Flexible load automatic optimization scheduling method and system based on carbon strength prediction

Abstract

The invention discloses a flexible load automatic optimization scheduling method and system based on carbon intensity prediction, and belongs to the technical field of smart power grids and carbon emission reduction. The method comprises the following steps: acquiring future power grid carbon intensity prediction data and user load demands; unified modeling is carried out on flexible loads such as an electric vehicle and an air conditioner, and constraint conditions are constructed; constructing a multi-objective optimization function by taking carbon emission minimization, power consumption cost minimization and user comfort degree maximization as objectives; an improved non-dominated sorting genetic algorithm (NSGA-II) is adopted for solving, and a Pareto optimal scheduling scheme is obtained; presenting the scheme through a visual interface and generating a control instruction to be issued and executed; and finally, tracking an actual emission reduction effect and feeding back an optimization model. The system correspondingly comprises a data interface module, an optimization decision module, a user interaction module, an instruction execution module and a feedback optimization module. According to the method, full-closed-loop automation of monitoring-prediction-decision-control-feedback is realized, the flexible load can be effectively guided to operate in a low-carbon period, the carbon emission of a power consumption side is remarkably reduced, and economical efficiency and user experience are considered at the same time.

Description

Technical Field

[0001] This invention relates to the field of smart grid and demand-side management technology, specifically to a flexible load automatic optimization scheduling method and system based on grid carbon intensity prediction. Background Technology

[0002] Achieving the "dual carbon" goals requires deep involvement from the electricity consumption side. Existing technologies suffer from two main shortcomings: First, carbon monitoring and accounting platforms can only calculate and display carbon emissions, essentially "monitoring but not controlling," and cannot provide actionable automated emission reduction solutions. Second, traditional demand-side response systems often rely on fixed electricity prices or simple rules for scheduling, resulting in singular and static optimization objectives and a lack of carbon awareness. This can lead to scheduling results that contradict carbon reduction targets and prevent predictive optimization of loads requiring advance planning, such as electric vehicle charging. Therefore, existing technologies suffer from a disconnect between carbon monitoring and control, and a lack of carbon awareness and predictability in scheduling strategies, making it difficult to support automated and optimized carbon emission reduction on the electricity consumption side. Summary of the Invention

[0003] This invention provides a flexible load automatic optimization scheduling method and system based on carbon intensity prediction to solve the problem that load scheduling in the prior art lacks dynamic carbon perception and multi-objective collaborative optimization capabilities.

[0004] The technical solution of the present invention is as follows:

[0005] A flexible load automatic optimization scheduling method based on carbon intensity prediction includes the following steps:

[0006] S1: Obtain the predicted carbon intensity data of the power grid and obtain the flexible load operation requirements set by the user; the flexible load includes energy load, power load, and temperature-controlled load, as follows: energy load refers to the load that needs to complete a fixed total energy supply within a specified time window (such as electric vehicle charging); power load refers to the load whose operating power needs to be maintained within a preset range; temperature-controlled load refers to the load whose controlled state parameters (such as temperature) need to be maintained within a comfortable range (such as air conditioning).

[0007] S2: Based on the type of flexible load, establish the corresponding schedulable model and constraints;

[0008] S3: Construct a multi-objective optimization function with the objectives of minimizing carbon emissions and minimizing electricity costs; optionally, the multi-objective optimization function also includes a user comfort optimization objective;

[0009] S4: Solve the multi-objective optimization function to obtain the optimal scheduling scheme;

[0010] S5: Generate control commands based on the optimal scheduling scheme and issue them for execution;

[0011] S6: Track actual operating data and actual carbon intensity data, and feed back deviation data to optimize the prediction model and the optimization model.

[0012] Furthermore, the grid carbon intensity prediction data obtained in S1 includes hourly predictions for the next 24 hours; the flexible load operation requirements include equipment type, rated power, dispatchable time window, and total energy required or state interval to be maintained.

[0013] Furthermore, in S2, for energy-type loads, the constraint condition is: within the time window [ , [Total energy completed within] The supply, namely For power-type loads, the constraint is that the operating power of the equipment is maintained within the preset power range; for temperature-controlled loads, the constraint is that the controlled state parameters of the equipment are maintained within the preset comfort range.

