Low-carbon power system optimization method and device, computer equipment and storage medium

Abstract

The invention relates to a low-carbon power system optimization method and device, computer equipment and a storage medium, and relates to the technical field of power system optimization. The method comprises the following steps: determining carbon emission factor historical data of a target area within a preset time; inputting the historical data of the carbon emission factors into a pre-configured carbon emission factor generation model to obtain a carbon emission factor baseline of the target area within a preset time; the carbon emission factor generation model is used for mapping the historical data of the carbon emission factors to a low-dimensional space and performing adversarial learning in the low-dimensional space to obtain an output result after fitting time sequence features corresponding to the historical data of the carbon emission factors; based on the carbon emission factor base line and a pre-configured power system time sequence simulation model, performing simulation on the target region to obtain a simulation result meeting a multi-target optimization condition; the multi-objective optimization condition at least comprises optimal network loss, optimal unit economic cost and optimal carbon emission.

Description

Technical Field

[0001] This application relates to the technical field of power system optimization, and particularly to a low-carbon power system optimization method, device, computer device, storage medium, and computer program product. Background Art

[0002] Currently, the power grids in various regions can plan the output of each unit through a low-carbon power system, and take into account aspects such as power generation cost, power generation amount, network loss, and carbon emission amount to achieve an optimal optimization result, thereby improving the economy and stability of the power grid. In related technologies, various pre-configured optimization models for low-carbon power systems can be used for optimization.

[0003] However, carbon emissions are mainly measured through carbon metering, that is, the carbon emissions of the regional power grid are evaluated by measuring the consumed energy and the corresponding carbon emission factors. Due to the intermittency and volatility of renewable energy, when the output of new energy units is insufficient to meet the current power demand, the energy storage system will provide power support, which will cause fluctuations in the carbon emission factors of the units. The optimization models of low-carbon power systems in related technologies do not consider the fluctuations of carbon emission factors, resulting in insufficient accuracy of the optimization results obtained when the low-carbon power system is optimized through the optimization model. Summary of the Invention

[0004] Based on this, it is necessary to provide a low-carbon power system optimization method, device, computer device, computer-readable storage medium, and computer program product that can improve the optimization accuracy for the above technical problems.

[0005] In a first aspect, this application provides a low-carbon power system method. The method includes:

[0006] Determine the historical data of carbon emission factors in a target region within a preset time; the historical data of carbon emission factors is determined based on the access power of new energy units and static carbon emission factors, and the access power of the energy storage system and static carbon emission factors;

[0007] Input the historical data of carbon emission factors into a pre-configured carbon emission factor generation model to obtain the baseline of carbon emission factors in the target region within a preset time; the carbon emission factor generation model is used to map the historical data of carbon emission factors to a low-dimensional space and perform adversarial learning in the low-dimensional space to obtain an output result after fitting the time series characteristics corresponding to the historical data of carbon emission factors;

[0008] Simulate the target area based on the carbon emission factor baseline and the pre-configured time-series simulation model of the power system to obtain simulation results that meet the multi-objective optimization conditions; the multi-objective optimization conditions at least include optimal network loss, optimal unit economic cost, and optimal carbon emissions; the simulation results include the access power of each new energy unit in the target area within a preset time.

[0009] In one embodiment, the carbon emission factor generation model is a time-series generative adversarial network; the time-series generative adversarial network includes a hidden Markov module, an embedding recovery module, a discriminator module, and a generator module;

[0010] The step of inputting the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline of the target area within a preset time includes:

[0011] Map the historical carbon emission factor data based on the embedding recovery module of the time-series generative adversarial network to obtain a low-dimensional output mapped to a low-dimensional space;

[0012] Based on the generator module of the time-series generative adversarial network, determine the output of the generation layer corresponding to the hidden space; the output of the generation layer is obtained by Gaussian sampling of the hidden space;

[0013] Based on the hidden Markov module of the time-series generative adversarial network, process the low-dimensional output and the output of the generation layer to obtain the first hidden Markov state parameters corresponding to the low-dimensional output and the second hidden Markov state parameters corresponding to the output of the generation layer;

[0014] Train the hidden Markov module, the embedding recovery module, the discriminator module, and the generator module based on the first hidden Markov state parameters, the second hidden Markov state parameters, and the low-dimensional output to obtain a trained time-series generative adversarial network;

[0015] Based on the generator module corresponding to the trained time-series generative adversarial network, determine the carbon emission factor baseline corresponding to the historical carbon emission factor data and determine it as the carbon emission factor baseline of the target area within a preset time.

[0016] In one embodiment, the step of training the hidden Markov module, the embedding recovery module, the discriminator module, and the generator module based on the first hidden Markov state parameters, the second hidden Markov state parameters, and the low-dimensional output to obtain a trained time-series generative adversarial network includes:

[0017] Based on the first hidden Markov state parameters and the second hidden Markov state parameters, determine the hidden Markov loss corresponding to the hidden Markov module;

[0018] Determine the embedding recovery loss corresponding to the embedding recovery module based on the predicted carbon emission factor baseline generated by the embedding recovery module and the historical carbon emission factor data; the predicted carbon emission factor baseline is obtained by the embedding recovery module processing the low-dimensional output and the second hidden Markov state parameter;

[0019] Process the low-dimensional output and the second hidden Markov state parameter through the discriminant module to obtain a discriminant result, and determine the discriminant loss corresponding to the discriminant module based on the discriminant result;

[0020] Determine the generation loss corresponding to the generation module based on the discriminant result and the predicted carbon emission factor baseline;

[0021] Based on the hidden Markov loss, the embedding recovery loss, the discriminant loss, and the generation loss, train the corresponding hidden Markov module, embedding recovery module, discriminant module, and generation module respectively to obtain a trained time series generative adversarial network.

[0022] In one embodiment, the power system time series simulation model includes a power simulation constraint sub-model and a multi-objective optimization sub-model;

[0023] Perform simulation on the target area based on the carbon emission factor baseline and a pre-configured power system time series simulation model to obtain a simulation result that meets the multi-objective optimization conditions, including:

[0024] Initialize the power simulation constraint sub-model corresponding to the target area; wherein, the power simulation constraint sub-model includes safety constraints, electrical nodes, and model variables; the electrical nodes include at least new energy units; the model variables include the access power, voltage, and phase angle of each electrical node in the power system;

[0025] Import the power simulation constraint sub-model into the multi-objective optimization model, and solve the power simulation constraint sub-model based on the network loss optimal objective function, unit economic cost optimal objective function, and carbon emission optimal objective function in the multi-objective optimization sub-model to obtain a simulation result that meets the multi-objective optimization conditions; the carbon emission optimal objective function is determined based on the carbon emission factor baseline.

[0026] In one embodiment, the safety constraints include power security adequacy constraints, generator output constraints, voltage magnitude constraints, power angle constraints, and line security constraints;

[0027] The initialization of the power simulation constraint sub-model corresponding to the target area includes:

[0028] Determine that the new energy units included in the target area are electrical nodes, and determine the access power, voltage, and phase angle corresponding to each electrical node as model variables; the access power includes active power and reactive power;

[0029] Configure the power security adequacy constraints and generator output constraints corresponding to the access power; the power security adequacy constraints are used to limit the value range of the remaining access power after the electrical node is used through the attached load;

[0030] Configure the voltage amplitude constraints corresponding to the voltage; the voltage amplitude constraints are used to limit the value range of the voltage;

[0031] Configure the power angle constraints corresponding to the phase angle; the power angle constraints are used to limit the value range of the phase angle;

[0032] Configure the line security constraints corresponding to the active power of the branch of the electrical node; the line security constraints are used to limit the value range of the remaining active power after the branch is used through the attached load.

[0033] In one embodiment, importing the power simulation constraint sub-model into the multi-objective optimization model, and based on the network loss optimal objective function, unit economic cost optimal objective function, and carbon emission optimal objective function in the multi-objective optimization sub-model, solving the power simulation constraint sub-model to obtain a simulation result that meets the multi-objective optimization conditions includes:

[0034] Based on the access power of the electrical node and the attached load of the electrical node, determine the network loss optimal objective function;

[0035] Based on the access power of the electrical node, determine the unit economic cost optimal objective function;

[0036] Based on the carbon emission factor baseline of the electrical node and the access power, determine the carbon emission optimal objective function;

[0037] Determine the differences between the network loss optimal objective function, the unit economic cost optimal objective function, and the carbon emission optimal objective function and their respective single-objective optimal solutions, and sum the weighted differences to obtain a multi-objective optimization function;

[0038] Based on the multi-objective optimization function, solve the power simulation constraint sub-model to obtain a simulation result that meets the multi-objective optimization conditions.

[0039] In one embodiment, the determining the historical carbon emission factor data of the target area within a preset time includes:

[0040] Obtain the first access power and the first static carbon emission factor of the new energy unit, as well as the second access power and the second static carbon emission factor of the energy storage system;

[0041] For each time point within a preset time, based on the first access power and the second access power, determine the first power ratio corresponding to the first access power and the second power ratio corresponding to the second access power;

[0042] Respectively, perform weighted summation of the first power ratio, the second power ratio with the corresponding first static carbon emission factor and the second static carbon emission factor to obtain the historical carbon emission factor data corresponding to the time point.

