Source-load joint scene construction method suitable for low-carbon dispatching of electric vehicle charging station

Abstract

The invention discloses a source-load joint scene construction method suitable for low-carbon dispatching of an electric vehicle charging station. The method comprises the following steps: constructing a carbon quota income-based electric vehicle charging station low-carbon scheduling model and a target function thereof; backtracking and solving the electric vehicle charging station low-carbon scheduling model based on carbon quota income by using historical scheduling data, generating a carbon quota sample, and calculating sensitivity gradients of different input variables to a carbon quota function to obtain a carbon sensitivity matrix; constructing a diffusion model comprising a forward diffusion process and a reverse denoising process; and constructing a diffusion model of carbon perception modulation, and generating a source-load joint scene sequence with a carbon perception structure by using the trained diffusion model of carbon perception modulation. The method can explicitly model a combined disturbance path of photovoltaic and electric vehicle loads, introduces a carbon sensitivity structure to carry out multi-level modulation, effectively improves the adaptability of a generated scene to carbon quota key variable disturbance, and can obtain a better scene sequence.

Description

TECHNICAL FIELD

[0001] The application relates to the field of low-carbon scheduling of electric vehicle charging stations and artificial intelligence scene generation modeling, and particularly relates to a source-load joint scene construction method suitable for low-carbon scheduling of electric vehicle charging stations. BACKGROUND

[0002] Under the background of the promotion of the "double carbon" goal and the rapid development of distributed energy, electric vehicle charging stations, as an important carrier connecting the transportation energy and the power system, play an increasingly important role in improving the flexibility of the power grid and realizing low-carbon operation. With the wide access of clean energy such as photovoltaic to the charging station, the output uncertainty and the random charging and discharging behavior of electric vehicles jointly cause the system source and load fluctuation to intensify, which brings challenges to the low-carbon scheduling of the charging station. In order to guarantee the stability and economy of scheduling, constructing a scene set that can accurately depict the randomness of source and load is a key prerequisite for low-carbon scheduling optimization.

[0003] Traditional scene generation methods mainly fall into three categories: one is the method based on sampling and statistical inference, such as Monte Carlo sampling and kernel density estimation, but it is difficult to accurately model the complex correlation between photovoltaic and electric vehicle load; the second is the parametric model method, such as Gaussian mixture model, which can handle certain nonlinear structure, but has limited adaptability to high-dimensional non-stationary time series; the third is the deep learning method based on generative model, such as variational autoencoder (VAE), generative adversarial network (GAN), Gaussian mixture model (GMM) and denoising diffusion probability model (DDPM), which has made certain progress in single-variable uncertainty modeling.

[0004] However, there are two core problems in the source-load scene generation for the low-carbon scheduling goal: first, most existing methods model photovoltaic output and electric vehicle load independently, ignoring the coupling structure of the two at the statistical and dynamic levels, and it is difficult to reflect the true joint disturbance characteristics; second, the low-carbon benefit of electric vehicles is mainly realized through the carbon quota mechanism, and the sensitivity of scheduling benefits to photovoltaic and load disturbance is significant, but existing methods generally lack modeling of the response structure between disturbance path and carbon benefit, which limits the adaptability and benefit of the generated scene in low-carbon scheduling. Therefore, it is urgent to propose a new scene construction method that can capture the joint uncertainty of source and load and the carbon-sensitive structure, so as to provide high-quality and strong-adaptive scene support for the low-carbon scheduling of electric vehicle charging stations, and improve the economy, stability and carbon reduction level of the system. SUMMARY

[0005] In view of the above deficiencies in the prior art, the source-load joint scene construction method suitable for low-carbon scheduling of electric vehicle charging stations is provided.

[0006] The source-load joint scene construction method suitable for low-carbon scheduling of electric vehicle charging stations comprises the following steps: S1. Considering the operating characteristics of distributed photovoltaic power output, electric vehicle charging load and energy storage system, construct a low-carbon scheduling model for electric vehicle charging stations based on carbon quota revenue. With the goal of maximizing carbon quota revenue, and taking into account the electricity purchase and sale cost and energy storage operation cost, construct the objective function of the low-carbon scheduling model for electric vehicle charging stations. S2. Based on the objective function of the low-carbon scheduling model of electric vehicle charging stations, the low-carbon scheduling model of electric vehicle charging stations based on carbon quota revenue is solved by backtracking using historical scheduling data, carbon quota samples are generated, and the carbon quotas obtained by electric vehicles are calculated using different input variables to obtain the sensitivity gradient of different carbon quotas in order to obtain the carbon sensitivity matrix. S3. Construct a diffusion model that includes the forward diffusion process and the reverse denoising process; S4. Embed the carbon sensitivity matrix into the noise perturbation intensity, network structure feature channels, and loss function weights of the diffusion model to construct a carbon-sensing modulation diffusion model. Use the trained carbon-sensing modulation diffusion model to generate a source-load joint scene sequence with a carbon-sensing structure.

