Dynamic sharing-charging coupling scheduling optimization method and system

Abstract

The invention relates to a dynamic sharing-charging coupling scheduling optimization method and system, and the method comprises the steps: dividing a target region into a plurality of sub-regions, obtaining the power grid data, order data, charging pile data and vehicle data of each region, and obtaining a target function; constructing a loss function of dynamic sharing-charging coupling scheduling based on the power grid data, the order data, the charging pile data and the vehicle data; and based on the objective function and the loss function, scheduling by using a federal learning algorithm to obtain the vehicle behavior when the objective function is optimal, and the system is used for implementing the method. Compared with the prior art, the method has the advantages that photovoltaic power generation is taken into consideration of vehicle charging scheduling, the objective function is constructed based on multiple main body benefits in the charging scheduling process, benefit maximization is ensured, and the privacy security of each charging aggregator in the scheduling process is ensured by utilizing federal learning.

Description

Technical Field

[0001] The present invention relates to the field of vehicle scheduling and energy, and in particular to a dynamic sharing-charging coupling scheduling optimization method and system. Background Art

[0002] New energy vehicles, as an emerging representative of flexible resources on the user side, have achieved rapid development worldwide in recent years. With the rapid increase in the number of electric vehicles, their charging load has become an important factor that cannot be ignored in the operation of the power grid. Especially in the context of the evening peak electricity consumption in urban power grids, the charging behavior of electric vehicles during the evening peak period further aggravates the burden on the power grid, resulting in the "duck curve" phenomenon (i.e., the characteristic that electricity demand is lower during the day and higher at night) becoming more significant; given the randomness and flexibility of electric vehicle charging time and location, they actually have huge scheduling potential, while the current research on electric vehicle charging scheduling issues mainly focuses on the perspective of shared travel platforms and urban power grids, but ignores the impact of photovoltaic power generation on new energy vehicle charging scheduling. This limitation leads to the scheduling strategy failing to fully utilize the potential of photovoltaic energy, and failing to effectively cope with the power grid load pressure brought by electric vehicle charging. In addition, when formulating collective scheduling strategies, it is inevitable that the charging-related data of various electric aggregators will be exposed, and privacy protection cannot be achieved.

[0003] Therefore, providing a method to incorporate PV generation into charging scheduling while respecting the data privacy of various electric aggregators is a problem that needs to be solved. Summary of the invention

[0004] The purpose of the present invention is to provide a dynamic sharing-charging coupling scheduling optimization method and system in order to overcome the defects of the above-mentioned prior art. The method constructs a benefit objective function for the entities involved in the charging scheduling process, such as photovoltaic power generation companies, traditional power grids and shared travel platforms, incorporates photovoltaic power generation into the scope of charging scheduling considerations, fully utilizes the potential of photovoltaic energy, and can generate more accurate scheduling plans.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] According to a first aspect of the present invention, a dynamic sharing-charging coupling scheduling optimization method is provided, the method comprising:

[0007] Divide the target area into multiple sub-areas and number them, obtain the power grid data, order data, charging pile data and vehicle data of each area, and obtain the objective function, which includes the sharing platform benefit function, the regional fairness benefit function and the photovoltaic power generation utilization function;

[0008] Based on the grid data, order data, charging pile data and vehicle data, a loss function for dynamic sharing-charging coupling scheduling is constructed, wherein the loss function includes a benefit loss function of a photovoltaic power generation company, a loss benefit function of a traditional grid and a benefit loss function of a shared travel platform;

[0009] Based on the objective function and loss function, a federated learning algorithm is used for scheduling to obtain the vehicle behavior when the objective function is optimal. The vehicle behavior includes: charging and charging sub-area number or receiving orders and going to the sub-area number for receiving orders.

[0010] As a preferred technical solution, the objective function is obtained by weighted summation of the sharing platform benefit function, the regional fairness benefit function and the photovoltaic power generation utilization rate function.

[0011] As a preferred technical solution, the expression of the benefit loss function of the photovoltaic power generation company is:

[0012]

[0013] Where T represents the total time, P t represents the electricity price in time period t, g t represents the photovoltaic power generation in time period t, sp t represents the penalty cost of overflow electricity, d t is the grid demand in time period t.

