self-reflection of the planning solution and improve the accuracy of the power system planning.

[0017] In some embodiments of the application, the planning optimization of the power system according to the planning scheme specifically includes: According to the planning scheme, simulation is performed in combination with the historical operation data to obtain simulation operation data; According to the simulation operation data, index verification is performed on the planning scheme, and when the index verification is passed, the power system is planned and optimized according to the planning scheme.

[0018] According to the planning scheme, simulation is performed in combination with the historical operation data to obtain simulation operation data, and then index verification is performed on the planning scheme, and after the index verification is passed, the power system is planned and optimized. Compared with the prior art of not verifying the planning, the simulation and index verification are performed on the planning scheme, and the planning optimization is performed when the verification is passed, so that the planning scheme can meet the optimization requirements and ensure the feasibility of the planning scheme.

[0019] According to a second aspect of the embodiments of the present application, a power system double-layer optimization planning device is provided, which comprises a data construction preparation module, an upper-layer model construction module, a lower-layer model construction module, and a model solution optimization module. The data construction preparation module is configured to construct a power source side operation model and a load side operation model according to historical operation data of a power system to be planned, and perform scene clustering on renewable energy output data in the historical operation data to construct a weather-driven scene set. The upper-layer model construction module is configured to construct an upper-layer planning model with the minimization of device full life cycle cost as a target according to the power source side operation model and the load side operation model, and determine upper-layer planning constraints of the upper-layer planning model based on planning scales. The lower-layer model construction module is configured to construct a lower-layer scheduling model with the minimization of periodic operation cost of the power system as a target according to the weather-driven scene set, and determine lower-layer scheduling constraints of the lower-layer scheduling model based on scheduling scales. The model solution optimization module is configured to combine the upper-layer planning model and the lower-layer scheduling model into a double-layer optimization model, iteratively solve the double-layer optimization model based on a Benders decomposition algorithm, obtain a planning scheme of the power system, and plan and optimize the power system according to the planning scheme.

[0020] In some embodiments of the present application, the historical operation data comprises renewable energy output data and traditional energy output data; the data construction preparation module comprises a model construction submodule; and the model construction submodule comprises a new energy model construction unit, a power source side model construction unit, a power demand model construction unit, and a load side model construction unit. The new energy model construction unit is configured to fit renewable energy expected output data according to the renewable energy output data, and obtain a renewable energy output model through scene disturbance correction. The power supply side model construction unit is configured to construct a power supply side operation model according to the renewable energy output model and a traditional energy output model, wherein the traditional energy output model is constructed based on the traditional energy output data. The power consumption demand model construction unit is configured to determine a plurality of power consumption behavior types according to the historical operation data, and model each power consumption behavior type to obtain a corresponding power consumption demand model, wherein the power consumption behavior types include industrial power consumption, interruptible load and transferable load. The load side model construction unit is configured to construct a load side operation model according to the power consumption demand model corresponding to each power consumption behavior type.

[0021] In some embodiments of the present application, the data construction preparation module includes a scene set construction submodule, which includes a data feature extraction unit, a vector iterative clustering unit and a scene set construction unit. The data feature extraction unit is configured to extract features from the renewable energy output data based on each data column in the renewable energy output data to construct a plurality of joint feature vectors, wherein the data columns include wind speed, irradiance, air temperature, load rate, wind power output rate, photovoltaic output rate and fault indication state. The vector iterative clustering unit is configured to iteratively cluster scenes based on unsupervised learning according to the plurality of joint feature vectors, and to represent the similarity between each joint feature vector based on Mahalanobis distance when clustering to cluster each scene cluster according to the similarity until the number of scene clusters meets a preset threshold to stop iteration. The scene set construction unit is configured to construct a weather-driven scene set according to each scene cluster obtained through the last iteration.

[0022] In some embodiments of the present application, the upper layer model construction module includes a coordination degree index determination unit, a decision cost construction and integration unit, an upper layer planning model construction unit and an upper layer planning constraint determination unit. The coordination degree index determination unit is configured to determine a device coordination degree index according to the power supply side operation model and the load side operation model. The decision cost construction and integration unit is configured to make decisions by taking power supply device commissioning and power grid device expansion as decision variables, to construct a plurality of decision costs of each device of the power system, and to integrate to obtain a device full life cycle cost, wherein the decision costs include commissioning cost, operation and maintenance cost, fuel cost and device residual cost. The upper-layer planning model construction unit is configured to construct an upper-layer planning model by taking minimization of a device full life cycle cost as a target, and combining the device full life cycle cost and the device coordination degree index. The upper-layer planning constraint determination unit is configured to predict an annual maximum predicted load and a flexible adjustment margin of the power system according to the historical operation data, and determine upper-layer planning constraints of the upper-layer planning model based on a planning scale and in combination with the annual maximum predicted load and the flexible adjustment margin.

[0023] In some embodiments of the present application, the lower-layer model construction module includes a cycle cost construction integration unit, a lower-layer scheduling model construction unit, and a lower-layer scheduling constraint determination unit. The cycle cost construction integration unit is configured to determine each cycle operation sub-cost of the power system according to the historical operation data, and integrate to obtain a cycle operation cost of the power system; wherein the cycle operation sub-cost includes a generation cost, a network loss cost, and a scenario penalty; wherein the network loss cost is calculated based on a node unit price and a power flow loss of each node in the power system; and the scenario penalty is obtained in combination with the weather-driven scenario set; The lower-layer scheduling model construction unit is configured to construct a lower-layer scheduling model by taking minimization of the cycle operation cost of the power system as a target. The lower-layer scheduling constraint determination unit is configured to determine a price scheduling constraint of the node unit price based on a scheduling scale, and obtain lower-layer scheduling constraints of the lower-layer scheduling model.