[0007] Furthermore, the objective function of the low-carbon scheduling model for electric vehicle charging stations is expressed as: , , ,

[0008] in: To take the sign of the maximum value, For joint scenarios The probability, For a set of joint scenarios, , , Joint scenarios Carbon trading revenue, electricity trading costs, and energy storage charging and discharging costs are all factors to consider. The price per unit of carbon credit for electric vehicles. For joint scenarios Down Carbon allowances earned by electric vehicles during certain time periods For the set of scheduling periods, for Electricity purchase cost during different time periods for Electricity sales cost during different time periods The charging and discharging cost of energy storage units, For scheduling intervals, For joint scenarios Down The amount of electricity purchased from the upper-level power grid during a given time period. For joint scenarios Down The amount of electricity sold to the upper-level power grid during a given time period. For energy storage In joint scenarios Down Charging power during the period A collection of energy storage devices in a charging station. For energy storage In joint scenarios Down Discharge power during the period This refers to the energy storage charging and discharging efficiency.

[0009] Furthermore, the carbon allowances earned by electric vehicles are calculated and expressed as follows:

[0010] in: for Carbon allowances earned by electric vehicles during certain time periods for The first time period with vehicle-to-grid interaction function The charging power of an electric vehicle The largest vehicle number in the set of controllable electric vehicles with vehicle-to-everything (V2X) connectivity. for Electric vehicle charging power during the time period This refers to the distance an electric vehicle can travel per unit of electricity. This refers to the carbon emissions per unit distance traveled by a gasoline-powered vehicle. for The amount of electricity purchased from the upper-level power grid during a given time period. for Actual grid-connected photovoltaic power during the period The proportion of thermal power in the electricity purchased by the power grid. This refers to the carbon emissions per unit power of thermal power units.

[0011] Furthermore, the operational constraints of the low-carbon scheduling model for electric vehicle charging stations include energy storage constraints, photovoltaic operation constraints, electricity trading constraints, controllable electric vehicle constraints, and power balance constraints.

[0012] Furthermore, the energy storage operation constraints are expressed as:

[0013] in:, For energy storage In joint scenarios Down Charging power during the period for the energy storage in the joint scenario under the discharging power of the period, the maximum charging and discharging power allowed for the energy storage, a set of joint scenarios, a set of energy storage devices in the charging station, a set of scheduling periods, for the energy storage in the joint scenario under the state of charge of the period, for the energy storage in the joint scenario under the state of charge of the period, the charging and discharging efficiency of the energy storage, a scheduling interval period, the capacity of the energy storage, , the minimum state of charge and the maximum state of charge allowed for the energy storage respectively; a photovoltaic operation constraint, denoted as:

[0014] wherein: the actual photovoltaic on-grid power of the period in the joint scenario under the photovoltaic output of the period in the joint scenario under the curtailed photovoltaic power of the period in the joint scenario under a power trading constraint, denoted as:

[0015] wherein: the power purchased from the upper-level power grid by the period in the joint scenario under the power sold to the upper-level power grid by the period in the joint scenario under the maximum value of the interaction power; a set of controllable electric vehicles with vehicle-to-grid interaction function is , and the constraint of the controllable electric vehicle is denoted as: ​​​​​

[0016] wherein: is the joint scenario is the lower is the charging power of the i-th electric vehicle with vehicle-to-grid function in the time period, is the joint scenario is the lower is the discharging power of the i-th electric vehicle with vehicle-to-grid function in the time period, is the maximum charging and discharging power of the electric vehicle, is the set of grid-joining time of the i-th electric vehicle with vehicle-to-grid function, is the joint scenario is the lower is the state of charge in the time period, is the joint scenario is the lower is the state of charge in the time period, is the joint scenario is the lower is the state of charge in the time period, is the battery capacity of the i-th electric vehicle with vehicle-to-grid function, , are the minimum and maximum state of charge of the i-th electric vehicle with vehicle-to-grid function, respectively, is the joint scenario is the lower is the state of charge in the time period, is the grid-leaving time of the i-th electric vehicle with vehicle-to-grid function, is the state of charge requirement of the i-th electric vehicle with vehicle-to-grid function at the user-set grid-leaving time; power balance constraint, denoted as: . further, the carbon sensitivity matrix, denoted as: ,