[0043] In a second aspect, the present application also provides a low-carbon power system optimization device. The device includes:

[0044] A data determination module, configured to determine the historical carbon emission factor data of a target area within a preset time; the historical carbon emission factor data is determined based on the access power and the static carbon emission factor of the new energy unit, as well as the access power and the static carbon emission factor of the energy storage system;

[0045] A baseline determination module, configured to input the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline of the target area within a preset time; the carbon emission factor generation model is used to map the historical carbon emission factor data to a low-dimensional space and perform adversarial learning in the low-dimensional space to obtain an output result after fitting the time series characteristics corresponding to the historical carbon emission factor data;

[0046] A simulation module, configured to perform simulation on the target area based on the carbon emission factor baseline and a pre-configured power system time series simulation model to obtain a simulation result that meets multi-objective optimization conditions; the multi-objective optimization conditions at least include optimal network loss, optimal unit economic cost, and optimal carbon emissions; the simulation result includes the access power of each new energy unit in the target area within a preset time.

[0047] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in the first aspect are implemented.

[0048] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0049] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program which, when executed by a processor, implements the steps of the method described in the first aspect.

[0050] The above-mentioned low-carbon power system optimization method, device, computer equipment, storage medium and computer program product can determine the historical data of carbon emission factors in a target area within a preset time, and the historical data of carbon emission factors is determined based on the access power and static carbon emission factors corresponding to new energy units and energy storage systems respectively. Then, the historical data of carbon emission factors is input into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline in the target area within the preset time. The carbon emission factor generation model can map the historical data of carbon emission factors to a low-dimensional space and perform adversarial learning in the low-dimensional space. After the learning is completed, an output result that fits the time series characteristics corresponding to the historical data of carbon emission factors is obtained. Finally, based on the carbon emission factor baseline and the pre-configured power system time series simulation model, a simulation is performed for the target area to obtain a simulation result that at least meets the multi-objective optimization conditions composed of optimal network loss, optimal unit economic cost, and optimal carbon emissions. Thus, the access power of each new energy unit in the target area within the preset time can be obtained. It can combine the access power and static carbon emission factors of new energy units and energy storage systems within the preset time, reduce the fluctuation of carbon emission factors when the output of new energy units is insufficient, improve the accuracy of determining the carbon emission factor baseline, and perform a simulation on the power system time series simulation model based on the carbon emission factor baseline and the multi-objective optimization conditions, taking into account the impact of the carbon emission factor baseline on the power system time series simulation model, thereby improving the accuracy of the simulation result. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0052] Figure 1 It is an application environment diagram of the low-carbon power system optimization method in an embodiment;

[0053] Figure 2 It is a flowchart of the low-carbon power system optimization method in an embodiment;

[0054] Figure 3 It is a flowchart of the steps to obtain the carbon emission factor baseline in an embodiment;

[0055] Figure 4Schematic diagram of the process of the low-carbon power system optimization method in another embodiment;

[0056] Figure 5 Schematic diagram of the network structure of the temporal generative adversarial network in one embodiment;

[0057] Figure 6 Schematic diagram of the optimization framework of the power system temporal simulation model in one embodiment;

[0058] Figure 7 Schematic diagram of the carbon emission factor baseline in one embodiment;

[0059] Figure 8 Schematic diagram of the power flow voltage and carbon emission factors in one embodiment;

[0060] Figure 9 Block diagram of the structure of the low-carbon power system optimization device in one embodiment;

[0061] Figure 10 Internal structure diagram of a computer device in one embodiment. Detailed implementation manners

[0062] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] The low-carbon power system optimization method provided by the embodiments of the present application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. The terminal 102 can obtain the access power and static carbon emission factors of each new energy unit in the target area within a preset time, as well as the access power and static carbon emission factors of the energy storage system, and send the above access power and static carbon emission factors to the server 104. The server can determine the historical carbon emission factor data of the target area within a preset time based on the access power and static carbon emission factors of the new energy unit and the energy storage system respectively. The server 104 can input the historical carbon emission factor data into a pre-configured carbon emission factor generation model to generate the carbon emission factor baseline of the target area within a preset time. The carbon emission factor generation model can perform adversarial learning based on the historical carbon emission factor data, so that the carbon emission factor generation model can generate the carbon emission factor baseline for the historical carbon emission factor data. The server 104 can simulate the power system of the target area based on the carbon emission factor baseline and the pre-configured power system time series simulation model, so as to obtain a simulation result that meets the optimal network loss, the optimal unit economic cost, and the optimal carbon emission, and obtain the access power of each new energy unit in the target area within a preset time.

[0064] Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, and tablet computers. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0065] In an exemplary embodiment, as Figure 2 shown, a carbon power system optimization method is provided. Taking the method applied to the Figure 1 server 104 in it as an example, it includes the following steps S202 to step S206. Among them:

[0066] Step S202, determine the historical carbon emission factor data of the target area within a preset time. The historical carbon emission factor data is determined based on the access power and static carbon emission factors of the new energy unit, and the access power and static carbon emission factors of the energy storage system.

[0067] Among them, the target area is a pre-configured geographical area, which may include multiple new energy units and energy storage systems. The target area may be, for example, a city, a district, a building, etc., and the embodiments of the present application do not make specific limitations. The preset time may be a time interval formed by two time points. For example, a time interval corresponding to one month, a time interval corresponding to one day, and a time interval corresponding to one year, etc., and the embodiments of the present application do not make specific limitations. In an exemplary embodiment, the server may determine the historical data of carbon emission factors in a certain city in each year.

[0068] The historical data of carbon emission factors includes the carbon emission factors actually recorded at each time point in the preset time in the target area. Since the new energy units in the power system have unstable power output, the energy storage system will output power when the new energy units are unstable, and the static carbon emission factor of the energy storage system is different from that of the new energy units. Therefore, at different time points, the actually recorded carbon emission factors are not the same. Based on this, the server can dynamically determine the historical data of carbon emission factors at different times according to the access power and static carbon emission factor of the new energy units in the preset time, and the access power and static carbon emission factor of the energy storage system.

[0069] It should be understood that the new energy unit is a set of devices that generate electric energy through new energy, such as the units corresponding to wind power generation and hydropower generation. The energy storage system may store the electric energy generated by various power generation means, and the energy storage system may also generate electric energy in real time through non-new energy methods and provide the stored or generated electric energy to the target area. For example, the energy storage system may provide electric energy to the power system of the target area through thermal power generation or energy storage devices. The access power may be the electric power provided by the new energy unit and the energy storage system to the power system in real time.

[0070] In an example, the server may obtain the new energy units and energy storage systems that provide electric energy to the target area, obtain the access power of the new energy units in the preset time, the corresponding static carbon emission factor of the new energy units, obtain the access power of the energy storage system in the preset time, and the corresponding static carbon emission factor of the energy storage system. Among them, the static carbon emission factor corresponding to the new energy unit is usually greater than the static carbon emission factor. The static carbon emission factor may be a fixed value and does not change with time. For example, the static carbon emission factor of each type of generator in the new energy unit is a first fixed value, and the static carbon emission factor corresponding to each type of generator or energy storage device in the energy storage system is a second fixed value. When the new energy unit and the energy storage system output power at the same time, the first fixed value and the second fixed value will exist at the same time, and the overall historical data of carbon emission factors in the preset time in the target area will be determined according to the proportion of their respective power outputs (the proportion of access power), the first fixed value, and the second fixed value.

[0071] Step S204: Input the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline for the target area within a preset time. The carbon emission factor generation model is used to map the historical carbon emission factor data to a low-dimensional space and perform adversarial learning in the low-dimensional space to obtain the output result after fitting the time series features corresponding to the historical carbon emission factor data.

[0072] Among them, the carbon emission factor generation model can be a pre-configured fitting model or prediction model. The carbon emission factor generation model can be based on the historical carbon emission factor data for fitting to obtain a carbon emission factor baseline that conforms to the time series features of the historical carbon emission factor data.

[0073] In an example, the server can input the historical carbon emission factor data into the carbon emission factor generation model. The carbon emission factor generation model can map the historical carbon emission factor data to a low-dimensional space and perform adversarial learning in the low-dimensional space. The server can continuously perform adversarial learning on the time series features in the low-dimensional space through the generator and discriminator. The generator can generate carbon emission factors, and the discriminator can discriminate the generated carbon emission factors according to the historical carbon emission factor data, so as to continuously optimize the generator and discriminator, thereby obtaining a trained carbon emission factor generation model, and through the trained generator, generate the carbon emission factor baseline for the target area within a preset time.

[0074] Step S206: Based on the carbon emission factor baseline and the pre-configured power system time series simulation model, perform simulation for the target area to obtain a simulation result that meets the multi-objective optimization conditions. The multi-objective optimization conditions at least include the optimal network loss, the optimal unit economic cost, and the optimal carbon emission; the simulation result includes the access power of each new energy unit in the target area within a preset time.