[0024] In some embodiments of the present application, the model solution optimization module includes a planning scheme solution sub-module; the planning scheme solution sub-module includes a double-layer model combination unit, a double-layer model solution unit, and a planning scheme determination unit. The double-layer model combination unit is configured to combine the upper-layer planning model and the lower-layer scheduling model into a double-layer optimization model. The double-layer model solution unit is configured to perform iterative solution on the double-layer optimization model based on a Benders decomposition algorithm; in the iteration, an initial planning scheme is solved by the upper-layer planning model and is passed to the lower-layer scheduling model, the lower-layer scheduling model simulates the initial planning scheme to update the cycle operation cost and passes it to the upper-layer planning model, so that the upper-layer planning model optimizes the initial planning scheme, until the double-layer optimization model converges to a preset threshold value, and the iteration is completed. The planning scheme determination unit is configured to take the initial planning scheme obtained in the last iteration as a planning scheme of the power system.

[0025] In some embodiments of the present application, the model solution optimization module comprises a system planning optimization sub-module; the system planning optimization sub-module comprises a data simulation simulation unit and a scheme verification optimization unit; The data simulation simulation unit is configured to simulate and verify, according to the planning scheme and in combination with the historical operation data, to obtain simulation operation data; The scheme verification optimization unit is configured to verify, according to the simulation operation data, the planning scheme, and when the verification is passed, to optimize the planning of the power system according to the planning scheme.

[0026] According to the historical operation data of the power system to be planned, the power source side operation model, the load side operation model and the weather driven scenario set are constructed, and then the upper planning model and the lower scheduling model are constructed with the minimum device life cycle cost as the target and the minimum periodic operation cost as the target, respectively, and then combined into a double-layer optimization model and iteratively solved based on the Benders decomposition algorithm. Compared with the existing single-layer planning model, the double-layer planning model can fully consider the different aspects of optimization in the power system, thereby improving the accuracy of the planning solution, i.e., improving the accuracy of the planning of the power system. At the same time, the upper planning model determines the upper planning constraints based on the planning scale, the lower scheduling model determines the lower scheduling constraints based on the scheduling scale, and the influence of different time scales of planning and scheduling is studied respectively, and the weather driven scenario set obtained by clustering the renewable energy output data is used to construct the lower scheduling model, which can fully reflect the dynamic characteristics of new energy, thereby improving the accuracy of the solution, i.e., improving the accuracy of the planning of the power system.

[0027] According to a third aspect of the embodiments of the present application, a computer device is provided, comprising: a processor; a memory; a computer program stored in the memory and configured to be executed by the processor; wherein the processor implements the power system double-layer optimization planning method according to the present application when executing the computer program.

[0028] According to a fourth aspect of the embodiments of the present application, a computer readable storage medium is provided, which stores a plurality of instructions adapted to be loaded by a processor to execute the power system double-layer optimization planning method according to the present application. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 A flowchart of a power system double-layer optimization planning method according to some embodiments of the present application is shown; Figure 2A module structure diagram of a power system double-layer optimization planning device shown in some embodiments of the present application. DETAILED DESCRIPTION

[0030] The embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by way of accompanying drawings are exemplary and are only used to explain some embodiments of the present application, and cannot be understood as limiting the embodiments of the present application. Based on the embodiments shown in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0031] In the description of the present application, it should be understood that the terms "first" and "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include at least one of the features. In the description of the present application, unless otherwise specifically limited, the meaning of "a plurality of" "several" is two or more.

[0032] The existing planning for the power system mainly uses a single-layer planning model to solve the planning scheme, the solution is insufficient, and the scheme is not verified, resulting in an inaccurate planning scheme and low practical feasibility. At the same time, the existing technology usually only studies the influence of new energy operation on the power system under a single time scale, and cannot reflect the different influences of new energy operation under multiple time scales, resulting in a planning scheme that cannot fully reflect the dynamic characteristics of new energy and is not accurate and reliable in actual application. Therefore, how to reasonably plan the power system to fully reflect the characteristics of new energy and improve the accuracy of the planning scheme is still a technical problem to be solved in the prior art.

[0033] It is easily understood by those skilled in the art that the renewable energy or new energy described in the present application actually refers to wind power generation using wind energy or photovoltaic power generation using solar energy; and the conventional energy described in the present application actually refers to thermal power generation using non-renewable fossil fuels.

[0034] Based on the above technical background, please refer to Figure 1 The embodiments of the present application provide a power system double-layer optimization planning method, which comprises steps S101 to S104, and each step is specifically as follows: Step S101: Based on the historical operating data of the power system to be planned, construct the power supply side operation model and the load side operation model, and perform scenario clustering on the renewable energy output data in the historical operating data to construct a weather-driven scenario set.

[0035] In some embodiments of this application, the historical operating data includes renewable energy output data and traditional energy output data; the step of constructing a power source-side operating model and a load-side operating model based on the historical operating data of the power system to be planned specifically includes: Based on the renewable energy output data, the expected renewable energy output data is fitted, and the renewable energy output model is obtained through scenario perturbation correction. Based on the renewable energy output model and the traditional energy output model, a power source operation model is constructed; wherein, the traditional energy output model is constructed based on the traditional energy output data; Based on the historical operating data, various electricity consumption behavior types are identified, and a model is created for each type to obtain a corresponding electricity demand model; wherein, the electricity consumption behavior types include industrial electricity consumption, interruptible loads, and transferable loads; Based on the electricity demand model corresponding to each type of electricity consumption behavior, a load-side operation model is constructed.

[0036] In some embodiments of this application, the renewable energy output model is specifically as follows: ; in, Indicates that renewable energy is at a certain time of effort, Indicates that renewable energy is at a certain time The expected contribution Indicates time Corresponding weather correction factor, This indicates scene perturbation. Generally, scene perturbation... The preferred preset value is 1.

[0037] The reason for considering correcting the output of renewable energy through weather changes and scenario disturbances is that these factors can reasonably reflect the load fluctuations of renewable energy caused by weather or scenario changes, thereby obtaining more realistic and accurate renewable energy output data based on the corresponding load fluctuations.

[0038] In some embodiments of this application, the power supply side operation model is specifically as follows: ; in, Indicates the power supply side at time Total output Indicates that traditional energy is at any time Total output.