[0017] , , ,

[0018] wherein: is the sensitivity of the electric vehicle to the carbon quota, is the carbon quota, is the is the charging power of the electric vehicle in the time period, is the mileage of the electric vehicle per unit of electricity, is the carbon emission of the gasoline vehicle per unit of mileage, is​​ Solar power output during different time periods For joint scenarios Down The amount of light discarded during a given period The proportion of thermal power in the electricity purchased by the power grid. for The amount of electricity purchased from the upper-level power grid during a given time period. Carbon emissions per unit power of thermal power units. The sensitivity of photovoltaics to carbon quotas, for The first time period with vehicle-to-grid interaction function The charging power of an electric vehicle This is the carbon sensitivity matrix. The sensitivity of electric vehicles to carbon allowances in time period 1. The sensitivity of electric vehicles to carbon allowances in two time periods. For electric vehicles The sensitivity of time period to carbon quotas The sensitivity of photovoltaics to carbon quotas in time period 1. The sensitivity of photovoltaics to carbon quotas in two time periods. For photovoltaics The sensitivity of time period to carbon quotas 2× T The set of real matrices, This represents the maximum time period within the scheduling period.

[0019] Furthermore, a diffusion model including the forward diffusion process and the reverse denoising process is constructed, the specific process being as follows: Based on the historical electric vehicle charging power and photovoltaic output, a joint time series sample is constructed for historical days, represented as follows: , ,

[0020] in: For history day Joint time series samples, For history day Electric vehicle charging power, For history day Photovoltaic power output, 2× T The set of real matrices, The longest time period in the scheduling period. For history day middle Electric vehicle charging power during the time period For history day middle The photovoltaic output during the period is: The joint input data is constructed based on the joint time series samples of historical days and is represented as follows: , in: For combined input data, The number of historical days; Based on a Markov process with a standard Gaussian distribution, let the predefined noise intensity be... And the signal preservation coefficient is determined based on the predefined noise intensity, expressed as:

[0021] in: For the first The signal retention coefficient of the step, The cumulative signal retention coefficient, Indicates from arrive The product of two products; Based on the joint input data and signal retention coefficients, the forward diffusion process is constructed and expressed as:

[0022] in: Forward diffusion in the first The conditional distribution of the steps, This represents the intermediate state of the combined input data during the diffusion process. The mean is covariance is Gaussian distribution at point The probability density value at that location. It is the identity matrix; A parameterized loss function for a deep neural network is constructed. Based on the parameterized loss function, the joint noise term at each step is predicted using the parameterized deep neural network. This process gradually denoises the data, restoring the intermediate state of the joint input data during the diffusion process to the true joint sample. This sample is then used as input for the low-carbon scheduling problem to construct an inverse denoising process.

[0023] The beneficial effects of this invention are as follows: The present application discloses a source-load joint scheduling method for low-carbon scheduling of an electric vehicle charging station, and specifically discloses a source-load joint scene construction method for low-carbon scheduling of an electric vehicle charging station. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A source-load joint scene construction method for low-carbon scheduling of an electric vehicle charging station; Figure 2 A schematic diagram of the influence of a source-load disturbance path on a carbon quota; Figure 3 A comparison diagram of photovoltaic output generation results and true values in a source-load scene; Figure 4 A comparison diagram of electric vehicle charging load generation results and true values in a source-load scene; Figure 5 A comparison diagram of mutual information distribution of scenes generated by different methods; Figure 6 A comparison of virtual power plant low-carbon scheduling strategy benefits in scenes generated by different methods. DETAILED DESCRIPTION

[0025] The specific embodiments of the present application are described below to facilitate understanding of the present application by those skilled in the art, but it should be clear that the present application is not limited to the scope of the specific embodiments, and for those skilled in the art, it is obvious that various changes are within the spirit and scope of the present application as defined and limited by the appended claims, and all applications utilizing the concept of the present application are within the scope of protection.

[0026] As shown in Figure 1 The source-load joint scene construction method for low-carbon scheduling of an electric vehicle charging station comprises steps S1-S4, and specifically as follows: S1. Considering the operating characteristics of distributed photovoltaic power output, electric vehicle charging load and energy storage system, construct a low-carbon scheduling model for electric vehicle charging stations based on carbon quota revenue. With the goal of maximizing carbon quota revenue, and taking into account the electricity purchase and sale cost and energy storage operation cost, construct the objective function of the low-carbon scheduling model for electric vehicle charging stations.