[0014] As a preferred technical solution, the expression of the traditional power grid loss benefit function is:

[0015]

[0016] L grid,peak =λ∑ t∈PeakHours s t ,

[0017] Among them, s t represents the total power supply from photovoltaic power generation and other methods during time period t, d t is the grid demand in period t, and λ represents the additional cost coefficient in the peak period.

[0018] As a preferred technical solution, the expression of the benefit loss function of the shared travel platform is:

[0019]

[0020] Among them, d j represents the mileage of vehicle j, c j represents the unit mileage cost, and M represents the total number of vehicles.

[0021] As a preferred technical solution, the loss function is obtained by weighted summing the benefit loss function of the photovoltaic power generation company, the loss benefit function of the traditional power grid and the benefit loss function of the shared travel platform, and its expression is:

[0022] L=αL PV +βL grid +γL platform ,

[0023] Among them, α, β and γ represent the weight coefficients of each loss function; L PV Represents the benefit loss function of the photovoltaic power generation company, based on L PV,usage and L PV,spill Get; L grid Represents the traditional power grid loss benefit function, based on L grid,demand and L grid,peak Get; L platform Represents the benefit loss function of the shared travel platform.

[0024] As a preferred technical solution, the method for scheduling using a federated learning algorithm includes:

[0025] Local update: Obtain all electric aggregators in each sub-region, calculate the loss function and gradient value of each electric aggregator based on its objective function, and perform local optimization update on the loss function;

[0026] Aggregate update: Aggregate the loss functions of all electric aggregators after local optimization and update to obtain an aggregate loss function, perform global optimization update based on the aggregate loss function, and distribute the results of the global optimization update to each electric aggregator for the next local optimization update;

[0027] Iteration: Repeat local updates and aggregate updates until the algorithm converges or reaches the training target;

[0028] Scheduling: Use the trained federated learning algorithm to obtain vehicle behavior when the objective function is optimal.

[0029] As a preferred technical solution, the polymerization method is:

[0030]

[0031] in, represents the local optimization update of the ith sub-region at time t, |D i | represents the local data of the i-th sub-region, N represents the total number of sub-regions, and j represents the j-th sub-region.

[0032] As a preferred technical solution, the training goal is: the rate of change of the loss function after the previous optimization update and the loss function after the next optimization update is less than a preset value.

[0033] According to a second aspect of the present invention, a dynamic sharing-charging coupling scheduling optimization system is provided, wherein the system is used to implement the above method.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] 1) The present invention provides a charging scheduling method for electric online car-hailing vehicles under the background of the new stage of photovoltaic development. By constructing an objective function and a loss function including the benefits of photovoltaic power generation, a more reasonable scheduling of online car-hailing vehicle charging is carried out for the purpose of improving the photovoltaic utilization rate, which can effectively reduce the adverse effects of the "duck curve", achieve the goal of peak shaving and valley filling, and promote the efficient use of green energy; it also ensures that the comprehensive benefits of photovoltaic power generation companies, online car-hailing platforms and power grid companies in the charging and scheduling process are maximized;

[0036] 2) The present invention also adopts federated learning technology to enable each charging aggregator to independently train based on local data, improve data privacy protection and model training efficiency, and avoid privacy leakage problems in centralized data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0039] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0040] This embodiment takes a certain city as an example to perform vehicle dynamic sharing-charging coupling scheduling optimization, and its process is as follows: Figure 1 As shown, the following steps are included:

[0041] S1. Data and target function acquisition:

[0042] S11. Divide the sample city into multiple sub-areas and number them, determine the subjects in each sub-area and the purpose that each subject hopes to achieve.

[0043] Among them, the main bodies include: traditional power companies, shared travel platforms, photovoltaic power generation companies, fleets, passengers, charging stations, etc.; specifically, 1) traditional power grid companies hope to reasonably dispatch vehicle charging to achieve peak shaving and valley filling and alleviate power grid pressure; 2) shared travel platforms hope to maximize profits while ensuring the quality of passenger services; 3) photovoltaic power generation companies hope to improve photovoltaic utilization and reduce photovoltaic waste; 4) the fleet is completely controlled by the shared travel platform in the model; 5) passengers hope that their own service quality can be guaranteed; 6) charging stations hope to serve as many vehicles as possible.