[0014] Furthermore, the multi-objective optimization function constructed in S3 also includes user comfort as the optimization objective.

[0015] The multi-objective optimization function is expressed as: ,in:

[0016] , for carbon emission targets. This represents the predicted value of the grid carbon intensity at time t, in gCO2 / kWh (grams of carbon dioxide per kilowatt-hour). The total power of all scheduled flexible loads at time t is expressed in kW (kilowatts); T is the total number of time periods in the scheduling cycle.

[0017] The economic cost objective is... The price is the electricity price at time t, expressed in yuan / kWh.

[0018] The goal is to improve user comfort. This represents the actual indoor temperature at time t, in °C. This indicates the user's set temperature, in °C.

[0019] Furthermore, in step S4, an improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization function.

[0020] Furthermore, S5 includes: presenting the optimal scheduling scheme to the user through a visual interface, the interface including graphical or data information to reflect the optimization effect; and generating and issuing device control commands after receiving user confirmation or modification instructions.

[0021] Furthermore, S6 specifically includes: collecting the actual operating power of the acquisition device. and actual grid carbon intensity The deviation between the predicted or optimized value and the actual value is calculated. When the deviation exceeds a preset threshold, the parameters of the carbon intensity prediction model and the multi-objective optimization model are retrained or fine-tuned.

[0022] A flexible load automatic optimization scheduling system based on carbon intensity prediction, used to implement the above method, includes:

[0023] The data interface module is used to acquire power grid carbon intensity forecast data and user load operation requirements;

[0024] The load modeling module is used to establish a schedulable model and constraints based on the load type.

[0025] The optimization decision module is used to construct and solve multi-objective optimization functions to generate the optimal scheduling scheme.

[0026] The user interaction module is used to visually present the optimal scheduling scheme and receive user feedback.

[0027] The instruction execution module is used to generate and issue device control instructions according to the optimal scheduling scheme;

[0028] The feedback optimization module is used to track actual operation and carbon intensity data, and to provide feedback to optimize the prediction model and the optimization model.

[0029] The data interface module executes step S1, the load modeling module executes step S2, the optimization decision module executes steps S3-S4, the user interaction module and the instruction execution module work together to execute step S5, and the feedback optimization module executes step S6.

[0030] Furthermore, the optimization decision module has a built-in improved non-dominated sorting genetic algorithm solver; the user interaction module is a responsive web application interface that supports access on PCs and mobile devices.

[0031] The beneficial effects of this invention are: for the first time, load scheduling is driven by dynamic carbon intensity prediction, realizing a paradigm shift of "prioritizing carbon emission reduction"; emission reduction, economic and comfort objectives are coordinated through multi-objective optimization; a fully closed-loop automated architecture of "prediction-decision-control-feedback" is constructed to ensure the real controllability and continuous optimization of emission reduction effects; and the improved optimization algorithm used enhances solution efficiency and low-carbon orientation. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall architecture of the system of the present invention. Figure 2 This is the overall flowchart of the method of the present invention. Figure 3 This is a flowchart of the improved NSGA-II algorithm used in this invention. Figure 4 This is a comparison graph of carbon intensity and load curves in one embodiment of the present invention. Figure 5 This is a comparison diagram of electric vehicle charging scheduling optimization before and after one embodiment of the present invention. Figure 6 This is a comparison diagram before and after air conditioning load scheduling optimization in one embodiment of the present invention. Detailed Implementation

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the following embodiments are only used to explain the present invention and do not constitute a limitation thereof.

[0034] Example 1: Overall System Operation Flow

[0035] Reference Figure 1 and Figure 2 The system's operation begins with the data interface module, which periodically retrieves hourly forecasts of grid carbon intensity for the next 24 hours from the grid data platform via a preset API, while simultaneously receiving operational demands from flexible loads at user terminals. Subsequently, the load modeling module transforms the demands into a mathematically constrained schedulable model based on load type (e.g., electric vehicles, air conditioning). The optimization decision module, as the core, calls its built-in, improved Non-Dominated Sorting Genetic Algorithm (NSGA-II) solver to solve an optimization function targeting carbon emissions and electricity costs (extendable to user comfort), generating the optimal scheduling scheme. The user interaction module then presents this scheme and its expected effects (e.g., ...) to the user. Figure 4 The comparison information shown is presented to the user. After confirmation, the instruction execution module compiles the scheme into specific equipment control instructions and issues them for execution. Throughout the entire scheduling cycle and after its end, the feedback optimization module continuously tracks the actual operating power of the equipment and the actual carbon intensity published by the power grid, calculates the deviation, and decides whether to trigger parameter updates of the prediction model or optimization model according to preset rules, thus forming a complete automated closed loop.