[0027] In an optional embodiment of the present invention, the objective function of the low-carbon scheduling model for electric vehicle charging stations is expressed as: , , ,

[0028] in: To take the sign of the maximum value, For joint scenarios The probability, For a set of joint scenarios, , , Joint scenarios Carbon trading revenue, electricity trading costs, and energy storage charging and discharging costs are all factors to consider. The price per unit of carbon credit for electric vehicles. For joint scenarios Down Carbon allowances earned by electric vehicles during certain time periods For the set of scheduling periods, for Electricity purchase cost during different time periods for Electricity sales cost during different time periods The charging and discharging cost of energy storage units, For scheduling intervals, For joint scenarios Down The amount of electricity purchased from the upper-level power grid during a given time period. For joint scenarios Down The amount of electricity sold to the upper-level power grid during a given time period. For energy storage In joint scenarios Down Charging power during the period A collection of energy storage devices in a charging station. For energy storage In joint scenarios Down Discharge power during the period This refers to the energy storage charging and discharging efficiency.

[0029] The carbon quota obtained by the electric vehicle is calculated by the present application, which is expressed as:

[0030] Wherein: is the carbon quota obtained by the electric vehicle in the time period, is the charging power of the i-th electric vehicle with vehicle-to-grid interaction function in the time period, is the maximum vehicle number in the set of controllable electric vehicles with vehicle-to-grid interaction function, is the charging power of the electric vehicle in the time period, is the mileage that the electric vehicle can travel per unit of electricity, is the carbon emission per unit of mileage of the fuel vehicle, is the actual photovoltaic on-grid power in the time period, is the proportion of thermal power in the grid purchase power, is

[0031] the carbon emission per unit power of the thermal power unit.

[0032] The operation constraints of the low-carbon scheduling model of the electric vehicle charging station include energy storage constraints, photovoltaic operation constraints, power trading constraints, controllable electric vehicle constraints and power balance constraints.

[0033] Wherein: is the charging power of the energy storage in the joint scenario in the time period, is the discharging power of the energy storage in the joint scenario in the time period, is the maximum charging and discharging power allowed by the energy storage, is the set of joint scenarios, is the set of energy storage devices in the charging station, is the set of scheduling time periods, is the state of charge of the energy storage in the joint scenario in the time period, is the state of charge of the energy storage in the joint scenario state of charge of the period, for energy storage charging and discharging efficiency, for dispatch interval period, for energy storage capacity, , for energy storage allowed minimum and maximum state of charge.

[0034] photovoltaic operation constraints, denoted as:

[0035] wherein: for joint scenario under actual photovoltaic on-grid power of the period, for joint scenario under photovoltaic output of the period, for joint scenario under curtailed photovoltaic power of the period.

[0036] power trading constraints, denoted as:

[0037] wherein: for joint scenario under power purchased from the upper-level power grid of the period, for joint scenario under power sold to the upper-level power grid of the period, for maximum interactive power.

[0038] a set of controllable electric vehicles with vehicle-to-grid interaction function is denoted as controllable electric vehicle constraints, denoted as:

[0039] wherein: for joint scenario under charging power of the period of the th electric vehicle with vehicle-to-grid interaction function, for joint scenario under discharging power of the period of the th electric vehicle with vehicle-to-grid interaction function, for maximum charging and discharging power of the electric vehicle, for the th electric vehicle with vehicle-to-grid interaction function.A set of electric vehicle network access times. For joint scenarios Down State of charge during a period of time For joint scenarios Down State of charge during a period of time For the first vehicle with vehicle-to-everything (V2X) connectivity function The battery capacity of an electric vehicle, , These are the first ones with vehicle-to-everything (V2X) connectivity features. The minimum and maximum state of charge (SOC) of a battery for an electric vehicle. For joint scenarios Down State of charge during a period of time For the first vehicle-to-everything (V2X) ... The departure time of each electric vehicle For the first vehicle-to-everything (V2X) ... Users of electric vehicles set their state of charge requirements when leaving the vehicle.

[0040] Power balance constraints are expressed as: .

[0041] S2. Based on the objective function of the low-carbon scheduling model of electric vehicle charging stations, the low-carbon scheduling model of electric vehicle charging stations based on carbon quota revenue is solved by backtracking using historical scheduling data, carbon quota samples are generated, and the carbon quotas obtained by electric vehicles are calculated using different input variables to obtain the sensitivity gradient of different carbon quotas, so as to obtain the carbon sensitivity matrix.