[0039] The reason for defining the total power output on the power supply side as the sum of the output of renewable energy and traditional energy is that this definition simplifies the power supply side model, thereby reducing the difficulty of subsequent model calculation and solving, and integrates the output of the two types of energy as power sources, providing a corresponding model foundation for the construction, solution and calculation of the upper-level planning model.

[0040] In some embodiments of this application, the electricity demand model corresponding to each type of electricity consumption behavior is specifically as follows: The electricity demand model corresponding to industrial electricity consumption is as follows: ; in, Indicates the time of industrial electricity users The active power demand, The total number of devices that require electricity. Indicates the first Load rate of the equipment. Indicates the first The equipment is at all times The running status, Indicates the first The power of the equipment under rated load; The electricity demand model corresponding to the interruptible load is as follows: ; in, Indicates commercial and residential users at any time Interruptible load power, The total number of interruptible commercial and residential users, For a collection of interruptible commercial and residential users, Indicates time The interruption rate Indicates the interruption of contract capacity; The electricity demand model corresponding to the transferable load is as follows: ; in, Indicates commercial and residential users at any time Transferable load power, The total number of transferable commercial and residential users, For a collection of transferable commercial and residential users, They are time points The original power curve and the power curve after load transfer.

[0041] Different models are respectively constructed for industrial electricity, interruptible load and transferable load because different electricity behaviors have different electricity demand curves and different electricity demand models, and the respective construction is beneficial to improve the matching degree of each electricity demand model and the corresponding electricity behavior, so as to obtain more actual and more accurate industrial electricity model, interruptible load model and transferable load model.

[0042] In some embodiments of the present application, the load side operation model, in particular: ; Among them, represents the total output of the load side at time .

[0043] The total output of the load side is defined as the sum of the outputs of industrial electricity, interruptible load and transferable load, because such definition can simplify the load side model, thereby reducing the difficulty of subsequent model calculation and solution, and integrating the outputs of the loads corresponding to different behaviors, providing a corresponding model basis for the construction, solution and calculation of the upper planning model.

[0044] The present application first determines the expected renewable energy output data according to the renewable energy output data, constructs the renewable energy output model in combination with the scene disturbance correction, and through data fitting and scene correction, the accuracy of the renewable energy output model constructed can be improved, and then in combination with the traditional energy output model constructed by the traditional energy output data, the accuracy of the power source side operation model obtained can be improved; at the same time, according to the historical operation data, a plurality of electricity behavior types including industrial electricity, interruptible load and transferable load are determined, and electricity demand models are respectively constructed to construct the load side operation model, and through modeling of different electricity behavior types, a more detailed load side operation model can be constructed, thereby improving the accuracy of the load side operation model.

[0045] In some embodiments of the present application, the renewable energy output data in the historical operation data is clustered according to scenes to construct a weather-driven scene set, specifically including: Based on each data column in the renewable energy output data, feature extraction is performed on the renewable energy output data to construct a plurality of joint feature vectors; wherein the data column includes wind speed, irradiance, air temperature, load rate, wind power output rate, photovoltaic output rate and fault indication state; According to the plurality of joint feature vectors, scenes are iteratively clustered based on unsupervised learning, and when clustering, the similarity between each joint feature vector is represented based on Mahalanobis distance, so as to cluster according to the similarity to obtain each scene cluster, and the iteration is stopped until the number of scene clusters meets the preset threshold. According to each scenario cluster obtained by the last iteration, a weather-driven scenario set is constructed.

[0046] Specifically, when constructing the joint feature vector, a sample data is selected from the renewable energy output data, and each data column of the renewable energy output data is taken as a data dimension to construct the joint feature vector corresponding to each sample data , wherein represents the wind speed, represents the irradiance, represents the air temperature, represents the load rate, represents the wind power output rate, represents the photovoltaic output rate, represents the fault indication state. Generally, the wind power output rate and the photovoltaic output rate are normalized to the target range [0, 1] in the process of constructing the joint feature vector; the fault indication state is a 0-1 identifier, and the value 0 indicates that the corresponding device has not failed, and the value 1 indicates that the corresponding device has failed.

[0047] Specifically, when iteratively clustering the scenarios based on unsupervised learning, the unsupervised clustering algorithm applied includes the K-means algorithm and its improved clustering algorithm, the Hierarchical Clustering algorithm, the grid-based clustering algorithm and the DBSCAN algorithm, and the K-means algorithm is preferred. More specifically, the Mahalanobis distance is used to represent the similarity between the joint feature vectors during clustering; wherein and are the i-th joint feature vector and the j-th joint feature vector, respectively, is the covariance matrix of the joint feature vector . Generally, in the clustering process, in order to represent the similarity between samples and classify the samples into corresponding clusters, the Euclidean distance between samples is usually taken as the similarity between samples, but the Euclidean distance is difficult to identify the correlation between variables and exclude its interference. In order to eliminate the clustering interference caused by the correlation between variables and avoid clustering errors, the Euclidean distance can be replaced by the Mahalanobis distance which can eliminate the correlation between variables.

[0048] More specifically, the preferred value of the preset threshold of the number of scenario clusters in the present application is 12. Exemplarily, the scenarios corresponding to each scenario cluster in an instance of the weather-driven scenario set are shown in Table 1 below.

[0049] Table 1: Examples of each scenario cluster of a weather-driven scenario set ​​ In the table, Indicates wind speed. Indicates temperature, Indicates irradiation. Indicates rainfall amount, These represent the actual power generation from wind power and the actual power generation from photovoltaic power, respectively. These represent the installed capacity of wind power and photovoltaic power, respectively. These represent the actual annual load and the maximum annual load, respectively. This indicates the average number of failures per day.

[0050] This application first extracts features from each data column in the renewable energy output data to construct multiple joint feature vectors. Then, it performs unsupervised learning-based iterative scene clustering on these joint feature vectors, using Mahalanobis distance to represent the similarity between the joint feature vectors. The clustering continues until the number of scene clusters meets the condition. Based on the scene clusters obtained in the last iteration, a weather-driven scene set is constructed. This clustering method can fully extract the features of different scenarios in the renewable energy output data, making each scene cluster more reflective of different situations in the operation of new energy. As a result, the constructed weather-driven scene set can more fully reflect the dynamic characteristics of new energy and improve the accuracy of the constructed weather-driven scene set.