[0036] Example 2: Scheduling for a Single Electric Vehicle Charging Load

[0037] This embodiment takes the charging scheduling of an electric vehicle as an example to illustrate in detail the specific implementation of each step of the method described in the claims.

[0038] S1 (Data and Demand Acquisition): The user sets via the app that the vehicle battery must be charged from the current 30% to 90% before 7:00 AM the following day. The vehicle battery capacity is 60kWh, therefore the required charging energy is 36kWh. The charging station's rated power is 7kW. The system synchronously acquires the grid carbon intensity prediction curve for the next 24 hours as follows: Figure 4 The upper curve in the diagram is shown.

[0039] S2 (Modeling and Constraints): This electric vehicle is an energy-type load. Its scheduleable time window is from the current time set by the user to 7:00 the next day, for example [22:00, 07:00]. The constraint is: the total charging energy must meet the following requirements. And power at any time Satisfy 0 ≤ ≤ 7 kW.

[0040] S3 (Objective Function Construction): This example considers two objectives: carbon emissions and economic costs. The function is constructed as follows: .in, , . for Figure 4 The predicted carbon intensity shown is The time-of-use electricity price is known.

[0041] S4 (Optimization Solution): Using the following... Figure 3 The improved NSGA-II algorithm shown in the flowchart is used to solve the problem. The main improvements of the algorithm include: (1) Adaptive crossover probability, using the formula Pc = 0.85 - 0.2×(current iteration number / maximum iteration number), where the preferred value of the maximum iteration number is 200 (the conventional value range is 100-300, and this invention has been verified by a large number of simulations, and 200 iterations can achieve the optimal balance between solution accuracy and efficiency). In the later stage of the iteration, the crossover probability is reduced to promote convergence, so that the crossover probability decreases linearly from 0.85 to about 0.65 as the iteration progresses; (2) Low-carbon oriented elite retention, when selecting individuals to enter the next generation, for individuals in the same non-dominated layer, their carbon emission objective function is retained first. Individuals with smaller values. After iterative calculation, a Pareto optimal solution set is obtained. The system selects the final scheme according to the "carbon emission minimization" strategy, which concentrates the charging load mainly in the early morning period when the predicted carbon intensity is low (see...). Figure 4 The specific optimized charging power curve is as follows: Figure 5 As shown.

[0042] S5 (Command Execution): The system generates control commands {Time: "01:00", Device: "EV_Charger_01", Command: "Start", Parameter: 7kW} and {Time: "06:00", Device: "EV_Charger_01", Command: "Stop"}, and sends them to the designated charging piles through the IoT platform.

[0043] S6 (Feedback Optimization): After charging is completed, the system collects actual charging records and calculates the total charging energy to be 35.8 kWh. The deviation between the average actual carbon intensity and the predicted value during this period is 4%. Since the deviation does not exceed the preset 5% threshold, no model update is triggered this time. This scheduling is expected to reduce carbon emissions by approximately 12 kg.

[0044] Example 3: Air Conditioning Load and Multi-Load Joint Dispatch

[0045] This embodiment demonstrates the applicability of the present invention to temperature-controlled loads and complex scenarios.

[0046] Taking the central air conditioning system of an office area as an example, a temperature control model is established. A simplified first-order equivalent thermal model is adopted: ,in For air conditioning cooling capacity, Indoor temperature, Outdoor temperature. User comfort constraint is 24℃ ≤ ≤ 26℃. In this case, the multi-objective optimization function constructed in step S3 needs to include a comfort objective. .