[0042] In an optional embodiment of the present invention, the carbon sensitivity matrix is ​​represented as follows: , ,

[0043] in: The sensitivity of electric vehicles to carbon quotas, For carbon quotas, for Electric vehicle charging power during the time period This refers to the distance an electric vehicle can travel per unit of electricity. This refers to the carbon emissions per unit distance traveled by a gasoline-powered vehicle. for Solar power output during different time periods For joint scenarios Down The amount of light discarded during a given period The proportion of thermal power in the electricity purchased by the power grid. for The amount of electricity purchased from the upper-level power grid during a given time period. Carbon emissions per unit power of thermal power units. The sensitivity of photovoltaics to carbon quotas, for The first time period with vehicle-to-grid interaction function The charging power of an electric vehicle This is the carbon sensitivity matrix. The sensitivity of electric vehicles to carbon allowances in time period 1. The sensitivity of electric vehicles to carbon allowances in two time periods. For electric vehicles The sensitivity of time period to carbon quotas The sensitivity of photovoltaics to carbon quotas in time period 1. The sensitivity of photovoltaics to carbon quotas in two time periods. For photovoltaics The sensitivity of time period to carbon quotas 2× T The set of real matrices, This represents the maximum time period within the scheduling period.

[0044] This indicates that photovoltaic deviation leads to carbon quota deviation, but its impact has a diminishing marginal effect. The impact of the photovoltaic-electric vehicle charging load disturbance path on carbon quotas, such as... Figure 2 As shown.

[0045] S3. Construct a diffusion model that includes the forward diffusion process and the reverse denoising process.

[0046] In an optional embodiment of the present invention, the diffusion model constructed by the present invention includes a forward diffusion process and a reverse denoising and reconstruction process to achieve source-load joint scene modeling. The specific process of constructing a diffusion model including a forward diffusion process and a reverse denoising process is as follows: Based on historical data of electric vehicle charging power and photovoltaic output, a joint time-series sample is constructed for historical days, represented as follows: , ,

[0047] in: For history day Joint time series samples, For history day Electric vehicle charging power, For history day Photovoltaic power output, 2× T The set of real matrices, The longest time period in the scheduling period. For history day middle Electric vehicle charging power during the time period For history day middle The photovoltaic output during the period is: The joint input data is constructed based on the joint time series samples of historical days and is represented as follows: , in: For combined input data, The number of historical days; Based on a Markov process with a standard Gaussian distribution, let the predefined noise intensity be... And the signal preservation coefficient is determined based on the predefined noise intensity, expressed as:

[0048] in: For the first The signal retention coefficient of the step, The cumulative signal retention coefficient, Indicates from arrive The product of two products; Based on the joint input data and signal retention coefficients, the forward diffusion process is constructed and expressed as:

[0049] in: Forward diffusion in the first The conditional distribution of the steps, This represents the intermediate state of the combined input data during the diffusion process. The mean is The covariance is Gaussian distribution at point The probability density value at that location, It is an identity matrix.

[0050] Specifically, the aforementioned Gaussian distribution is a multivariate Gaussian distribution, wherein, For the intermediate state of the joint samples during the diffusion process, the covariance matrix adopts a simplified isotropic form. That is, noise is added independently and identically distributed in each variable dimension, but its pre-calculation coefficients are... Ensure that the electric vehicle and photovoltaics maintain the coupling relationship of the initial structure along the diffusion path.

[0051] The loss function of the parameterized deep neural network is constructed, the loss function of the parameterized deep neural network is constructed, the parameterized deep neural network is used to predict each step of the joint noise item, the joint input data is gradually denoised in the diffusion process, and the intermediate state of the joint input data in the diffusion process is restored to the real joint sample as the input of the low-carbon scheduling problem to construct the reverse denoising process.

[0052] Specifically, the reverse denoising process is a process of gradually denoising from random noise to restore the real joint sample, and a parameterized deep neural network is trained , which predicts each step of the joint noise item . In a specific implementation, a one-dimensional U-Net network structure is used in this paper, which encodes and decodes two-dimensional time series data and has the ability to capture cross-variable dependencies between photovoltaic and electric vehicles.

[0053] In the reverse denoising process, starting from the initial Gaussian noise , the iteration is gradually performed to , and the iteration expression is:

[0054] , wherein: is a parameterized deep neural network, is an independent standard Gaussian noise introduced in the reverse sampling step .