[0044] S12. Acquire the data generated by each subject in each sub-area, including power grid data, order data, charging pile data, and vehicle data.

[0045] The sample cities were intercepted for part of the data requested throughout the day, where order data included: 30,309 orders, order ID, order start time, order start point, and the latitude and longitude of the order end point; vehicle data included: vehicle maximum endurance, full charge time, and mileage per kilowatt-hour; power grid data included: photovoltaic power generation, photovoltaic power generation share, photovoltaic subsidies, power grid demand, and total power grid power generation; charging pile data included: number of charging piles, charging pile numbers, and charging pile charging voltage.

[0046] S13. Combine the purpose that the subject wants to achieve and the data obtained in step S12 to obtain the objective function. The objective function is expressed as: F = max{F1, F2, F3}. In more detail, the objective function can be expressed by constructing a Pareto solution, which is expressed as:

[0047]

[0048] in, represents the normalized function of the ζth objective function in the kth optimal solution; represents the ζth objective function in the kth optimal solution; represents the minimum value of the ζth objective function; It represents the maximum value of the ζth objective function; ζ represents the objective function type, and when ζ=1, it represents the travel benefit of the sharing platform; when ζ=2, it represents the regional fairness benefit; and when ζ=3, it represents the utilization rate of photovoltaic power generation.

[0049] The objective functions include:

[0050] A. The sharing platform benefit function is obtained from the charging pile data and vehicle data, and its expression is:

[0051]

[0052] Among them, F1 represents the travel benefits of the sharing platform. represents the set of online ride-hailing vehicles participating in the dispatch, represents the set of travel requests involved in scheduling, Indicates the benefits generated by vehicle pick-up and drop-off orders, Represents the cost of vehicle pick-up and drop-off orders, represents the regional photovoltaic power generation subsidy coefficient, represents the cost of the vehicle going to charge, represents the set of charging stations participating in the scheduling, k represents the charging station number, x ij represents the decision variable for vehicle i to pick up order j, y ik represents the decision variable for vehicle i to go to charging station k, P pv (t) represents photovoltaic power generation, P max (t) represents the vehicle charging power.

[0053] The benefits generated by vehicle pick-up and drop-off orders are composed of two parts: the commission of the sum of the starting price, mileage and driving time, and the labor fee of the remaining amount. The calculation formula is as follows:

[0054]

[0055] Among them, p sta Indicates the starting price of the order, p km Indicates the cost per kilometer of the order, s j represents the order distance, p min represents the service fee per minute for an order, t j Indicates the order time, c pla represents the commission of the total amount of the order, θ pla Indicates the labor fee commission of the remaining amount of the order.

[0056] The cost of vehicle pick-up orders is mainly determined by the distance between the vehicle's current location and the order pick-up point. What needs to be explained is the loss of the platform when the vehicle runs idly for 1 km. km " is calculated based on the hourly wages of local drivers, platform order benefits and other information, and the formula is as follows:

[0057]

[0058] And d ij Indicates the distance between the vehicle and the order.

[0059] The regional photovoltaic power generation subsidy coefficient is determined by the local photovoltaic power generation company, because the photovoltaic power generation company hopes that vehicles can be charged during the peak period of photovoltaic power generation to increase the utilization rate of photovoltaics.

[0060] The cost of a vehicle going to a charging station is mainly determined by the current electricity price, the distance between the vehicle and the charging station, and the estimated waiting time at the charging station. It is expressed in three parts, representing the distance cost from the current location of the vehicle to the charging station, the estimated waiting cost at the charging station, and the charging cost, as shown below:

[0061]

[0062] in, Indicates the estimated waiting time for the vehicle to charge at the charging station, L min It represents the loss caused to the platform for every idle minute. Indicates the estimated time for charging the vehicle, e price Indicates the electricity price.