[0051] Step S102: Based on the power supply side operation model and the load side operation model, construct an upper-level planning model with the goal of minimizing the equipment's total life cycle cost, and determine the upper-level planning constraints of the upper-level planning model based on the planning scale.

[0052] In some embodiments of this application, the step of constructing an upper-level planning model based on the power supply-side operation model and the load-side operation model, with the goal of minimizing the equipment's total lifecycle cost, and determining the upper-level planning constraints of the upper-level planning model based on the planning scale, specifically includes: Based on the power supply side operation model and the load side operation model, determine the equipment coordination index; The decision-making process uses the commissioning of power equipment and the expansion of grid equipment as decision variables to construct various decision costs for each piece of equipment in the power system, and integrates them to obtain the equipment's total life cycle cost; wherein, the decision costs include commissioning costs, operation and maintenance costs, fuel costs, and equipment residual costs; With the goal of minimizing the total lifecycle cost of equipment, an upper-level planning model is constructed by combining the total lifecycle cost of equipment and the equipment synergy index. According to the historical operation data, the annual maximum predicted load and the flexible regulation margin of the power system are predicted, and based on the planning scale, the upper planning constraint of the upper planning model is determined in combination with the annual maximum predicted load and the flexible regulation margin.

[0053] In some embodiments of the present application, the equipment coordination degree index, in particular, ; Among them, represents the equipment coordination degree index, represents the total output of the power supply side at time , represents the total output of the load side at time , represents the available capacity of the energy storage end at time , represents the power fluctuation amplitude at time , is a weight factor. Generally, the weight factor is preferably 0.5, 0.5.

[0054] Considering that the equipment coordination degree index is determined by the ratio of the power supply side output to the load side output and the ratio of the energy storage end available capacity to the power fluctuation amplitude, the two ratios can measure the degree of coordinated operation between the power supply side and the load side equipment and the degree of coordinated operation of the equipment corresponding to the available capacity of the energy storage side and the power fluctuation, respectively. By measuring the equipment coordination degree, the problems of supply and demand imbalance and low equipment utilization that may occur during planning can be avoided.

[0055] In some embodiments of the present application, the equipment life cycle cost, in particular, ; ; ; ; ; Among them, represents the equipment life cycle cost, represents the total planning year, are the commissioning cost, operation and maintenance cost, fuel cost and equipment residual cost of the th year, respectively, is the discount rate; represents the type of the th equipment, including power generation equipment , line network equipment and energy storage equipment ; Indicates the first New capacity for similar devices Indicates the first The one-time investment per unit capacity of this type of equipment, Indicates the first Physical lifespan of this type of equipment Indicates the first Annual operation and maintenance cost rate for this type of equipment Indicates that traditional energy is at any time The cost of electricity generation, The total length of the presidential term throughout the year (unit: hours). Indicates time interval, Indicates the first The average annual residual value factor for this type of equipment. Generally, the discount rate... The preferred preset value is 8%; annual maintenance cost rate The preferred preset value is 2% when the equipment is a power generation device using traditional energy, 5% when the equipment is a power generation device using renewable energy, 0.2% when the equipment is a transmission line, and 2% when the equipment is an energy storage device; the annual average residual value factor The preferred preset value is .

[0056] By considering the total discounted value of various costs in each planning year, i.e. by constructing the equipment life cycle cost, we can reflect the cost changes of the equipment in each planning year within the total planning period, so that the equipment life cycle cost is more in line with the cost changes of the equipment in actual operation, thus obtaining a more accurate equipment life cycle cost.

[0057] In some embodiments of this application, the upper-level planning model specifically refers to: ; in, Indicates to make The planning scheme corresponding to the minimum, This is the synergy coefficient.

[0058] In some embodiments of this application, the upper-level planning constraint specifically includes: ; in, Indicates planning scheme Corresponding system capacity These are the annual maximum predicted load and the flexible adjustment margin, respectively.

[0059] By defining the upper planning constraint as the size relationship between the system capacity and the sum of the annual maximum predicted load and the flexible regulation margin, the available flexible regulation margin can be reserved for load fluctuations, equipment failures, load fluctuations of renewable energy and other sudden situations during model solving, ensuring stable and reliable operation of the system, making the planning scheme obtained by solving more consistent with actual implementation, and thus obtaining a more accurate planning scheme.

[0060] The application first determines the equipment coordination degree index according to the power supply side operation model and the load side operation model, then makes decisions by taking the power equipment commissioning and the power grid equipment expansion as decision variables, constructs various decision costs of each equipment including commissioning cost, operation and maintenance cost, fuel cost and equipment residual cost, integrates to obtain the equipment life cycle cost, then constructs the upper planning model by taking the minimization of the equipment life cycle cost as the target, combining the equipment life cycle cost and the equipment coordination degree index, and determines the upper planning constraint based on the planning scale, combining the annual maximum predicted load and the flexible regulation margin predicted through historical operation data, constructs the equipment life cycle cost by various decision costs, and constructs the upper planning model by combining the equipment coordination degree index, which can fully and comprehensively construct the upper target, improve the matching degree of the upper planning model and the actual demand, and at the same time, study the upper planning constraint in the time scale of planning, match the actual demand of the upper planning model, thereby improving the accuracy of the upper planning model and its constraints.

[0061] Step S103: According to the weather-driven scenario set, an underlayer scheduling model is constructed by taking minimization of the periodic operation cost of the power system as the target, and underlayer scheduling constraints of the underlayer scheduling model are determined based on the scheduling scale.

[0062] In some embodiments of the application, the underlayer scheduling model is constructed by taking minimization of the periodic operation cost of the power system as the target according to the weather-driven scenario set, and the underlayer scheduling constraints of the underlayer scheduling model are determined based on the scheduling scale, which specifically includes: According to the historical operation data, each periodic operation sub-cost of the power system is determined, and the periodic operation cost of the power system is integrated; wherein the periodic operation sub-cost includes generation cost, network loss cost and scenario penalty; wherein the network loss cost is calculated based on the node unit price in the power system and the power flow loss of each node; the scenario penalty is obtained in combination with the weather-driven scenario set; The underlayer scheduling model is constructed by taking minimization of the periodic operation cost of the power system as the target; Based on the scheduling scale, the price scheduling constraint of the node unit price is determined to obtain the underlayer scheduling constraints of the underlayer scheduling model.