[0055] Finally, the joint sample is generated as the input of the low-carbon scheduling problem.

[0056] The loss function of the parameterized deep neural network is constructed, and the basic goal is to minimize the difference between the predicted noise and the real noise, and the expression of the loss function of the parameterized deep neural network is:

[0057] , wherein: is the loss of the parameterized deep neural network, is the joint expectation of the data sample, time step and noise, is the square of the Euclidean norm. S4, embed the carbon sensitivity matrix into the noise disturbance intensity, network structure feature channel and loss function weight of the diffusion model respectively, construct the carbon perception modulated diffusion model, and generate the source-load joint scene sequence with carbon perception structure by using the trained carbon perception modulated diffusion model.

[0058] In an optional embodiment of the present invention, the carbon sensitivity matrix is ​​embedded into the noise perturbation intensity, network structure feature channels, and loss function weights of the diffusion model, thereby improving the adaptability between the model generation scenario and the actual low-carbon scheduling decision.

[0059] Considering the source-load variable difference sensitive structure, through the carbon sensitivity matrix The noise perturbation intensity of each dimension at each time point is modulated to embed the carbon sensitivity matrix into the diffusion model, and is expressed as: , exist In the process of introducing noise into the forward diffusion prediction process in the network, it is necessary to enhance the expressive ability of the electric vehicle and photovoltaic joint structure. This can be achieved by introducing a cross-variable structure-aware weighting mechanism into the one-dimensional U-Net, assuming an intermediate feature tensor. ,in This refers to the batch size.

[0060] The channel embedding is obtained by averaging each variable channel along the time dimension, and is represented as follows:

[0061] in: For the first Individual samples, channels ,time eigencomponents, Indicates electric vehicle, It refers to photovoltaics.

[0062] The structure-sensing weight vector is constructed based on the carbon sensitivity matrix and is expressed as follows:

[0063] in: This is the structure-aware weight vector.

[0064] Weight modulation is applied to the channel embedding based on the structure-aware weight vector, as follows:

[0065] in: For channel embedding after weight modulation, For channel embedding, This is an element-wise product.

[0066] The original features are modulated using channel embedding with applied weights to complete the embedding of the carbon sensitivity matrix into the network structure feature channels of the diffusion model, as shown below:

[0067] in: The first characteristic modulation Individual samples, channels ,time eigencomponents.

[0068] The modulated features As The intermediate network representation continues to participate in subsequent residual connections and reverse denoising processes. This invention guides the information flow between channels by embedding the carbon sensitivity matrix into the network structure feature channels of the diffusion model, enabling the denoising network to learn the cooperative relationships across variable structures and improve its ability to model joint perturbations.

[0069] During the training phase, the loss function of the parameterized deep neural network is structurally adjusted to a carbon quota-sensitive weighted loss, so that the carbon sensitivity matrix is ​​embedded into the weights of the loss function of the diffusion model, as shown below:

[0070] in: Denoising loss weighted by carbon sensitivity, The joint expectation of data samples, time steps, and noise. It is the square of the Euclidean norm.

[0071] This invention enhances the model's learning of sensitive structures by embedding the carbon sensitivity matrix into the weights of the loss function of the diffusion model. This multi-layered structured modulation enables the constructed carbon-sensing modulation diffusion model to have a higher ability to capture sensitive structures, and ultimately, it can be used to generate source-load joint scene sequences with carbon-sensing structures.

[0072] Figure 3 and Figure 4 The comparison between the source-load scenarios generated by the proposed method and the actual values ​​is shown. The scenarios generated by the proposed method can cover the actual values ​​well, and the average value of the generated scenarios is basically consistent with the actual scenarios. For photovoltaic scenarios with regular output, the average value of the generated scenarios can basically match the actual scenarios, and for electric vehicle charging loads with large fluctuations, this method also shows good performance.

[0073] Figure 5 The mutual information (MI) distribution of 1000 source-load joint scenarios generated by our method and other comparative methods is shown, with the MI values ​​of real scenarios marked by red dashed lines. The mutual information distribution of the scenarios generated by our method is the most concentrated, the median is close to the real value, and the volatility is the smallest compared to other schemes.

[0074] Figure 6The benefits of the low-carbon scheduling strategy of the electric vehicle charging station under the proposed method and other comparative methods are compared. By using the generated multiple groups of scenarios as the input of the optimization problem, the optimal value gap, the benefit standard deviation, and the minimum operating benefit under the scenarios generated by different methods are shown.