[0063] B. Regional fairness benefit function, obtained from vehicle data, is expressed as:

[0064]

[0065] Among them, p represents a sub-region, P represents a sub-region set, SAV p represents the number of vehicles in area p after scheduling, TOTAL SAV Represents the total number of vehicles involved in the dispatch, RE p Indicates the predicted number of requests from region p to this sub-region, TOTAL RE Represents the total number of requests; the number of vehicles in area p after scheduling includes the original number of vehicles in area p, the number of vehicles going to area p due to orders, and the number of vehicles going to area p due to charging.

[0066] Among them, the number of vehicles in area p after scheduling is the original number of vehicles in the area, the number of vehicles going to the area due to orders, and the number of charging vehicles going to the area due to charging. The calculation formula for the number of vehicles going to the area due to orders is:

[0067]

[0068] The calculation formula for the number of vehicles going to the area for charging is:

[0069]

[0070] C. Photovoltaic power generation utilization function, obtained from power grid data, is expressed as:

[0071]

[0072] Among them, F3 represents the utilization rate of photovoltaic power generation, T represents the target period, P pv (t) represents photovoltaic power generation, P sav (t) represents the vehicle charging power.

[0073] S2. Loss function construction:

[0074] The loss function of dynamic sharing-charging coupling scheduling is constructed based on power grid data, order data, charging pile data and vehicle data. The loss function includes the benefit loss function of photovoltaic power generation company, the loss benefit function of traditional power grid and the benefit loss function of shared travel platform. Its expression is:

[0075] L=αL PV +βL grid +γL platform ,

[0076] Among them, α, β and γ represent the weight coefficients of each loss function; L PV Represents the benefit loss function of the photovoltaic power generation company, based on L PV,usage and L PV,spill Get; L grid Represents the traditional power grid loss benefit function, based on L grid,demand and L grid,peak Get; L platform Represents the benefit loss function of the shared travel platform.

[0077] Specifically, the expression of the benefit loss function of the photovoltaic power generation company is:

[0078]

[0079] Where T represents the total time, P t represents the electricity price in time period t, g t represents the photovoltaic power generation in time period t, sp t represents the penalty cost of overflow electricity, d t is the grid demand in time period t.

[0080] The expression of the traditional power grid loss benefit function is:

[0081]

[0082] L grid,peak =λ∑ t∈PeakHours s t ,

[0083] Among them, s t represents the total power supply from photovoltaic power generation and other methods during time period t, d t is the grid demand in period t, and λ represents the additional cost coefficient in the peak period.

[0084] The expression of the benefit loss function of the shared travel platform is:

[0085]

[0086] Among them, d j represents the mileage of vehicle j, c j represents the unit mileage cost, and M represents the total number of vehicles.

[0087] S3. Scheduling using federated learning algorithms:

[0088] Based on the constructed objective function and loss function, the vehicle behavior when the objective function has the optimal solution is obtained by using the federated learning algorithm, wherein the federated learning algorithm includes multiple local blocks, namely, electric aggregators, and a central server.

[0089] The detailed steps include:

[0090] S31. Local update: Obtain all electric charging aggregators in each sub-area, calculate the loss function and gradient value for each charging aggregator based on its own local data, and perform local optimization update on the loss function. After completing the local optimization update, each electric charging aggregator uploads the results of the local optimization update to the central server.

[0091] S32, aggregate update: The central server aggregates the locally optimized and updated loss functions of all electric aggregators to obtain the aggregate loss function. This aggregation process is usually a weighted average:

[0092]

[0093] represents the local optimization update result of the i-th charging aggregator, N represents the total number of charging aggregators, and D i represents the result of the local optimization update of the charging aggregator, and j represents the jth charging aggregator.

[0094] A global optimization update is performed based on the aggregate loss function, and the results of the global optimization update are distributed to each electric charging aggregator for the next charging aggregator local optimization update.

[0095] S33, iteration: Repeat local update and aggregate update until the algorithm converges or the change rate between the loss function after the previous optimization update and the loss function after the next optimization update is less than 0.01%.

[0096] S34. Scheduling: Utilize the trained federated learning algorithm to obtain the vehicle behavior when the objective function is optimal, i.e., the solution of the Pareto solution in step S13 is the maximum value. The specific vehicle behaviors include: charging and charging sub-area number or receiving orders and going to the sub-area number for receiving orders.