[0063] In some embodiments of this application, the cycle operating cost specifically includes: ; ; ; in, Indicates the cycle operating cost, The length of the presidential time during the planning period (unit: hours). For a moment The cost of electricity generation, For a moment Network loss cost To constrain the penalty coefficient for violations, For weather-driven scene sets, For the scene Weather-driven scenario set The percentage of samples belonging to the scene cluster in the middle. For the scene At any moment The amount of violations that are restricted This represents the cost coefficient for conventional peak shaving. This indicates the minimum output for regular peak shaving. This represents the cost coefficient for deep peak shaving. This represents the minimum output for deep peak tuning. This is the critical point for deep peak scaling. For a moment The unit electricity price at the node For a moment Total node power flow loss.

[0064] By defining the cycle operating cost as the sum of generation cost, grid loss cost, and scenario penalty cost obtained by combining grid loss cost with scenario penalty, the cost of system operation during the planning period can be integrated and simplified. This allows for accurate simulation and convenient solution. Furthermore, by using these three types of costs, the actual cycle operating cost can be reflected more accurately and comprehensively, thus obtaining a more precise cycle operating cost.

[0065] In some embodiments of this application, the lower-level scheduling model specifically includes: ; in, Indicates to make The range of values ​​for the decision variable group corresponding to the minimum value.

[0066] In some embodiments of this application, the lower-level scheduling constraint specifically includes: The lower-level scheduling constraints include the electricity price scheduling constraints; the electricity price constraints are specifically as follows: ; wherein, denotes the node price fluctuation vector at time , is a joint probability distribution constructed based on a Copula function, respectively denote the price vector of the node after the node is connected with wind power or photovoltaic.

[0067] By defining the lower layer scheduling constraint as the size relationship between the node unit price and the node price fluctuation vector, the nonlinear correlation between the wind and light output and the price can be accurately and flexibly described by the node price fluctuation vector obtained from the joint probability distribution of wind power and photovoltaic constructed by the Copula function, and then the fluctuation of the node unit price is regulated by the node price fluctuation vector, the rationality of the price regulation in the lower layer scheduling is ensured, the planning scheme obtained by solving is more in line with the actual implementation, and a more accurate planning scheme is obtained.

[0068] In the present application, the periodic operation sub-costs including generation cost, network loss cost and scenario penalty are determined according to historical operation data, and then the periodic operation cost is integrated, and then the lower layer scheduling model is constructed with the objective of minimizing the periodic operation cost, and the price scheduling constraint of the node unit price in the network loss cost is determined based on the scheduling scale, and then the lower layer scheduling constraint is determined, the periodic operation cost is constructed by multiple periodic operation sub-costs, and then the lower layer scheduling model is constructed, which can fully and comprehensively construct the lower layer target, improve the matching degree of the lower layer scheduling model and the actual demand, and at the same time, the lower layer scheduling constraint is studied in the time scale of scheduling, which matches the actual demand of the lower layer scheduling model, thereby improving the accuracy of the lower layer scheduling model and its constraint; at the same time, the scenario penalty is determined by combining the weather-driven scenario set constructed by the renewable energy output data, which can integrate the dynamic characteristics of new energy into the lower layer scheduling model, and then the lower layer scheduling model is more in line with the actual demand when solving in the scheduling scale, thereby improving the accuracy of the lower layer scheduling model.

[0069] Step S104: combining the upper layer planning model and the lower layer scheduling model into a bi-level optimization model, iteratively solving the bi-level optimization model based on the Benders decomposition algorithm, obtaining the planning scheme of the power system, and planning and optimizing the power system according to the planning scheme.

[0070] In some embodiments of the present application, the combination of the upper layer planning model and the lower layer scheduling model into a bi-level optimization model, the iterative solving of the bi-level optimization model based on the Benders decomposition algorithm, and the obtaining of the planning scheme of the power system specifically include: combining the upper layer planning model and the lower layer scheduling model into a bi-level optimization model; Based on the Benders decomposition algorithm, the two-level optimization model is iteratively solved. During the iteration, the initial planning scheme is solved by the upper-level planning model and passed to the lower-level scheduling model. The lower-level scheduling model simulates the initial planning scheme to update the cycle running cost and passes it to the upper-level planning model so that the upper-level planning model can optimize the initial planning scheme. The iteration is completed when the two-level optimization model converges to a preset threshold. The initial planning scheme obtained from the last iteration is used as the planning scheme for the power system.

[0071] This application first combines the upper-level planning model and the lower-level scheduling model into a two-level optimization model, and then iteratively solves the problem based on the Benders decomposition algorithm to obtain the planning scheme of the power system. Compared with the existing method of solving the planning problem using only a single-level planning model, the two-level planning model can fully consider different aspects of optimization in the power system, thereby improving the introspection of the main group of the planning solution and improving the accuracy of the power system planning.

[0072] In some embodiments of this application, the step of optimizing the power system according to the planning scheme specifically includes: Based on the planning scheme, simulation is performed using the historical operating data to obtain simulation operating data; Based on the simulation data, the planning scheme is validated. When the validation is successful, the power system is optimized according to the planning scheme.

[0073] In some embodiments of this application, the indicators that need to be verified include economic indicators and reliability indicators; the economic indicators specifically include: ; in, This represents the system's cost per kilowatt-hour. Indicates the total planning period. The first Annual production costs, operation and maintenance costs, fuel costs, and remaining equipment costs. They represent the first Annual power generation and equivalent discharge on the energy storage side. This indicates the charging and discharging efficiency of the energy storage system. This represents the power curtailment penalty coefficient. Indicates the first Annual abandoned electricity Indicates the first Theoretical annual power generation; The reliability indicators are specifically as follows: ; ; wherein, respectively represent line overload risk index and voltage out-of-limit sensitivity index, respectively represent total number of transmission lines and total number of buses, respectively represent rated transmission power and maximum transmission power of the i-th transmission line, respectively represent rated transmission power and maximum transmission power of the i-th transmission line, represent overload duration of the i-th transmission line, represent total length of evaluation period, represent voltage sensitivity to reactive power of the i-th bus, represent reactive power and voltage of the i-th bus. respectively represent reactive power and voltage of the i-th bus.