[0075] Among them, the power system time series simulation model can model the power system of the target area and perform simulation on the power system according to the multi-objective optimization conditions. By simulating the access power of each new energy unit in the target area within a preset time, the network loss, unit economic cost, and carbon emission in the power system can be determined, and in the three directions of the optimal network loss, the optimal unit economic cost, and the optimal carbon emission, continuously optimize the access power of each new energy unit, thereby obtaining a simulation result that meets the multi-objective optimization conditions.

[0076] Specifically, for a target area and within a preset time, the server can input the corresponding carbon emission factor baseline into a pre-configured power system time-series simulation model. The power system time-series simulation model can determine the access power of each new energy unit within the preset time through simulation, and can determine the carbon emissions of the power system based on the carbon emission factor baseline. Through continuous simulation, the server can obtain a simulation result that meets the multi-objective optimization conditions, and then obtain the access power of each new energy unit in the target area within the preset time.

[0077] In the above low-carbon power system optimization method, the historical carbon emission factor data of the target area within the preset time can be determined, and the historical carbon emission factor data is determined based on the access power and static carbon emission factors corresponding to the new energy unit and the energy storage system respectively, and the historical carbon emission factor data is input into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline of the target area within the preset time. The carbon emission factor generation model can map the historical carbon emission factor data to a low-dimensional space and perform adversarial learning in the low-dimensional space. After the learning is completed, the output result after fitting the time-series characteristics corresponding to the historical carbon emission factor data is obtained. Finally, based on the carbon emission factor baseline and the pre-configured power system time-series simulation model, simulation is performed for the target area to obtain a simulation result that at least meets the multi-objective optimization conditions composed of the optimal network loss, the optimal unit economic cost, and the optimal carbon emissions, so as to obtain the access power of each new energy unit in the target area within the preset time. It can combine the access power and static carbon emission factors of the new energy unit and the energy storage system within the preset time, reduce the fluctuation of the carbon emission factor when the output of the new energy unit is insufficient, improve the accuracy of determining the carbon emission factor baseline, and perform simulation on the power system time-series simulation model based on the carbon emission factor baseline and the multi-objective optimization conditions, taking into account the impact of the carbon emission factor baseline on the power system time-series simulation model, thereby improving the accuracy of the simulation result.

[0078] In an exemplary embodiment, as Figure 3 shown, the carbon emission factor generation model is a time-series generative adversarial network; the time-series generative adversarial network includes a hidden Markov module, an embedding recovery module, a discriminator module, and a generator module; the specific implementation process of the step "input the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline of the target area within the preset time" includes steps S302 to S310. Among them:

[0079] Step S302, map the historical carbon emission factor data based on the embedding recovery module of the time-series generative adversarial network to obtain a low-dimensional output mapped to the low-dimensional space.

[0080] Among them, the carbon emission factor generation model can be a time series generative adversarial network; the time series generative adversarial network includes a hidden Markov module, an embedding and recovery module, a discriminant module, and a generation module. The hidden Markov module can process the observed values through a statistical model of a Markov process with hidden unknown parameters to obtain corresponding hidden state parameters. The observed values can be the historical carbon emission factor data mapped to a low-dimensional space or the generated data generated by the generator in the generation module. The hidden state parameters can be the hidden Markov state parameters corresponding to the historical carbon emission factor data and the hidden Markov state parameters corresponding to the generated data.

[0081] The embedding and recovery module can respectively include an embedding layer and a recovery layer. The server can map the historical carbon emission factor data to a low-dimensional space through the embedding layer, and the server can map the time series features in the low-dimensional space back to the real space through the recovery layer, so as to obtain the fitted carbon emission factor baseline.

[0082] The discriminant module can include a discriminator, so as to compare the historical carbon emission factor data in the low-dimensional space with the generated data generated by the generator through the discriminator to determine the discriminant result of the generated data. The generation module can include a generator, and the generator can sample the random parameters in the hidden space to obtain the generated data.

[0083] Specifically, the server can input the historical carbon emission factor data into the input layer of the embedding and recovery module, send the data of the input layer to the embedding layer, and map the historical carbon emission factor data to a low-dimensional space through the embedding layer to obtain the low-dimensional output corresponding to the historical emission factor data.

[0084] Step S304, based on the generation module of the time series generative adversarial network, determine the output of the generation layer corresponding to the hidden space; the output of the generation layer is obtained by performing Gaussian sampling on the hidden space.

[0085] Among them, the hidden space generator converts the input noise into an intermediate representation space of observable data, and the hidden space can contain noise. The generation layer can perform Gaussian sampling on the noise in the hidden space to obtain noise data that conforms to the Gaussian distribution and use it as the output of the generation layer.

[0086] Step S306, based on the hidden Markov module of the time series generative adversarial network, process the low-dimensional output and the output of the generation layer to obtain the first hidden Markov state parameter corresponding to the low-dimensional output and the second hidden Markov state parameter corresponding to the output of the generation layer.

[0087] In one example, the server processes the low-dimensional output and the output of the generation layer through the hidden Markov layer in the hidden Markov module to determine the first hidden Markov state parameter and the second hidden Markov state parameter corresponding to the low-dimensional output and the output of the generation layer respectively.

[0088] Step S308: Based on the first Hidden Markov state parameter, the second Hidden Markov state parameter, and the low-dimensional output, train the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module to obtain a trained temporal generative adversarial network.

[0089] Specifically, the first Hidden Markov state parameter, the second Hidden Markov state parameter, and the low-dimensional output can be used as the parameters of the loss functions corresponding to the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module, so as to determine the loss functions corresponding to each module, and continuously update the loss values corresponding to the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module based on the first Hidden Markov state parameter, the second Hidden Markov state parameter, and the low-dimensional output in each round of training, thereby adjusting the internal parameters of each module, completing the training of the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module, and further training the temporal generative adversarial network. When the trained temporal generative adversarial network meets the training end condition, a trained temporal generative adversarial network is obtained.

[0090] Step S310: Based on the generation module corresponding to the trained temporal generative adversarial network, determine the carbon emission factor baseline corresponding to the historical carbon emission factor data, and determine it as the carbon emission factor baseline for the target area within a preset time.

[0091] Specifically, the server can call the generation module corresponding to the trained temporal generative adversarial network to generate the carbon emission factor baseline for the target area within a preset time. Since the generator of the generation module is continuously trained based on the historical carbon emission factor data, the carbon emission factor baseline generated by the generator conforms to the temporal characteristics of the historical carbon emission factor data.

[0092] In this embodiment, the temporal generative adversarial network is trained with the historical carbon emission factor data, and based on the historical carbon emission factor data, the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module in the temporal generative adversarial network are respectively trained through adversarial training to obtain a trained temporal generative adversarial network, so as to generate the corresponding carbon emission factor baseline through the trained generation module, which can improve the accuracy of the carbon emission factor baseline.

[0093] In an exemplary embodiment, the specific implementation process of the step "Based on the first Hidden Markov state parameter, the second Hidden Markov state parameter, and the low-dimensional output, train the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module to obtain a trained temporal generative adversarial network" includes:

[0094] Step S402: Based on the first Hidden Markov state parameter and the second Hidden Markov state parameter, determine the Hidden Markov loss corresponding to the Hidden Markov module.

[0095] Specifically, the server may determine the Hidden Markov loss based on the difference between the first Hidden Markov state parameter and the second Hidden Markov state parameter, as well as the differences between the first Hidden Markov state parameter and the second Hidden Markov state parameter at all time points before the current time point.

[0096] In one example, the server may generate a Hidden Markov loss function according to the difference between the first Hidden Markov state parameter and the second Hidden Markov state parameter, as well as the differences between the first Hidden Markov state parameter and the second Hidden Markov state parameter at all time points before the current time point, and determine the Hidden Markov loss based on the Hidden Markov loss function after obtaining the first Hidden Markov state parameter and the second Hidden Markov state parameter corresponding to each time point.

[0097] In one example, the Hidden Markov layer loss function may be expressed as:

[0098] ℒ s = 1 τ ∑ [ ( h * hmm − g * hmm ) 2 + ( h * − g * ) 2 ]

[0099] where is a constant, is the first Hidden Markov state parameter, is the second Hidden Markov state parameter, is the first Hidden Markov state parameter at all previous t - 1 time moments, is the second Hidden Markov state parameter at all previous t - 1 time moments.

[0100] Step S404: Determine the embedding recovery loss corresponding to the embedding recovery module based on the predicted carbon emission factor baseline generated by the embedding recovery module and the historical carbon emission factor data.

[0101] Among them, the predicted carbon emission factor baseline is obtained by the embedding recovery module processing the low - dimensional output and the second Hidden Markov state parameter.

[0102] Specifically, the server may process the low - dimensional output and the second Hidden Markov state parameter through the recovery layer of the embedding recovery module to obtain the predicted carbon emission factor baseline. The predicted carbon emission factor baseline is obtained by mapping the processed low - dimensional output and the second Hidden Markov state parameter back to the original space. Based on this, the server may construct an embedding recovery loss function according to the predicted carbon emission factor baseline and the historical carbon emission factor data, and calculate the loss value corresponding to the embedding recovery loss function.