[0097] This embodiment also provides a dynamic sharing-charging coupling scheduling optimization system, which is used to implement the above method. Technical personnel in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working process described can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0098] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A dynamic sharing-charging coupling scheduling optimization method, characterized in that: The method includes: Divide the target area into multiple sub-areas and number them, obtain the power grid data, order data, charging pile data and vehicle data of each area, and obtain the objective function, which includes the sharing platform benefit function, the regional fairness benefit function and the photovoltaic power generation utilization function; Based on the grid data, order data, charging pile data and vehicle data, a loss function for dynamic sharing-charging coupling scheduling is constructed, wherein the loss function includes a benefit loss function of a photovoltaic power generation company, a loss benefit function of a traditional grid and a benefit loss function of a shared travel platform; Based on the objective function and loss function, a federated learning algorithm is used for scheduling to obtain the vehicle behavior when the objective function is optimal. The vehicle behavior includes: charging and charging sub-area number or receiving orders and going to the sub-area number for receiving orders.

2. A dynamic sharing-charging coupling scheduling optimization method according to claim 1, characterized in that: The objective function is obtained by weighted summation of the sharing platform benefit function, the regional fairness benefit function and the photovoltaic power generation utilization rate function.

3. The method for optimizing dynamic sharing-charging coupling scheduling according to claim 1, characterized in that: The expression of the benefit loss function of the photovoltaic power generation company is: Where T represents the total time, P t represents the electricity price in time period t, g t represents the photovoltaic power generation in time period t, sp t represents the penalty cost of overflow electricity, d t is the grid demand in time period t.

4. The method for optimizing dynamic sharing-charging coupling scheduling according to claim 3, characterized in that: The expression of the traditional power grid loss benefit function is: L grid,peak =λ∑ t∈PeakHours s t , Among them, s t represents the total power supply from photovoltaic power generation and other methods during time period t, d t is the grid demand in period t, and λ represents the additional cost coefficient in the peak period.

5. A dynamic sharing-charging coupling scheduling optimization method according to claim 4, characterized in that: The expression of the shared travel platform benefit loss function is: Among them, d j represents the mileage of vehicle j, c j represents the unit mileage cost, and M represents the total number of vehicles.

6. A dynamic sharing-charging coupling scheduling optimization method according to claim 5, characterized in that: The loss function is obtained by weighted summing the benefit loss function of the photovoltaic power generation company, the loss benefit function of the traditional power grid and the benefit loss function of the shared travel platform, and its expression is: L=αL PV +βL grid +γL Platform , Among them, α, β and γ represent the weight coefficients of each loss function; L PV Represents the benefit loss function of the photovoltaic power generation company, based on L PV,usage and L PV,spill Get; L grid Represents the traditional power grid loss benefit function, based on L grid,demand and L grid,peak Get; L platform Represents the benefit loss function of the shared travel platform.

7. The method for optimizing dynamic sharing-charging coupling scheduling according to claim 1, characterized in that: The method for scheduling using a federated learning algorithm includes: Local update: Obtain all electric aggregators in each sub-region, calculate the loss function and gradient value of each electric aggregator based on its objective function, and perform local optimization update on the loss function; Aggregate update: Aggregate the loss functions of all electric aggregators after local optimization and update to obtain an aggregate loss function, perform global optimization update based on the aggregate loss function, and distribute the results of the global optimization update to each electric aggregator for the next local optimization update; Iteration: Repeat local updates and aggregate updates until the algorithm converges or reaches the training target; Scheduling: Use the trained federated learning algorithm to obtain vehicle behavior when the objective function is optimal.

8. The method for optimizing dynamic sharing-charging coupling scheduling according to claim 7, characterized in that: The polymerization method is: in, represents the local optimization update of the ith sub-region at time t, |D i | represents the local data of the i-th sub-region, N represents the total number of sub-regions, and j represents the j-th sub-region.

9. The method for optimizing dynamic sharing-charging coupling scheduling according to claim 7, characterized in that: The training goal is that the rate of change of the loss function after the previous optimization update and the loss function after the next optimization update is less than a preset value.

10. A dynamic sharing-charging coupling scheduling optimization system, characterized in that: The system is used to implement the method according to any one of claims 1 to 9.

Images