[0074] The economic index is constructed by the system generation cost, and the reliability index is constructed by the line overload risk index and the voltage out-of-limit sensitivity index, because the system generation cost, the line overload risk index and the voltage out-of-limit sensitivity index can all represent the current economy and reliability of the system in the most significant way, i.e., represent the rationality of the system cost and the risk of system operation under the current planning scheme, so as to ensure that the cost calculation is reasonable and the system has a certain reliability when the planning scheme is solved, thereby ensuring that the planning scheme obtained by solving has feasibility / implementation.

[0075] In some embodiments of the present application, when the index verification fails, the coefficients of the upper planning model and the lower scheduling model in the double-layer optimization model are adjusted, and the planning scheme is solved again.

[0076] The present application first simulates the simulation operation data according to the planning scheme combined with the historical operation data, and then verifies the planning scheme by the index, and optimizes the power system after the index verification passes. Compared with the prior art which does not verify the planning, the simulation and index verification are performed through the planning scheme, and the planning optimization is executed when the verification passes, which can ensure that the planning scheme meets the optimization demand and guarantees the feasibility of the planning scheme.

[0077] ​​​​Compared to existing technologies, this application first constructs a power source-side operation model, a load-side operation model, and a weather-driven scenario set based on historical operating data of the power system to be planned. Then, it constructs an upper-level planning model and a lower-level scheduling model with the objectives of minimizing the equipment's total life-cycle cost and minimizing the cycle operating cost, respectively. These are then combined into a two-level optimization model and iteratively solved using the Benders decomposition algorithm. Compared to existing methods that only use a single-level planning model, this application, through the combination of two-level planning models, can fully consider different aspects of optimization in the power system, thereby improving the accuracy of the planning solution, i.e., improving the accuracy of power system planning. At the same time, the upper-level planning model determines upper-level planning constraints based on the planning scale, and the lower-level scheduling model determines lower-level scheduling constraints based on the scheduling scale. The impact of planning and scheduling at different time scales is studied separately. Furthermore, the lower-level scheduling model is constructed using weather-driven scenario sets obtained by clustering renewable energy output data, which can fully reflect the dynamic characteristics of new energy sources, thereby improving the accuracy of the solution during planning, i.e., improving the accuracy of power system planning.

[0078] For a method corresponding to the one described above, please refer to [link / reference]. Figure 2 This application provides a two-layer optimization planning device for a power system, including a data construction and preparation module 210, an upper-layer model construction module 220, a lower-layer model construction module 230, and a model solving and optimization module 240. The data construction and preparation module 210 is used to construct a power source side operation model and a load side operation model based on the historical operation data of the power system to be planned, and to perform scenario clustering on the renewable energy output data in the historical operation data to construct a weather-driven scenario set. The upper-level model construction module 220 is used to construct an upper-level planning model based on the power supply side operation model and the load side operation model, with the goal of minimizing the equipment's total life cycle cost, and to determine the upper-level planning constraints of the upper-level planning model based on the planning scale. The lower-level model construction module 230 is used to construct a lower-level scheduling model based on the weather-driven scenario set, with the goal of minimizing the cycle operation cost of the power system, and to determine the lower-level scheduling constraints of the lower-level scheduling model based on the scheduling scale. The model solving and optimization module 240 is used to combine the upper-level planning model and the lower-level scheduling model into a two-level optimization model, iteratively solve the two-level optimization model based on the Benders decomposition algorithm to obtain the planning scheme of the power system, and optimize the power system according to the planning scheme.

[0079] In some embodiments of the present application, the historical operation data comprises the renewable energy output data and the traditional energy output data; the data construction preparation module 210 comprises a model construction submodule; the model construction submodule comprises a new energy model construction unit, a power supply side model construction unit, a power consumption demand model construction unit and a load side model construction unit; The new energy model construction unit is configured to fit renewable energy expected output data according to the renewable energy output data, and obtain renewable energy output model through scene disturbance correction. The power supply side model construction unit is configured to construct a power supply side operation model according to the renewable energy output model and a traditional energy output model; wherein the traditional energy output model is constructed based on the traditional energy output data. The power consumption demand model construction unit is configured to determine a plurality of power consumption behavior types according to the historical operation data, and model each power consumption behavior type to obtain a corresponding power consumption demand model; wherein the power consumption behavior types comprise industrial power consumption, interruptible load and transferable load. The load side model construction unit is configured to construct a load side operation model according to the power consumption demand model corresponding to each power consumption behavior type.

[0080] In some embodiments of the present application, the data construction preparation module 210 comprises a scene set construction submodule; the scene set construction submodule comprises a data feature extraction unit, a vector iterative clustering unit and a scene set construction unit. The data feature extraction unit is configured to extract features of the renewable energy output data based on each data column in the renewable energy output data, and construct a plurality of joint feature vectors; wherein the data columns comprise wind speed, irradiance, air temperature, load rate, wind power output rate, photovoltaic output rate and fault indication state. The vector iterative clustering unit is configured to iteratively cluster scenes based on unsupervised learning according to the plurality of joint feature vectors; when clustering, the similarity between each joint feature vector is represented based on Mahalanobis distance, so as to cluster according to the similarity to obtain each scene cluster until the number of scene clusters meets a preset threshold to stop iteration. The scene set construction unit is configured to construct a weather-driven scene set according to each scene cluster obtained through the last iteration.