[0103] In one example, the embedding recovery loss function may be expressed as:

[0104]

[0105] where is the historical data of carbon emission factors, is the predicted carbon emission factor baseline.

[0106] Step S406: The discrimination module processes the low-dimensional output and the second Hidden Markov state parameter to obtain a discrimination result, and determines the discrimination loss corresponding to the discrimination module based on the discrimination result.

[0107] Specifically, the server can use the discrimination module to discriminate the low-dimensional output and the second Hidden Markov state parameter to obtain a discrimination result, which can include true or false. For a true discrimination result, the first discrimination loss function can be determined; for a false discrimination result, the second discrimination loss function can be determined. The server can obtain the first discrimination loss corresponding to the discrimination module through the first discrimination loss function, and obtain the second discrimination loss corresponding to the discrimination module through the second discrimination loss function.

[0108] In one example, the first discrimination loss function and the second discrimination loss function can be expressed as follows:

[0109] ℒ d − r e a l = − 1 τ ∑ [ ones ( y real )log( y real ) + [ 1 − ones ( y real )]log( 1 − y real )] ℒ d − f a k e = − 1 τ ∑ [ zeros ( y fake )log( y fake ) + [ 1 − zeros ( y fake )]log( 1 − y fake )]

[0110] where ones() represents a matrix with a value of 1, zeros() represents a matrix with a value of 0, and y real represents that the discrimination result is true, and y fake represents that the discrimination result is false.

[0111] Step S408: Determine the generation loss corresponding to the generation module based on the discrimination result and the predicted carbon emission factor baseline.

[0112] Specifically, the server can determine the first generation loss function and the second generation loss function according to the discrimination result of the discrimination module and the predicted carbon emission factor baseline output by the recovery layer. The first generation loss function can determine the loss value of the generation layer, and the second generation loss function can determine the loss value between the generated data (predicted carbon emission factor baseline) and the initial data (historical carbon emission factor data).

[0113] In one example, the first generation loss function and the second generation loss function can be expressed as follows:

[0114] ℒ g 1 = − 1 τ ∑ [ ones ( y fake )log( y fake ) + [ 1 − ones ( y fake )]log( 1 − y fake )] ℒ g 2 = 1 τ ∑ [ μ ( x , x * ) + σ 2 ( x , x * ) ]

[0115] where ones() represents a matrix with a value of 1, zeros() represents a matrix with a value of 0, is the historical data of carbon emission factors, is the predicted carbon emission factor baseline. and respectively represent the mean and variance operators.

[0116] Step S410: Based on the Hidden Markov loss, embedding recovery loss, discriminative loss, and generative loss, train the corresponding Hidden Markov module, embedding recovery module, discriminative module, and generative module respectively to obtain a trained temporal generative adversarial network.

[0117] In one example, for one round of training, the server can perform Gaussian sampling on the hidden space through the generation layer of the generative module to obtain the generation layer output g. The server can determine the low-dimensional output h corresponding to the historical carbon emission factor data through the embedding layer in the embedding recovery module. Based on this, the server can input the generation layer output g and the low-dimensional output h into the Hidden Markov module to obtain the Hidden Markov state parameters corresponding to the generation layer output g and the low-dimensional output h respectively, which may include the first Hidden Markov state parameter , the second Hidden Markov state parameter , the first Hidden Markov state parameters at all previous t - 1 moments , the second Hidden Markov state parameters at all previous t - 1 moments . The server can map the low-dimensional output h and the second Hidden Markov state parameter through the recovery layer of the embedding recovery module to the original space to obtain the predicted carbon emission factor baseline .

[0118] Based on this, for the Hidden Markov module, the server can determine the Hidden Markov loss according to the Hidden Markov layer loss function. For the embedding recovery module, the server can determine the embedding recovery loss according to the embedding recovery loss function. For the discriminative module, the server can determine the first discriminative loss and the second discriminative loss according to the first discriminative loss function and the second discriminative loss function. For the generative module, the server can determine the first generative loss and the second generative loss according to the first generative loss function and the second generative loss function.

[0119] For the above Hidden Markov loss, embedding recovery loss, discriminative loss, and generative loss, the stochastic gradient descent method can be used to adjust the module parameters of the Hidden Markov module, the module parameters of the embedding recovery module, the module parameters of the discriminative module, and the module parameters of the generative module respectively. After the adjustment, the updated low-dimensional output is obtained again through the generation layer of the generative module. Then, repeat each round of training process described above, continuously update the module parameters of each module through the Hidden Markov loss, embedding recovery loss, discriminative loss, and generative loss, so that the Hidden Markov loss, embedding recovery loss, discriminative loss, and generative loss are reduced to the range of the loss value configured by the training end condition. After multiple rounds of training and when the preset training end condition is met, a trained temporal generative adversarial network is obtained.

[0120] In this embodiment, the module parameters of each module are updated by determining the Hidden Markov loss, the embedding recovery loss, the discriminant loss, and the generation loss, and the Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module are continuously updated through iteration, so as to train the time series generative adversarial network, thereby improving the training efficiency and the training convergence speed, and improving the fitting accuracy of the time series generative adversarial network.

[0121] In an exemplary embodiment, the power system time series simulation model includes a power simulation constraint sub-model and a multi-objective optimization sub-model; the specific implementation process of the step "simulating the target area based on the carbon emission factor baseline and the pre-configured power system time series simulation model to obtain a simulation result that meets the multi-objective optimization conditions" includes:

[0122] Step S502, initialize the power simulation constraint sub-model corresponding to the target area.

[0123] Among them, the power system time series simulation model includes a power simulation constraint sub-model and a multi-objective optimization sub-model; the power simulation constraint sub-model includes safety constraints, electrical nodes, and model variables; the electrical nodes at least include new energy units; the model variables include the access power, voltage, and phase angle of each electrical node in the power system.

[0124] Specifically, the server can configure the safety constraints, electrical nodes, and model variables corresponding to the target area. The server can determine the nodes corresponding to the new energy units and energy storage systems in the target area as electrical nodes. The server can configure the access power, voltage, and phase angle of each electrical node at each time point as model variables. The server can configure the safety constraints corresponding to the model variables for each model variable, so as to limit the value range of the model variables when simulating through the multi-objective optimization sub-model.

[0125] Step S504, import the power simulation constraint sub-model into the multi-objective optimization model, and solve the power simulation constraint sub-model based on the network loss optimal objective function, the unit economic cost optimal objective function, and the carbon emission optimal objective function in the multi-objective optimization sub-model to obtain a simulation result that meets the multi-objective optimization conditions; the carbon emission optimal objective function is determined based on the carbon emission factor baseline.

[0126] Specifically, the server can configure the optimal objective functions of network loss, unit economic cost, and carbon emission at each time point within a preset time. For example, the time point can be a moment. The server can simulate each electrical node and the model variables of the electrical node in the power simulation constraint sub-model through a multi-objective optimization model, and make the model variables satisfy the security constraints during the simulation, so as to obtain multiple sets of simulation results. Through the optimal objective function of network loss, the optimal objective function of unit economic cost, and the optimal objective function of carbon emission in the multi-objective optimization sub-model, the server can determine the final objective value corresponding to each set of simulation results and determine whether the final objective value meets the multi-objective optimization conditions. In one example, generally, minimizing the final objective value can be used as the multi-objective optimization condition, and it can be determined that when the obtained final objective value is less than the pre-configured minimum threshold value, the obtained simulation results meet the multi-objective optimization conditions, so as to obtain the simulation results that meet the multi-objective optimization conditions, that is, obtain the access power of each new energy unit in the target area within the preset time.

[0127] In this embodiment, by initializing the power simulation constraint sub-model corresponding to the target area, importing the power simulation constraint sub-model into the multi-objective optimization model, and solving the power simulation constraint sub-model based on the optimal objective function of network loss, the optimal objective function of unit economic cost, and the optimal objective function of carbon emission in the multi-objective optimization sub-model, the simulation results that meet the multi-objective optimization conditions can be obtained, which can comprehensively consider multiple objective functions to make the simulation results maximally meet the multi-objective optimization conditions, thereby improving the accuracy and rationality of the simulation results.

[0128] In an exemplary embodiment, the security constraints include power security adequacy constraints, generator output constraints, voltage magnitude constraints, power angle constraints, and line security constraints. The specific implementation process of the step "initializing the power simulation constraint sub-model corresponding to the target area" includes:

[0129] Step S602, determine the new energy units included in the target area as electrical nodes, and determine the access power, voltage, and phase angle corresponding to each electrical node as model variables.

[0130] Among them, the access power includes active power and reactive power; the active power refers to the power actually consumed and converted into electrical energy, which is the part that actually does work in the power system. The reactive power can be the power to establish an electric field or a magnetic field. This part of the power exchanges back and forth between the power source and energy storage elements (such as inductors, capacitors) and does not directly do work.

[0131] Specifically, the server can obtain the new energy units included in the target area and determine the new energy units as electrical nodes. The server can also determine the energy storage system corresponding to the target area as an electrical node. The server can configure the access power, voltage, and phase angle of the electrical node as model variables.