[0075] Those skilled in the art will understand that the embodiments described herein are for the purpose of helping the reader understand the principles of the present application and should be understood as not limiting the scope of protection of the present application to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the scope of protection of the present application.

Claims

1. A source-load joint scenario construction method suitable for low-carbon scheduling of an electric vehicle charging station, characterized in that, The method comprises the following steps: S1, considering the distributed photovoltaic output, electric vehicle charging load and the operating characteristics of the energy storage system, a low-carbon scheduling model of the electric vehicle charging station based on carbon quota income is constructed, and a target function of the low-carbon scheduling model of the electric vehicle charging station is constructed with the maximization of carbon quota income as the target, and the power purchase and sale cost and the energy storage operation cost are taken into account; S2, based on the target function of the low-carbon scheduling model of the electric vehicle charging station, the low-carbon scheduling model of the electric vehicle charging station based on carbon quota income is solved by using historical scheduling data backtracking to generate carbon quota samples, and the carbon quota obtained by the electric vehicle is calculated by using different input variables to obtain the sensitivity gradient of different carbon quotas, so as to obtain a carbon sensitivity matrix; S3, a diffusion model including a forward diffusion process and a reverse denoising process is constructed; S4, the carbon sensitivity matrix is embedded into the noise disturbance intensity, the network structure feature channel and the loss function weight of the diffusion model respectively to construct a carbon-aware modulation diffusion model, and a source-load joint scene sequence with carbon-aware structure is generated by using the trained carbon-aware modulation diffusion model. 2.The source-load joint scenario construction method for low-carbon scheduling of electric vehicle charging stations according to claim 1, characterized in that, The target function of the low-carbon scheduling model of the electric vehicle charging station is represented as: ， ， ， in: To determine the sign of the maximum value, For joint scenarios The probability, For a set of joint scenarios, , , Joint scenarios Carbon trading revenue, electricity trading costs, and energy storage charging and discharging costs are all factors to consider. The price per unit of carbon credit for electric vehicles. For joint scenarios Down Carbon allowances earned by electric vehicles during certain time periods For the set of scheduling periods, for Electricity purchase cost during different time periods for Electricity sales cost during different time periods The charging and discharging cost of energy storage units, For scheduling intervals, For joint scenarios Down The amount of electricity purchased from the upper-level power grid during a given time period. For joint scenarios Down The amount of electricity sold to the upper-level power grid during a given time period. For energy storage In joint scenarios Down Charging power during the period A collection of energy storage devices in a charging station. For energy storage In joint scenarios Down Discharge power during the period This refers to the energy storage charging and discharging efficiency. 3.The source-load joint scenario construction method for low-carbon scheduling of electric vehicle charging stations according to claim 2, characterized in that, The carbon quota obtained by the electric vehicle is calculated and represented as: in: for Carbon allowances earned by electric vehicles during certain time periods for The first time period with vehicle-to-grid interaction function The charging power of an electric vehicle The largest vehicle number in the set of controllable electric vehicles with vehicle-to-everything (V2X) connectivity. for Electric vehicle charging power during the time period This refers to the distance an electric vehicle can travel per unit of electricity. This refers to the carbon emissions per unit distance traveled by a gasoline-powered vehicle. for The amount of electricity purchased from the upper-level power grid during a given time period. for Actual grid-connected photovoltaic power during the period The proportion of thermal power in the electricity purchased by the power grid. This refers to the carbon emissions per unit power of thermal power units. 4.The source-load joint scenario construction method for low-carbon scheduling of electric vehicle charging stations according to claim 1, characterized in that, The operating constraints of the low-carbon scheduling model of the electric vehicle charging station include energy storage constraints, photovoltaic operating constraints, power transaction constraints, controllable electric vehicle constraints and power balance constraints.