[0081] In some embodiments of the present application, the upper layer model construction module 220 comprises a coordination degree index determination unit, a decision cost construction integration unit, an upper layer planning model construction unit and an upper layer planning constraint determination unit. The coordination degree index determination unit is configured to determine a device coordination degree index according to the power supply side operation model and the load side operation model. The decision cost construction and integration unit is configured to make decisions with power equipment commissioning and power grid equipment expansion as decision variables, construct multiple decision costs of each device of the power system, and integrate device full life cycle costs; wherein the decision costs include commissioning costs, operation and maintenance costs, fuel costs, and device residual costs. The upper-layer planning model construction unit is configured to construct an upper-layer planning model with minimization of the device full life cycle costs as a target, in combination with the device full life cycle costs and the device coordination degree index. The upper-layer planning constraint determination unit is configured to predict annual maximum predicted load and flexible regulation margin of the power system according to the historical operation data, and determine upper-layer planning constraints of the upper-layer planning model based on planning scales, in combination with the annual maximum predicted load and the flexible regulation margin.

[0082] In some embodiments of the present application, the lower-layer model construction module 230 includes a period cost construction and integration unit, a lower-layer scheduling model construction unit, and a lower-layer scheduling constraint determination unit. The period cost construction and integration unit is configured to determine each period operation sub-cost of the power system according to the historical operation data, and integrate period operation costs of the power system; wherein the period operation sub-costs include generation costs, network loss costs, and scenario penalties; wherein the network loss costs are calculated based on node unit prices and power flow losses of each node in the power system; and the scenario penalties are obtained in combination with the weather-driven scenario set. The lower-layer scheduling model construction unit is configured to construct a lower-layer scheduling model with minimization of the period operation costs of the power system as a target. The lower-layer scheduling constraint determination unit is configured to determine price scheduling constraints of the node unit prices based on scheduling scales, to obtain lower-layer scheduling constraints of the lower-layer scheduling model.

[0083] In some embodiments of the present application, the model solution optimization module 240 includes a planning scheme solution sub-module; the planning scheme solution sub-module includes a double-layer model combination unit, a double-layer model solution unit, and a planning scheme determination unit. The double-layer model combination unit is configured to combine the upper-layer planning model and the lower-layer scheduling model into a double-layer optimization model. The double-layer model solving unit is configured to solve the double-layer optimization model based on a Benders decomposition algorithm, and in the solving process, an initial planning scheme is solved by the upper-layer planning model and passed to the lower-layer scheduling model, the lower-layer scheduling model simulates the initial planning scheme to update a periodic operation cost and pass the updated periodic operation cost to the upper-layer planning model, and the upper-layer planning model optimizes the initial planning scheme until the double-layer optimization model converges to a preset threshold value, and the iteration is completed. The planning scheme determination unit is configured to determine the initial planning scheme obtained in the last iteration as the planning scheme of the power system.

[0084] In some embodiments of the present application, the model solving optimization module 240 comprises a system planning optimization sub-module, and the system planning optimization sub-module comprises a data simulation simulation unit and a scheme verification optimization unit. The data simulation simulation unit is configured to simulate and obtain simulation operation data according to the planning scheme and in combination with the historical operation data. The scheme verification optimization unit is configured to verify indexes of the planning scheme according to the simulation operation data, and when the index verification is passed, the power system is planned and optimized according to the planning scheme.

[0085] In the present application, first, the historical operation data of the power system to be planned is used to construct a power source side operation model, a load side operation model and a weather driven scenario set, then an upper-layer planning model and a lower-layer scheduling model are constructed with the minimization of the device full life cycle cost as the target and the minimization of the periodic operation cost as the target, and then the double-layer optimization model is combined and solved based on the Benders decomposition algorithm. Compared with the existing single-layer planning model, the double-layer planning model can fully consider different aspects of optimization in the power system, thereby improving the accuracy of planning and optimization, and improving the accuracy of planning of the power system. Meanwhile, the upper-layer planning model determines the upper-layer planning constraint based on the planning scale, the lower-layer scheduling model determines the lower-layer scheduling constraint based on the scheduling scale, the influence of different time scales of planning and scheduling is studied respectively, and the weather driven scenario set obtained by clustering renewable energy output data is used to construct the lower-layer scheduling model, which can fully reflect the dynamic characteristics of new energy, thereby improving the accuracy of planning and optimization, and improving the accuracy of planning of the power system.

[0086] It should be understood that the device provided by the embodiments of the present application corresponds to the foregoing method, and the double-layer optimization planning device of the power system provided by the embodiments of the present application can realize the double-layer optimization planning method of the power system provided by any one of the embodiments of the present application.

[0087] Adaptively, the embodiment of the application further provides a computer device and a computer readable storage medium.

[0088] The computer device comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor; When the processor executes the computer program, the power system bi-level optimization planning method of the application is implemented.

[0089] The computer readable storage medium stores a plurality of instructions, which are adapted to be loaded by the processor to execute the power system bi-level optimization planning method of the application.

[0090] The above is part of the embodiments of the application, which further details the purpose, technical solutions and beneficial effects of the application. It should be clear that the above part of the embodiments of the application cannot be understood as the limitation of the application. It is particularly pointed out that any change, modification, equivalent replacement and variation, etc. made by the person skilled in the art within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A two-level optimization planning method for power systems, characterized in that, include: Based on the historical operating data of the power system to be planned, a power source side operation model and a load side operation model are constructed, and the renewable energy output data in the historical operating data are clustered by scenario to construct a weather-driven scenario set; Based on the power supply side operation model and the load side operation model, with the goal of minimizing the total life cycle cost of the equipment, an upper-level planning model is constructed, and the upper-level planning constraints of the upper-level planning model are determined based on the planning scale. Based on the weather-driven scenario set, and with the goal of minimizing the cycle operating cost of the power system, a lower-level scheduling model is constructed, and the lower-level scheduling constraints of the lower-level scheduling model are determined based on the scheduling scale. The upper-level planning model and the lower-level scheduling model are combined into a two-level optimization model. The two-level optimization model is iteratively solved based on the Benders decomposition algorithm to obtain the planning scheme of the power system. The power system is then optimized according to the planning scheme.