[0132] In one example, the model variables of the power simulation constraint sub-model can be expressed as:

[0133] { P i , t = P i , t G − P i , t D Q i , t = Q i , t G − Q i , t D P i , t = ∑ j V i , t V j , t [ G ij cos( δ i , t − δ j , t ) + B ij sin( δ i , t − δ j , t )] Q i , t = ∑ j V i , t V j , t [ G ij sin( δ i , t − δ j , t ) − B ij cos( δ i , t − δ j , t )]

[0134] The above formula is the model of the node injection power balance constraint. Among them, and respectively represent the active power and reactive power injected into the node, and respectively represent the active power and reactive power of the unit connected to the node, and respectively represent the active power and reactive power of the load mounted on the node. and respectively represent the voltage and phase angle of the unit, and respectively represent the conductance and susceptance of the branch i-j of the unit. N is the set of system nodes, and the node , at time .

[0135] Step S604, configure the power security adequacy constraint and generator output constraint corresponding to the access power.

[0136] Among them, the power security adequacy constraint is used to limit the value range of the remaining access power after the electrical node uses the mounted load.

[0137] In one example, at time t, the power security adequacy constraint is as follows:

[0138]

[0139] Among them, is the network loss (network power loss) of the power system at time t. represents the active power of the unit connected to the node, represents the active power of the load mounted on the node.

[0140] Step S606, configure the voltage amplitude constraint corresponding to the voltage.

[0141] Among them, the voltage amplitude constraint is used to limit the value range of the voltage.

[0142] Step S608, configure the power angle constraint corresponding to the phase angle.

[0143] Among them, the power angle constraint is used to limit the value range of the phase angle.

[0144] Step S610, configure the line security constraint corresponding to the active power of the branch of the electrical node.

[0145] Among them, the line safety constraint is used to limit the value range of the remaining active power of the branch after being used with the attached load.

[0146] In an example, the generator output constraint, voltage amplitude constraint, power angle constraint, and line safety constraint are as follows:

[0147]

[0148] Among them, the subscripts "min" and "max" respectively represent the minimum and maximum values of the corresponding electrical parameters. is the active power limit of the branch. 、 respectively represent the injected active power and reactive power of the node. 、 respectively represent the active and reactive power of the units connected to the node. 、 respectively represent the active and reactive power of the load attached to the node. 、 respectively represent the voltage and phase angle of the unit.

[0149] In this embodiment, by configuring the electrical nodes, model variables, and safety constraints in the power simulation constraint sub-model, the integrity and richness of the power simulation constraint sub-model can be improved.

[0150] In an exemplary embodiment, the specific implementation process of the step "import the power simulation constraint sub-model into the multi-objective optimization model, and solve the power simulation constraint sub-model based on the network loss optimal objective function, unit economic cost optimal objective function, and carbon emission optimal objective function in the multi-objective optimization sub-model to obtain a simulation result that meets the multi-objective optimization conditions" includes:

[0151] Based on the access power of the electrical node and the attached load of the electrical node, determine the network loss optimal objective function; based on the access power of the electrical node, determine the unit economic cost optimal objective function; based on the carbon emission factor baseline and access power of the electrical node, determine the carbon emission optimal objective function; determine the differences between the network loss optimal objective function, unit economic cost optimal objective function, and carbon emission optimal objective function and their respective single-objective optimal solutions, and sum the weighted differences to obtain the multi-objective optimization function; solve the power simulation constraint sub-model based on the multi-objective optimization function to obtain a simulation result that meets the multi-objective optimization conditions.

[0152] In an example, the network loss optimal objective function can be expressed as:

[0153]

[0154] Unit economic cost optimal objective function It can be expressed as:

[0155]

[0156] where a i , b i and r i respectively represent the quadratic term, linear term and constant term of the polynomial modeling of the economic cost of the unit output.

[0157] The optimal objective function of carbon emissions can be expressed as:

[0158]

[0159] In one example, the specific implementation process of the step "determine the differences between the optimal objective function of network loss, the optimal objective function of unit economic cost and the optimal objective function of carbon emissions and their respective corresponding single-objective optimal solutions, and weight and sum the differences to obtain a multi-objective optimization function" can be expressed as:

[0160] { min f t BCP = min λ loss , λ cost , λ CO 2 ] [ f t loss − f t ,min loss f t cost − f t ,min cost f t cost − f t ,min CO 2 ] s.t. λ loss + λ cost + λ CO 2 = 1

[0161] In the formula, , , are respectively the weight parameters of the optimal objective of network loss, the optimal objective of economy and the optimal objective of carbon emissions. The server subtracts the objective functions of the optimal network loss, the optimal economy and the optimal carbon emissions from their corresponding single-objective optimal solutions , , in turn, then linearly sums them, and configures the corresponding weights, so as to realize the overall optimization of the three groups of objective functions.

[0162] In this embodiment, by solving the power simulation constraint sub-model based on the multi-objective optimization function, the simulation results satisfying the multi-objective optimization conditions are obtained, which can improve the optimization efficiency, as well as the accuracy and rationality of the obtained simulation results.

[0163] In an exemplary embodiment, the specific implementation process of the step "determine the historical data of carbon emission factors in the target area within the preset time" includes:

[0164] Obtain the first access power and the first static carbon emission factor of the new energy unit, as well as the second access power and the second static carbon emission factor of the energy storage system; for each time point within a preset time, based on the first access power and the second access power, determine the first power ratio corresponding to the first access power and the second power ratio corresponding to the second access power; respectively perform weighted summation of the first power ratio, the second power ratio with the corresponding first static carbon emission factor and the second static carbon emission factor to obtain the historical carbon emission factor data corresponding to the time point.

[0165] In one example, the ideal carbon emission factor of the energy storage system-containing is defined as follows:

[0166]

[0167] Among them, at time t, for the new energy unit i, 、 respectively represent the access powers of the new energy unit and the energy storage system, that is, the first access power and the second access power, 、 respectively represent the first static carbon emission factor and the second static carbon emission factor of the new energy unit and the energy storage system. Thus, the historical carbon emission factor data of the new energy unit i at time t can be obtained . By calculating the carbon emission factors over the entire time series, a historical data set can be obtained, denoted as X, and its data feature space is denoted as F.

[0168] In this embodiment, by performing weighted summation of the first power ratio, the second power ratio with the corresponding first static carbon emission factor and the second static carbon emission factor to obtain the historical carbon emission factor data corresponding to the time point, the accuracy of the historical carbon emission factor data can be improved.

[0169] In an exemplary embodiment, the solution of this application embodiment includes two contents: (1) A time-series dynamic carbon emission factor generation method based on a generative framework; (2) Propose a low-carbon power system time-series multi-objective optimization modeling and framework based on carbon emission factors and compromise programming. The following is an introduction respectively:

[0170] (1) New energy has intermittent and volatile characteristics, and its carbon emission factors also have time-series volatility. Based on a time-series generative deep learning framework, integrating the dynamic characteristics of the energy storage system, capture the data intermittency and volatility characteristics of historical carbon emission factors to generate a carbon emission factor baseline.

[0171] (2)Based on the carbon emission factor baseline generated in (1), combined with the actual power grid parameters and the provincial power grid planning scheme, a multi-objective optimization framework for the power system based on the Basic Compromise Programming (BCP) is constructed, and the time-series simulation results of the low-carbon power system are obtained by combining the open-source simulation framework PYOMO and the Interior-Point Optimizer (IPOPT).

[0172] In this embodiment, by integrating the power system carbon emission factor baseline of the energy storage dynamic system and the multi-objective optimization simulation framework, it can be used to solve the problem of integrating economic - security - low-carbon multi-objective optimization in practical engineering, and promote the application of the electric-carbon coupling technology in energy enterprises.

[0173] As Figure 4 shown, the following combines a specific embodiment to describe in detail the specific execution process of the above low-carbon power system optimization method, including the following steps:

[0174] Step 1: Construct a historical dataset of the power system carbon emission factors integrating the energy storage system. Step 2: Build a training model for carbon emission factors based on a generative framework. Step 3: Complete the training and export the trained model and the carbon emission factor baseline; for example, export the trained model and the carbon emission factor baseline from the trained neural network model. Step 4: Construct a time-series simulation calculation model for the low-carbon power system and perform iterative solution; for example, based on the PYOMO open-source architecture, build a multi-objective time-series simulation calculation model for the power system based on the Basic Compromise Programming (BCP) and perform iterative solution. Step 5: Output the results of the optimal system output, economic cost, and carbon emissions.