5. The source-load joint scenario construction method suitable for low-carbon scheduling of electric vehicle charging stations according to claim 4, characterized in that, The energy storage operating constraints are represented as: in:, For energy storage In joint scenarios Down Charging power during the period For energy storage In joint scenarios Down Discharge power during the period The maximum allowable charge and discharge power for energy storage. For a set of joint scenarios, A collection of energy storage devices in a charging station. For the set of scheduling periods, For energy storage In joint scenarios Down State of charge during a period of time For energy storage In joint scenarios Down State of charge during a period of time For energy storage charging and discharging efficiency, For scheduling intervals, For energy storage capacity, , Energy storage The minimum and maximum states of charge that are allowed; The photovoltaic operating constraints are represented as: wherein: is the joint scenario the actual photovoltaic on-grid power of the time period, is the joint scenario the photovoltaic power output of the time period, is the joint scenario the curtailed photovoltaic power of the time period; The power transaction constraints are represented as: wherein: is the joint scenario is the joint scenario is the power purchased from the superior grid in the time period, is the joint scenario is the joint scenario is the power sold to the superior grid in the time period, is the maximum value of the interaction power; A set of controllable electric vehicles with vehicle-to-grid interaction function , controllable electric vehicles with constraints, expressed as: in: For joint scenarios Down The first time period with vehicle-to-grid interaction function The charging power of an electric vehicle For joint scenarios Down The first time period with vehicle-to-grid interaction function The discharge power of an electric vehicle The maximum charging and discharging power of electric vehicles, For the first vehicle-to-everything (V2X) ... A set of electric vehicle network access times. For joint scenarios Down State of charge during a period of time For joint scenarios Down State of charge during a period of time For the first vehicle-to-everything (V2X) ... The battery capacity of an electric vehicle, , These are the first ones with vehicle-to-everything (V2X) connectivity features. The minimum and maximum state of charge (SOC) of a battery for an electric vehicle. For joint scenarios Down State of charge during a period of time For the first vehicle-to-everything (V2X) ... The departure time of each electric vehicle For the first vehicle-to-everything (V2X) ... The user of an electric vehicle sets the required state of charge when leaving the vehicle. The power balance constraints are represented as: 。 6.The source-load joint scenario construction method for low-carbon scheduling of electric vehicle charging stations according to claim 1, characterized in that, The carbon sensitivity matrix is represented as: ， ， in: The sensitivity of electric vehicles to carbon quotas, For carbon quotas, for Electric vehicle charging power during the time period This refers to the distance an electric vehicle can travel per unit of electricity. This refers to the carbon emissions per unit distance traveled by a gasoline-powered vehicle. for Solar power output during different time periods For joint scenarios Down The amount of light discarded during a given period The proportion of thermal power in the electricity purchased by the power grid. for The amount of electricity purchased from the upper-level power grid during a given time period. Carbon emissions per unit power of thermal power units. The sensitivity of photovoltaics to carbon quotas, for The first time period with vehicle-to-grid interaction function The charging power of an electric vehicle This is the carbon sensitivity matrix. The sensitivity of electric vehicles to carbon allowances in time period 1. The sensitivity of electric vehicles to carbon allowances in two time periods. For electric vehicles The sensitivity of time period to carbon quotas The sensitivity of photovoltaics to carbon quotas in time period 1. The sensitivity of photovoltaics to carbon quotas in two time periods. For photovoltaics The sensitivity of time period to carbon quotas 2× T The set of real matrices, This represents the maximum time period within the scheduling period. 7.The source-load joint scenario construction method for low-carbon scheduling of electric vehicle charging stations according to claim 1, characterized in that, The diffusion model including the forward diffusion process and the reverse denoising process is constructed, and the specific process is as follows: Based on the electric vehicle charging power and the photovoltaic output of the historical day, the joint time sequence sample of the historical day is constructed and represented as: ， ， wherein: is the joint timing sample of the historical day , is the electric vehicle charging power of the historical day , is the photovoltaic power output of the historical day , is the set of real matrices of 2 x T , is the maximum period in the scheduling period, is the electric vehicle charging power of the historical day in the period , is the photovoltaic power output of the historical day in the period is the photovoltaic power output of the historical day Based on the joint time sequence sample of the historical day, the joint input data is constructed and represented as: ， wherein: is the joint input data, is the number of historical days; Based on the Markov process of the standard Gaussian distribution, let the predefined noise strength be and determine the signal preservation coefficient based on the predefined noise strength, denoted as: wherein: is the first signal preservation coefficient of the step, is the cumulative signal preservation coefficient, denotes the product of the to cumulative signal preservation coefficients. Based on the joint input data and the signal reservation coefficient, the forward diffusion process is constructed and represented as: wherein: is the forward diffusion at the step of the conditional distribution, is the intermediate state of the joint input data during the diffusion process, is a Gaussian distribution with mean and covariance at point , and is the identity matrix. The loss function of the parameterized deep neural network is constructed, and based on the loss function of the parameterized deep neural network, the parameterized deep neural network is used to predict the joint noise item at each step, and the joint input data is gradually denoised to restore the intermediate state in the diffusion process to the real joint sample as the input of the low-carbon scheduling problem to construct the reverse denoising process.

Images