2. The two-level optimization planning method for power systems according to claim 1, characterized in that, The historical operating data includes renewable energy output data and traditional energy output data; the construction of power source-side operating models and load-side operating models based on the historical operating data of the power system to be planned specifically includes: Based on the renewable energy output data, the expected renewable energy output data is fitted, and the renewable energy output model is obtained through scenario perturbation correction. Based on the renewable energy output model and the traditional energy output model, a power source operation model is constructed; wherein, the traditional energy output model is constructed based on the traditional energy output data; Based on the historical operating data, various electricity consumption behavior types are identified, and a model is created for each type to obtain a corresponding electricity demand model; wherein, the electricity consumption behavior types include industrial electricity consumption, interruptible loads, and transferable loads; Based on the electricity demand model corresponding to each type of electricity consumption behavior, a load-side operation model is constructed.

3. The two-level optimization planning method for power systems according to claim 1, characterized in that, The step of performing scenario clustering on renewable energy output data from the historical operational data to construct a weather-driven scenario set specifically includes: Based on each data column in the renewable energy output data, feature extraction is performed on the renewable energy output data to construct multiple joint feature vectors; wherein, the data columns include wind speed, irradiance, temperature, load factor, wind power output rate, photovoltaic power output rate, and fault indication status; Based on the multiple joint feature vectors, the scene is iteratively clustered based on unsupervised learning. During clustering, Mahalanobis distance is used to characterize the similarity between the joint feature vectors, and the scene clusters are obtained based on the similarity. The iteration stops when the number of scene clusters meets a preset threshold. Based on the scene clusters obtained in the last iteration, construct a weather-driven scene set.

4. The two-level optimization planning method for power systems according to claim 1, characterized in that, Based on the power supply-side operation model and the load-side operation model, and with the goal of minimizing the equipment's total lifecycle cost, a higher-level planning model is constructed, and the higher-level planning constraints of the higher-level planning model are determined based on the planning scale. Specifically, this includes: Based on the power supply side operation model and the load side operation model, determine the equipment coordination index; The decision-making process uses the commissioning of power equipment and the expansion of grid equipment as decision variables to construct various decision costs for each piece of equipment in the power system, and integrates them to obtain the equipment's total life cycle cost; wherein, the decision costs include commissioning costs, operation and maintenance costs, fuel costs, and equipment residual costs; With the goal of minimizing the total lifecycle cost of equipment, an upper-level planning model is constructed by combining the total lifecycle cost of equipment and the equipment synergy index. Based on the historical operating data, the annual maximum predicted load and flexible adjustment margin of the power system are predicted. Based on the planning scale, and in combination with the annual maximum predicted load and the flexible adjustment margin, the upper-level planning constraints of the upper-level planning model are determined.

5. The two-level optimization planning method for power systems according to claim 1, characterized in that, The step of constructing a lower-level scheduling model based on the weather-driven scenario set, with the goal of minimizing the cycle operating cost of the power system, and determining the lower-level scheduling constraints of the lower-level scheduling model based on the scheduling scale, specifically includes: Based on the historical operating data, the operating sub-costs for each cycle of the power system are determined and integrated to obtain the cycle operating cost of the power system; wherein, the cycle operating sub-costs include generation cost, network loss cost, and scenario penalty; wherein, the network loss cost is calculated based on the unit electricity price of nodes in the power system and the power flow loss of each node; the scenario penalty is obtained by combining the weather-driven scenario set; A lower-level scheduling model is constructed with the goal of minimizing the cycle operating cost of the power system. Based on the scheduling scale, the electricity price scheduling constraint of the unit electricity price of the node is determined, and the lower-level scheduling constraint of the lower-level scheduling model is obtained.

6. The two-level optimization planning method for power systems according to claim 1, characterized in that, The process of combining the upper-level planning model and the lower-level scheduling model into a two-level optimization model, and iteratively solving the two-level optimization model based on the Benders decomposition algorithm to obtain the planning scheme of the power system, specifically includes: The upper-level planning model and the lower-level scheduling model are combined into a two-layer optimization model; Based on the Benders decomposition algorithm, the two-level optimization model is iteratively solved. During the iteration, the initial planning scheme is solved by the upper-level planning model and passed to the lower-level scheduling model. The lower-level scheduling model simulates the initial planning scheme to update the cycle running cost and passes it to the upper-level planning model so that the upper-level planning model can optimize the initial planning scheme. The iteration is completed when the two-level optimization model converges to a preset threshold. The initial planning scheme obtained from the last iteration is used as the planning scheme for the power system.

7. A two-level optimization planning method for power systems according to any one of claims 1 to 6, characterized in that, The step of optimizing the power system according to the planning scheme specifically includes: Based on the planning scheme, simulation is performed using the historical operating data to obtain simulation operating data; Based on the simulation data, the planning scheme is validated. When the validation is successful, the power system is optimized according to the planning scheme.

8. A two-level optimization planning device for a power system, characterized in that, It includes a data construction and preparation module, an upper-level model construction module, a lower-level model construction module, and a model solving and optimization module; The data construction and preparation module is used to construct a power source side operation model and a load side operation model based on the historical operation data of the power system to be planned, and to perform scenario clustering on the renewable energy output data in the historical operation data to construct a weather-driven scenario set. The upper-level model construction module is used to construct an upper-level planning model based on the power supply side operation model and the load side operation model, with the goal of minimizing the equipment's total life cycle cost, and to determine the upper-level planning constraints of the upper-level planning model based on the planning scale. The lower-level model construction module is used to construct a lower-level scheduling model based on the weather-driven scenario set, with the goal of minimizing the cycle operation cost of the power system, and to determine the lower-level scheduling constraints of the lower-level scheduling model based on the scheduling scale. The model solving and optimization module is used to combine the upper-level planning model and the lower-level scheduling model into a two-level optimization model, iteratively solve the two-level optimization model based on the Benders decomposition algorithm to obtain the planning scheme of the power system, and optimize the power system according to the planning scheme.

9. A computer device, characterized in that, include: processor; Memory; A computer program stored in the memory and configured to be executed by the processor; When the processor executes the computer program, it implements a two-level optimization planning method for a power system as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to execute a two-level optimization planning method for a power system as described in any one of claims 1 to 7.

Images