[0175] Specifically, Step 1: Construct a historical dataset of the power system carbon emission factors integrating the energy storage system. The ideal carbon emission factor of the energy storage system is defined as follows:

[0176] (1)

[0177] Among them, at time t, for new energy unit i, 、 respectively represent the access powers of the new energy unit and the energy storage system, that is, the first access power and the second access power, 、 respectively represent the first static carbon emission factor and the second static carbon emission factor of the new energy unit and the energy storage system. Thus, the historical data of the carbon emission factor of new energy unit i at time t can be obtained. By calculating the carbon emission factors over the entire time series, a historical dataset can be obtained, denoted as X, and its data feature space is denoted as 。

[0178] Step 2: Build a carbon emission factor training model based on a generative framework. As Figure 5 shown, select the Hidden Markov Model-based Temporal Generative Adversarial Network (HMM-TSGAN) as the generative artificial intelligence framework, including: an embedding recovery module, a generation module, a discriminant module, and a hidden Markov module.

[0179] The embedding layer maps the original high-dimensional information to a low-dimensional space, improving the ability to learn core features while reducing the dimension. The recovery layer can restore the real sequence to the initial state space. The generation layer and the discriminant layer perform adversarial learning on Gaussian sampling in the latent space through a game, enabling the generation layer model to simulate and generate the distribution of the real space. A supervised learning model with the real distribution as the label is set in the supervision layer to fit the temporal features of the generation layer distribution and the real distribution in the low-dimensional space. For carbon emission factor data, the embedding recovery module model is:

[0180] (2)

[0181] where h and r are the models of the embedding layer and the recovery layer respectively, 、 are the corresponding parameters. is the low-dimensional output mapped by the embedding layer, is the output restored to the real space by the recovery layer. Assume that the low-dimensional space of the embedding layer is denoted as The mathematical form of model optimization is:

[0182] { ℒ h =  x : ℱ , h : ℋ | | x − x ^ | | 2 m i n ℒ h ⇐ ∇ [ ∂ ℒ h ∂ ( θ h + θ r ) ] (3)

[0183] where represents the loss function of the embedding-recovery pooling module, and regularization is achieved by the L2 norm. represents the update of the model parameters and using the stochastic gradient descent method. The instance of the generation layer is as follows:

[0184] (4)

[0185] where s, g, d, and 、 、 represent the instantiated hidden Markov layer, generation layer, discriminant layer, and their corresponding model parameters respectively. represents the Gaussian sampling z in the hidden space Z. g, 、 represent the corresponding outputs of the generation layer, hidden Markov layer, and recovery layer respectively, ​Denotes the output of the discriminative layer, Denotes the hidden Markov state parameters of the generation layer. The model is optimized as follows:

[0186] { ℒ g 1 =  z :  l o g ( 1 − y fake ) ℒ g 2 =  x , x * : ℱ [ μ ( x , x * ) + σ 2 ( x , x * ) ] (5)

[0187] m i n ( ℒ g 2 + ℒ g 1 ) ⇐ ∇ [ ∂ ( κ 1 ℒ g 1 + κ 2 ℒ g 2 ) ∂ ( θ g + θ r ) ] (6)

[0188] In the above formula, Denotes the loss of the generation layer, Denotes the joint loss of the generated data and the initial data, where And Denote the mean and variance operators respectively. Equation (6) gives the optimization solution process of the joint loss, where k1 and k2 are the hyperparameters of the joint loss.

[0189] The discriminative layer instance and the optimization process are as follows:

[0190] { y real = d ( h ; θ d ) ℒ d − r e a l = −  h : ℋ l o g ( y real ) ℒ d − f a k e =  z :  log( 1 − y fake ) min( ℒ d − r e a l + ℒ d − f a k e ) ⇐∇ [ ∂ ( ℒ d − r e a l + ℒ d − f a k e ) ∂ θ d ] (7)

[0191] In the formula, Denotes the output of the discriminative layer, And Denote the classification learning losses of the discriminator for different output labels respectively. For a data with time series distribution characteristics, the state at this moment should be jointly determined by the hidden states of the previous multiple moments, such as:

[0192] (8)

[0193] Among them, p(s t ) represents the probability of being in state s t at time t. Then p(s t-1 , s t-2 ,...) represents the probability of the hidden states of all previous moments. For the supervised learning model with parameters , assuming that at time t, the previous t - 1 moments are the Markov hidden state parameters denoted as , then the instantiation of the hidden Markov - supervised model is as follows:

[0194] (9)

[0195] In the above formula Denotes the output of the hidden Markov layer. Through the supervised learning of the above formula, the generation layer can use the low - dimensional mapping of the initial data as the learning label, thereby improving the learning efficiency of the generation model for the real distribution.

[0196] Step 3: Complete the training and export the trained model and carbon emission factors .

[0197] Step 4-1: Construct a low-carbon power system timing simulation model and perform iterative solution. Let N be the set of system nodes, and the nodes ,time . Power flow constraints:

[0198] { P i , t = P i , t G − P i , t D Q i , t = Q i , t G − Q i , t D P i , t = ∑ j V i , t V j , t [ G ij cos( δ i , t − δ j , t ) + B ij sin( δ i , t − δ j , t )] Q i , t = ∑ j V i , t V j , t [ G ij sin( δ i , t − δ j , t ) − B ij cos( δ i , t − δ j , t )] (10)

[0199] The above formula is the model of node injection power balance constraint. Among them, , They represent the injected active power and reactive power of the node respectively, , They represent the active power and reactive power of the units connected to the node respectively. , They represent the active power and reactive power of the node's load respectively. , Respectively represent the voltage and phase angle of the unit, , They represent the conductance and susceptance of the branch ij of the unit respectively.

[0200] At time t, the power security and sufficiency constraints are as follows:

[0201] (11)

[0202] in, is the grid loss (network loss) of the power system at time t. Indicates the active power of the unit connected to the node, Indicates the active power of the load mounted on the node.

[0203] Generator output constraints, voltage amplitude constraints, power angle constraints and line safety constraints are as follows:

[0204] (12)

[0205] The subscripts "min" and "max" represent the minimum and maximum values ​​of the corresponding electrical parameters, respectively. is the active power limit of the branch. , They represent the injected active power and reactive power of the node respectively, , They represent the active and reactive power of the units connected to the node respectively. , They represent the active and reactive power of the node mounted load respectively. , represent the voltage and phase angle of the unit respectively.

[0206] The optimal objective function of network loss at time t can be expressed as:

[0207] (13)

[0208] The optimal objective function of the economic cost of the unit at time t can be expressed as:

[0209] (14)

[0210] where a i , b i and r i represent the quadratic term, linear term and constant term of the polynomial modeling of the economic cost of the unit output respectively.

[0211] The optimal objective function of carbon emissions at time t can be expressed as:

[0212] (15)

[0213] As can be seen from step (1), the carbon emission factor of the unit will fluctuate with time, and the carbon emission factor here is taken from the learning results of steps 2 - 3. Note that, combining equations (10) - (12) will establish a non-convex and non-linear power simulation constraint model, and cooperating with the single-objective functions (13) - (15) can construct a multi-objective optimization model. As Figure 6 shown, the power time-series multi-objective optimization model based on the BCP theory is as follows:

[0214] { min f t BCP = min λ loss , λ cost , λ CO 2 ] [ f t loss − f t ,min loss f t cost − f t ,min cost f t cost − f t ,min CO 2 ] s.t. λ loss + λ cost + λ CO 2 = 1 (16)

[0215] In the formula, , , are the weight parameters of the optimal objective of network loss, economic optimum and carbon emission optimum respectively. The server will subtract the optimal objective functions of network loss optimum, economic optimum and carbon emission optimum from their corresponding single-objective optimal solutions , , in turn, then linearly sum them, and configure the corresponding weights, so as to realize the overall optimization of the three groups of objective functions.

[0216] Step 4-2: The multi-objective optimization framework of the low-carbon power system based on PYOMO-IPOPT, as Figure 6 shown.

[0217] Step 5: Output the results of the optimal system output, economic cost, and carbon emissions.

[0218] In an exemplary embodiment, the power grid framework parameter information and carbon emission factor modeling are set as follows:

[0219] Table 1: Static carbon emission factors of each type of unit

[0220]

[0221] The simulation example selects the basic power grid framework of a certain provincial power grid in the extreme operating mode in summer of a certain year. The test system has 466 nodes, including 42 generator nodes and 79 load nodes, including 406 branches and 298 groups of transformer branches.

[0222] As Figure 7 shown, the abscissa represents the time from 00:00 to 23:00. The gray dotted line represents the actual change curve of the carbon emission factors of wind power and photovoltaic power throughout 2022. The blue solid line is the daily average change curve of the carbon emission factors of wind power and photovoltaic power. The thick green solid line is the time series feature distribution curve generated by the HMM-TSGAN model. To quantitatively compare the advantages and disadvantages of the two groups of curves, the standard deviation of the carbon emission factors within 24 hours of the day is taken as the volatility index in the example. The daily average baseline volatility is 1.7 kg∙CO2 / MW∙h, the daily average volatility of historical data is 5.6 kg∙CO2 / MW∙h, and the volatility of the generated feature curve is 3.2 kg∙CO2 / MW∙h. The generated feature curve covers both the stability and volatility of historical data and is a balanced result considering historical extreme values and other data.

