A method and system for coordinated dispatching of a light storage system, a light storage system and a medium

Abstract

The application relates to a photovoltaic storage system cooperative scheduling method, a scheduling system, a photovoltaic storage system and a medium, wherein the method comprises the following steps: predicting photovoltaic predicted power and load predicted power in a scheduling period based on historical photovoltaic power data, historical meteorological data, historical load data, power grid price data and hard constraint parameters; obtaining a benchmark scheduling strategy in the scheduling period based on a photovoltaic predicted power sequence, a load predicted power sequence and constraint conditions in a day-ahead stage; correcting the benchmark scheduling strategy in a current scheduling period based on the deviation of real-time photovoltaic power data and photovoltaic predicted power and the deviation of real-time load power data and load predicted power in a rolling window in an intra-day stage, so as to obtain an optimal scheduling strategy in the current scheduling period; and issuing an execution instruction generated based on the optimal scheduling strategy to a bidirectional energy storage converter, thereby improving the long-term stability and accuracy of photovoltaic storage system cooperative scheduling.

Description

Technical Field

[0001] This application relates to the field of photovoltaic control technology, and in particular to a method, scheduling system, photovoltaic-storage system and medium for coordinated scheduling of photovoltaic and energy storage systems. Background Technology

[0002] With the advancement of dual-carbon goals, distributed photovoltaic (PV) systems are rapidly gaining popularity among industrial, commercial, and industrial park users. However, their output is significantly affected by weather, exhibiting strong intermittency and volatility. Simultaneously, user loads exhibit peak-valley differences, compounded by time-of-use / dynamic pricing mechanisms, leading to high electricity costs. Furthermore, when PV-storage systems participate in grid interaction, they must also meet multiple constraints such as feed-in power and grid-connected capacity.

[0003] Among the related methods, the patent with publication number CN116316767B controls the timing of electricity purchase or sale by presetting parameters such as "electricity purchase / sale period, state of charge (SOC) threshold, and execution power." However, its scheduling logic is rigid, lacks global optimization capabilities, and is difficult to adapt to electricity price fluctuations and load uncertainties, easily getting trapped in local optima. The patent with publication number CN119109019B uses real-time electricity price clustering to divide time periods and formulates charging and discharging strategies for the next moment. However, it is based only on single-step forward decision-making, lacks a day-ahead and intraday coordination mechanism, and has no error feedback or strategy self-updating capabilities, thus limiting its economic efficiency and robustness.

[0004] Against this backdrop, there is an urgent need for an intelligent scheduling method that, while prioritizing the reliability of users' electricity supply, coordinates and optimizes charging and discharging behavior and power grid purchase and sale strategies to achieve the dual goals of reducing electricity costs and increasing electricity sales revenue. Summary of the Invention

[0005] Therefore, it is necessary to provide a method for coordinated scheduling of optical-storage systems, a scheduling system, an optical-storage system, and a medium to address the aforementioned technical problems.

[0006] In a first aspect, embodiments of this application provide a method for coordinated scheduling of a photovoltaic-storage system, wherein the photovoltaic-storage system includes an energy storage system, and the method includes:

[0007] Based on historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the energy storage system, the photovoltaic power forecast and load forecast during the scheduling cycle are predicted.

[0008] In the day-ahead phase, based on the photovoltaic predicted power, the load predicted power, and the constraints, a baseline scheduling strategy for the scheduling period is obtained.

[0009] During the intraday phase, based on the deviation between real-time photovoltaic power data and photovoltaic predicted power within the rolling window, and the deviation between real-time load power data and load predicted power, the baseline scheduling strategy for the current scheduling period is corrected to obtain the optimal scheduling strategy for the current scheduling period; and, based on the optimal scheduling strategy, an execution instruction is generated and the execution instruction is sent to the bidirectional energy storage converter.

[0010] In one embodiment, the hard constraint parameters of the energy storage system include: initial state of charge of energy storage, total energy storage capacity, rated power of energy storage, charging efficiency, discharging efficiency, state of charge safety boundary in each operating mode, grid power limit, and feeder power limit.