[0047] When jointly optimizing the electric vehicle fleet and central air conditioning system of the office area, the optimization decision module will establish a joint optimization problem that includes all load models and constraints, and will still use the improved NSGA-II algorithm described in Example 2 (see Example 2). Figure 3 The solution yields a coordinated scheduling scheme: guiding the dense distribution of electric vehicle charging loads during the deep low-carbon period at night; simultaneously, while ensuring comfortable indoor temperatures, appropriately adjusting the air conditioning operation strategy, such as pre-cooling before the afternoon carbon intensity peak, to smooth out peak power consumption. A schematic diagram illustrating the coordinated relationship between the load curve and the carbon intensity curve of this joint optimization scheme is shown below. Figure 6 As shown in the figure. Simulation results show that this solution can effectively reduce the overall carbon emissions of electricity consumption in the park by more than 30% while ensuring user experience.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A flexible load automatic optimization scheduling method based on carbon intensity prediction, characterized in that, Includes the following steps: S1: Obtain grid carbon intensity prediction data and obtain user-defined flexible load operation requirements; the flexible loads include energy-type loads, power-type loads, and temperature-controlled loads; S2: Based on the type of flexible load, establish the corresponding schedulable model and constraints; S3: Construct a multi-objective optimization function with the objectives of minimizing carbon emissions and minimizing electricity costs; Optionally, the multi-objective optimization function further includes a user comfort optimization objective; S4: Solve the multi-objective optimization function to obtain the optimal scheduling scheme; S5: Generate control commands based on the optimal scheduling scheme and issue them for execution; S6: Track actual operating data and actual carbon intensity data, and feed back deviation data to optimize the prediction model and the optimization model.

2. The method according to claim 1, characterized in that, The power grid carbon intensity prediction data obtained in S1 includes hourly predictions for the next 24 hours; the flexible load operation requirements include equipment type, rated power, dispatchable time window, and total energy required or state range to be maintained.

3. The method according to claim 1, characterized in that, In S2, for energy-type loads, the constraint condition is: within the time window [ , [Total energy completed within] The supply, namely For power-type loads, the constraint is that the equipment operating power must be maintained within a preset power range; For temperature-controlled loads, the constraint is that the controlled state parameters of the equipment are maintained within a preset comfort range.

4. The method according to claim 1, characterized in that, The multi-objective optimization function constructed in S3 also includes user comfort as the optimization objective.

5. The method according to claim 4, characterized in that, The multi-objective optimization function is expressed as: ,in: For carbon emission targets, Predict carbon intensity at time t. Let t be the total scheduling power at time t; For the economic cost objective, Let t be the electricity price at time t; For user comfort goals, Let t be the indoor temperature. To set the temperature.

6. The method according to claim 1, characterized in that, In step S4, an improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization function.

7. The method according to claim 1, characterized in that, S5 includes: presenting the optimal scheduling scheme to the user through a visual interface, the interface including graphic or data information to reflect the optimization effect; and generating and issuing device control commands after receiving user confirmation or modification instructions.

8. The method according to claim 1, characterized in that, S6 specifically includes: the actual operating power of the data acquisition device. and actual grid carbon intensity The deviation between the predicted or optimized value and the actual value is calculated. When the deviation exceeds a preset threshold, the parameters of the carbon intensity prediction model and the multi-objective optimization model are retrained or fine-tuned.

9. A flexible load automatic optimization scheduling system based on carbon intensity prediction, characterized in that, To implement the method of any one of claims 1 to 8, comprising: The data interface module is used to acquire power grid carbon intensity forecast data and user load operation requirements; The load modeling module is used to establish a schedulable model and constraints based on the load type. The optimization decision module is used to construct and solve multi-objective optimization functions to generate the optimal scheduling scheme. The user interaction module is used to visually present the optimal scheduling scheme and receive user feedback. The instruction execution module is used to generate and issue device control instructions according to the optimal scheduling scheme; The feedback optimization module is used to track actual operation and carbon intensity data, and to provide feedback to optimize the prediction model and the optimization model. The data interface module executes step S1, the load modeling module executes step S2, the optimization decision module executes steps S3-S4, the user interaction module and the instruction execution module work together to execute step S5, and the feedback optimization module executes step S6.

10. The system according to claim 9, characterized in that, The optimization decision module has a built-in improved non-dominated sorting genetic algorithm solver; the user interaction module is a responsive web application interface that supports access on PC and mobile devices.

Images