[0223] To better combine with the actual business needs of the power grid, the planning scheme of Province A in 2025 is selected as the basic business scenario for the case study simulation in this paper. According to the planning scheme, the maximum load in 2025 is 10,500 MW, the total installed capacity of traditional coal and gas units is 10,400 MW, the installed capacity of nuclear power is 2,600 MW, the installed capacity of hydropower units is 930 MW, the installed capacity of pumped storage is 600 MW, the installed capacity of wind power is 4,000 MW, the installed capacity of photovoltaic systems is 6,500 MW, and the energy storage system is 1,200 MW. In addition, coal-fired power, gas-fired power, and nuclear power are respectively configured with 15% - 20% capacity reserve, and hydropower and pumped storage are respectively configured with 30% - 40% capacity reserve. According to the local hydrological characteristics, January to May is set as the dry season, and the rest of the time is the wet season. In the wet season, the adjustment parameters of the available capacity of hydropower and pumped storage are 85% - 95%, and in the dry season, they are 15% - 40%. The network loss boundary of the system is set at 200 MW. For the quadratic term and linear term of the economic cost model of traditional units, [0.05, 0.15] (yuan / MW²) and [20, 60] (yuan / MW) are taken respectively. For new energy sources such as wind power and photovoltaic, they are estimated based on the one-time installation cost and a 20-year service life, and the constant term interval is [57,000, 62,000] (yuan).

[0224] As Figure 8 shown, the power flow, voltage distribution, and carbon potential distribution diagrams of each voltage level of the power grid in Province A at a certain moment in winter 2025 under low wind and no light conditions are given. Among them, the carbon potential calculation is derived from reference [1], which will not be elaborated here in detail. At this moment, the system load is 6,444 MW, the system unit output is 6,586 MW, and the system network loss is 135.06 MW. Figure 8 Part (a) in it details the optimal branch power flow and voltage results of the 500 kV grid. To ensure readability, the power flow calculation results of voltage levels below 200 kV are not shown in detail in the text. Corresponding to (a), (b) gives the detailed carbon potential distribution of the 500 kV nodes and the carbon potential histograms of other voltage level grids. The total node carbon potential of the test system at the current moment is 76,724.66 kg∙CO₂ / MW∙h, among which the sum of the node carbon potentials of the 500 kV voltage level is 1,049.92 kg∙CO₂ / MW∙h, the sum of the node carbon potentials of the 220 kV voltage level is 29,331.38 kg∙CO₂ / MW∙h, and the sum of the carbon potentials of voltage levels below 110 kV is 46,298.36 kg∙CO₂ / MW∙h.

[0225] Table 2: Seasonal Comparison of Multi-objective Optimization Results

[0226]

[0227] Table 2 presents a comparison of the BCP method with the results of the unoptimized case study. Based on the BCP model, the optimal annual carbon dioxide equivalent emissions result is 18.5746 million tons, the optimal annual power generation economic cost is 6.417 billion yuan, and the minimum network loss is 109.59 MW. The average carbon emissions throughout the period are reduced by 23.96%, and the economic efficiency is improved by 14.25%.

[0228] In this embodiment, a time-series carbon emission factor baseline of the power system can be generated to optimize the quality of the basic carbon emission data of the new power system. In this embodiment, by proposing a time-series low-carbon simulation method for the power system of BCP and an open-source simulation framework, and combining with the actual planning scheme, the transformation path of the future power system can be accurately evaluated, and the economic, safety, and carbon emission indicators can be quantified, with strong engineering feasibility.

[0229] It should be understood that although the steps in the flowcharts involved in the above embodiments are sequentially shown according to the indications of the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same moment, but can be executed at different moments. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or alternately with at least a part of the steps or stages in other steps or other steps.

[0230] Based on the same inventive concept, an embodiment of the present application also provides a low-carbon power system optimization device for implementing the above-mentioned low-carbon power system optimization method. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the low-carbon power system optimization device provided below can refer to the limitations on the low-carbon power system optimization method in the above text, and will not be repeated here.

[0231] In an exemplary embodiment, as Figure 9 shown, a low-carbon power system optimization device 900 is provided, including: a data determination module 901, a baseline determination module 902, and a simulation module 903, where:

[0232] The data determination module 901 is configured to determine the historical data of carbon emission factors in the target area within a preset time; the historical data of carbon emission factors is determined based on the access power of new energy units and static carbon emission factors, and the access power of energy storage systems and static carbon emission factors;

[0233] A baseline determination module 902 is configured to input historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain a carbon emission factor baseline for a target region within a preset time period. The carbon emission factor generation model is used to map the historical carbon emission factor data into a low-dimensional space and perform adversarial learning in the low-dimensional space to obtain an output result after fitting the temporal features corresponding to the historical carbon emission factor data.

[0234] A simulation module 903 is configured to perform a simulation for the target region based on the carbon emission factor baseline and a pre-configured power system temporal simulation model to obtain a simulation result that meets multi-objective optimization conditions. The multi-objective optimization conditions at least include optimal network loss, optimal unit economic cost, and optimal carbon emissions. The simulation result includes the access power of each new energy unit in the target region within the preset time period.

[0235] Further, the carbon emission factor generation model is a temporal generative adversarial network. The temporal generative adversarial network includes a hidden Markov module, an embedding recovery module, a discriminator module, and a generator module. The baseline determination module 902 is specifically configured to: map the historical carbon emission factor data based on the embedding recovery module of the temporal generative adversarial network to obtain a low-dimensional output mapped into the low-dimensional space; determine the output of the generation layer corresponding to the hidden space based on the generator module of the temporal generative adversarial network. The output of the generation layer is obtained by performing Gaussian sampling on the hidden space; process the low-dimensional output and the output of the generation layer based on the hidden Markov module of the temporal generative adversarial network to obtain the first hidden Markov state parameters corresponding to the low-dimensional output and the second hidden Markov state parameters corresponding to the output of the generation layer; train the hidden Markov module, the embedding recovery module, the discriminator module, and the generator module based on the first hidden Markov state parameters, the second hidden Markov state parameters, and the low-dimensional output to obtain a trained temporal generative adversarial network; determine the carbon emission factor baseline corresponding to the historical carbon emission factor data based on the generator module corresponding to the trained temporal generative adversarial network, and determine it as the carbon emission factor baseline for the target region within the preset time period.

[0236] Further, the baseline determination module 902 is specifically further configured to: determine the hidden Markov loss corresponding to the hidden Markov module based on the first hidden Markov state parameter and the second hidden Markov state parameter; determine the embedding recovery loss corresponding to the embedding recovery module based on the predicted carbon emission factor baseline generated by the embedding recovery module and the historical carbon emission factor data; the predicted carbon emission factor baseline is obtained by the embedding recovery module processing the low-dimensional output and the second hidden Markov state parameter; the discriminant module processes the low-dimensional output and the second hidden Markov state parameter to obtain a discriminant result, and determines the discriminant loss corresponding to the discriminant module based on the discriminant result; determine the generation loss corresponding to the generation module based on the discriminant result and the predicted carbon emission factor baseline; respectively train the corresponding hidden Markov module, embedding recovery module, discriminant module, and generation module based on the hidden Markov loss, embedding recovery loss, discriminant loss, and generation loss to obtain a trained time series generative adversarial network.

[0237] Further, the power system time series simulation model includes a power simulation constraint sub-model and a multi-objective optimization sub-model. The simulation module 903 is specifically configured to: initialize the power simulation constraint sub-model corresponding to the target area; wherein, the power simulation constraint sub-model includes safety constraints, electrical nodes, and model variables; the electrical nodes include at least new energy units; the model variables include the access power, voltage, and phase angle of each electrical node in the power system; import the power simulation constraint sub-model into the multi-objective optimization model, and solve the power simulation constraint sub-model based on the network loss optimal objective function, unit economic cost optimal objective function, and carbon emission optimal objective function in the multi-objective optimization sub-model to obtain a simulation result that meets the multi-objective optimization conditions; the carbon emission optimal objective function is determined based on the carbon emission factor baseline.

[0238] Further, the safety constraints include power safety adequacy constraints, generator output constraints, voltage magnitude constraints, power angle constraints, and line safety constraints. The simulation module 903 is specifically further configured to: determine the new energy units included in the target area as electrical nodes, and determine the access power, voltage, and phase angle corresponding to each electrical node as model variables; the access power includes active power and reactive power; configure the power safety adequacy constraints and generator output constraints corresponding to the access power; the power safety adequacy constraints are used to limit the value range of the remaining access power after the electrical node is used with the attached load; configure the voltage magnitude constraints corresponding to the voltage; the voltage magnitude constraints are used to limit the value range of the voltage; configure the power angle constraints corresponding to the phase angle; the power angle constraints are used to limit the value range of the phase angle; configure the line safety constraints corresponding to the active power of the branch of the electrical node; the line safety constraints are used to limit the value range of the remaining active power after the branch is used with the attached load.

[0239] Further, the simulation module 903 is specifically further configured to: determine an optimal objective function for network loss based on the access power of the electrical node and the load mounted on the electrical node; determine an optimal objective function for the economic cost of the unit based on the access power of the electrical node; determine an optimal objective function for carbon emissions based on the carbon emission factor baseline of the electrical node and the access power; determine the differences between the optimal objective function for network loss, the optimal objective function for the economic cost of the unit, and the optimal objective function for carbon emissions and their respective corresponding single-objective optimal solutions, and sum the differences after weighting to obtain a multi-objective optimization function; solve the power simulation constraint sub-model based on the multi-objective optimization function to obtain a simulation result that meets the multi-objective optimization conditions.

[0240] Further, the data determination module 901 is specifically configured to: obtain the first access power and the first static carbon emission factor of the new energy unit, and the second access power and the second static carbon emission factor of the energy storage system; for each time point within a preset time period, determine the first power ratio corresponding to the first access power and the second power ratio corresponding to the second access power based on the first access power and the second access power; respectively sum the first power ratio, the second power ratio, the corresponding first static carbon emission factor, and the second static carbon emission factor after weighting to obtain the historical carbon emission factor data corresponding to the time point.

[0241] Each module in the above low-carbon power system optimization device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above respective modules.

[0242] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 10 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store historical carbon emission factor data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a low-carbon power system optimization method.

[0243] Those skilled in the art can understand that Figure 10 The structure shown in Figure 10 is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component arrangement.

[0244] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0245] In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0246] In an embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0247] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include Read-Only Memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0248] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0249] The above-described embodiments only represent several implementation manners of this application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of this application. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application should be subject to the appended claims.

Claims

1. A method for optimizing a low-carbon power system, characterized in that, The method includes: Determining historical carbon emission factor data of a target area within a preset time; the historical carbon emission factor data is determined based on the access power and static carbon emission factors of new energy units, as well as the access power and static carbon emission factors of an energy storage system; Inputting the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain a carbon emission factor baseline of the target area within the preset time; the carbon emission factor generation model is used to map the historical carbon emission factor data to a low-dimensional space and perform adversarial learning in the low-dimensional space to obtain an output result after fitting the time series characteristics corresponding to the historical carbon emission factor data; Performing simulation on the target area based on the carbon emission factor baseline and a pre-configured power system time series simulation model to obtain a simulation result that meets multi-objective optimization conditions; the multi-objective optimization conditions at least include optimal network loss, optimal unit economic cost, and optimal carbon emissions; the simulation result includes the access power of each new energy unit in the target area within the preset time.

2. The method according to claim 1, wherein The carbon emission factor generation model is a time series generative adversarial network; the time series generative adversarial network includes a hidden Markov module, an embedding recovery module, a discriminant module, and a generation module; The step of inputting the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain a carbon emission factor baseline of the target area within the preset time includes: Mapping the historical carbon emission factor data based on the embedding recovery module of the time series generative adversarial network to obtain a low-dimensional output mapped to the low-dimensional space; Based on the generation module of the time series generative adversarial network, determining the output of the generation layer corresponding to the hidden space; the output of the generation layer is obtained by performing Gaussian sampling on the hidden space; Based on the hidden Markov module of the time series generative adversarial network, processing the low-dimensional output and the output of the generation layer to obtain a first hidden Markov state parameter corresponding to the low-dimensional output and a second hidden Markov state parameter corresponding to the output of the generation layer; Training the hidden Markov module, the embedding recovery module, the discriminant module, and the generation module based on the first hidden Markov state parameter, the second hidden Markov state parameter, and the low-dimensional output to obtain a trained time series generative adversarial network; Based on the generation module corresponding to the trained time series generative adversarial network, determining the carbon emission factor baseline corresponding to the historical carbon emission factor data and determining it as the carbon emission factor baseline of the target area within the preset time.

3. The method according to claim 2, wherein The step of training the hidden Markov module, the embedding recovery module, the discriminant module, and the generation module based on the first hidden Markov state parameter, the second hidden Markov state parameter, and the low-dimensional output to obtain a trained time series generative adversarial network includes: Determining the hidden Markov loss corresponding to the hidden Markov module based on the first hidden Markov state parameter and the second hidden Markov state parameter; Based on the predicted carbon emission factor baseline generated by the embedding recovery module and the historical carbon emission factor data, determine the embedding recovery loss corresponding to the embedding recovery module; the predicted carbon emission factor baseline is obtained by the embedding recovery module processing the low-dimensional output and the second Hidden Markov state parameter; Process the low-dimensional output and the second Hidden Markov state parameter through the discriminant module to obtain a discriminant result, and determine the discriminant loss corresponding to the discriminant module based on the discriminant result; Based on the discriminant result and the predicted carbon emission factor baseline, determine the generation loss corresponding to the generation module; Based on the Hidden Markov loss, the embedding recovery loss, the discriminant loss, and the generation loss, train the corresponding Hidden Markov module, the embedding recovery module, the discriminant module, and the generation module respectively to obtain a trained time series generative adversarial network.

4. The method according to claim 1, wherein The power system time series simulation model includes a power simulation constraint sub-model and a multi-objective optimization sub-model; Perform simulation on the target area based on the carbon emission factor baseline and the pre-configured power system time series simulation model to obtain a simulation result that meets the multi-objective optimization conditions, including: Initialize the power simulation constraint sub-model corresponding to the target area; wherein, the power simulation constraint sub-model includes safety constraints, electrical nodes, and model variables; the electrical nodes at least include new energy units; the model variables include the access power, voltage, and phase angle of each electrical node in the power system; Import the power simulation constraint sub-model into the multi-objective optimization model, and solve the power simulation constraint sub-model based on the network loss optimal objective function, the unit economic cost optimal objective function, and the carbon emission optimal objective function in the multi-objective optimization sub-model to obtain a simulation result that meets the multi-objective optimization conditions; the carbon emission optimal objective function is determined based on the carbon emission factor baseline.

5. The method according to claim 4, characterized in that, The safety constraints include power security adequacy constraints, generator output constraints, voltage magnitude constraints, power angle constraints, and line security constraints; The initialization of the power simulation constraint sub-model corresponding to the target area includes: Determine the new energy units included in the target area as electrical nodes, and determine the access power, voltage, and phase angle corresponding to each electrical node as model variables; the access power includes active power and reactive power; Configure the power security adequacy constraint and the generator output constraint corresponding to the access power; the power security adequacy constraint is used to limit the value range of the remaining access power after the electrical node is used with the attached load; Configure the voltage magnitude constraint corresponding to the voltage; the voltage magnitude constraint is used to limit the value range of the voltage; Configure the power angle constraint corresponding to the phase angle; the power angle constraint is used to limit the value range of the phase angle; Configure the line security constraint corresponding to the active power of the branch of the electrical node; the line security constraint is used to limit the value range of the remaining active power after the branch is used with the attached load.

6. The method according to claim 4, wherein Importing the power simulation constraint sub-model into the multi-objective optimization model, and solving the power simulation constraint sub-model based on the network loss optimal objective function, the unit economic cost optimal objective function, and the carbon emission optimal objective function in the multi-objective optimization sub-model to obtain a simulation result that meets the multi-objective optimization conditions, including: Determining the network loss optimal objective function based on the access power of the electrical node and the load carried by the electrical node; Determining the unit economic cost optimal objective function based on the access power of the electrical node; Determining the carbon emission optimal objective function based on the carbon emission factor baseline of the electrical node and the access power; Determining the differences between the network loss optimal objective function, the unit economic cost optimal objective function, and the carbon emission optimal objective function and their respective single-objective optimal solutions, and weighted summing the differences to obtain a multi-objective optimization function; Solving the power simulation constraint sub-model based on the multi-objective optimization function to obtain a simulation result that meets the multi-objective optimization conditions.

7. The method according to claim 1, characterized in that, The determining of the historical carbon emission factor data of the target area within a preset time includes: Obtaining the first access power and the first static carbon emission factor of the new energy unit, and the second access power and the second static carbon emission factor of the energy storage system; For each time point within the preset time, determining the first power ratio corresponding to the first access power and the second power ratio corresponding to the second access power based on the first access power and the second access power; Weighted summing the first power ratio, the second power ratio with the corresponding first static carbon emission factor and the second static carbon emission factor respectively to obtain the historical carbon emission factor data corresponding to the time point.

8. An optimization device for a low-carbon power system, characterized in that, The device includes: A data determination module, configured to determine the historical carbon emission factor data of the target area within a preset time; the historical carbon emission factor data is determined based on the access power and the static carbon emission factor of the new energy unit, and the access power and the static carbon emission factor of the energy storage system; A baseline determination module, configured to input the historical carbon emission factor data into a pre-configured carbon emission factor generation model to obtain the carbon emission factor baseline of the target area within a preset time; the carbon emission factor generation model is used to map the historical carbon emission factor data to a low-dimensional space and perform adversarial learning in the low-dimensional space to obtain an output result after fitting the time series characteristics corresponding to the historical carbon emission factor data; A simulation module, configured to perform simulation on the target area based on the carbon emission factor baseline and a pre-configured power system time series simulation model to obtain a simulation result that meets the multi-objective optimization conditions; the multi-objective optimization conditions at least include optimal network loss, optimal unit economic cost, and optimal carbon emissions; the simulation result includes the access power of each new energy unit in the target area within a preset time.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 7.

Images