[0011] In one embodiment, the constraints include: power balance constraints, energy storage dynamic constraints, power limit constraints, mode-dependent constraints, and state-of-charge safety boundary constraints.

[0012] In one embodiment, obtaining the baseline scheduling strategy within the scheduling period based on the photovoltaic predicted power, the load predicted power, and constraints includes:

[0013] Based on explicit electricity sales revenue, explicit electricity purchase expenditure, and implicit electricity savings, an objective function for maximizing comprehensive revenue within the scheduling cycle is constructed.

[0014] The photovoltaic predicted power, the load predicted power, and the constraints are input into a mixed-integer linear programming model to solve the objective function, thereby obtaining the baseline scheduling strategy within the scheduling cycle.

[0015] In one embodiment, the baseline scheduling strategy includes operating modes and grid interaction power. The step of correcting the baseline scheduling strategy for the current scheduling period based on the deviation between real-time photovoltaic power data and predicted photovoltaic power within the rolling window, and the deviation between real-time load power data and predicted load power, to obtain the optimal scheduling strategy for the current scheduling period includes:

[0016] Based on the cumulative deviation between real-time photovoltaic power data and photovoltaic power prediction within the rolling window, the photovoltaic power prediction for the current scheduling period is corrected to obtain the photovoltaic power prediction correction value; and based on the cumulative deviation between real-time load power data and load power prediction within the rolling window, the load power prediction for the current scheduling period is corrected to obtain the load power prediction correction value.

[0017] Based on the photovoltaic power forecast correction value, load power forecast correction value, real-time photovoltaic power data, and real-time load power data for the current scheduling period, the baseline scheduling strategy for the current scheduling period is corrected to obtain the optimal grid interaction power for the current scheduling period.

[0018] Based on the energy storage state of charge during the current scheduling period and the state of charge safety boundary under each working mode, all feasible working modes are determined, and the working mode with the greatest immediate benefit is selected as the optimal working mode for the current scheduling period.

[0019] Based on the optimal power grid interaction power and the optimal operating mode, the optimal scheduling strategy for the current scheduling period is obtained.

[0020] In one embodiment, the formula for calculating the optimal grid interaction power for the current scheduling period by correcting the baseline scheduling strategy based on the photovoltaic power forecast correction value, load power forecast correction value, real-time photovoltaic power data, and real-time load power data for the current scheduling period is as follows:

[0021]

[0022] in, for Time-corrected grid interaction power for Uncorrected grid interaction power during the time period for Net power deviation over the time period;

[0023] ;

[0024] in, for Real-time photovoltaic power data for the time period for Correction values ​​for photovoltaic power forecasts for the specified period. for Real-time load power data for the time period for The power correction value for load forecasting during the time period.

[0025] In one embodiment, the method further includes:

[0026] A reward function is constructed based on the cumulative revenue within the statistical time window, and the state-of-charge safety boundary for each working mode is updated based on the feedback of the reward function.

[0027] Secondly, embodiments of this application also provide a collaborative scheduling system for a photovoltaic-storage system, the system comprising:

[0028] The prediction module is used to predict the photovoltaic power and load power during the scheduling period based on historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the photovoltaic and energy storage system.

[0029] The acquisition module is used to obtain a baseline scheduling strategy for the scheduling period based on the photovoltaic predicted power, the load predicted power, and constraints during the day-ahead phase.

[0030] The correction module is used to correct the baseline scheduling strategy for the current scheduling period during the intraday phase based on the deviation between the real-time photovoltaic power data and the photovoltaic power forecast within the rolling window, as well as the deviation between the real-time load power data and the load forecast, so as to obtain the optimal scheduling strategy for the current scheduling period.

[0031] The execution module is used to generate execution instructions based on the optimal scheduling strategy and send the execution instructions to the bidirectional energy storage converter.

[0032] Thirdly, embodiments of this application also provide an optical storage system, including a controller, which is used to implement the method described in the first aspect above.

[0033] Fourthly, embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in the first aspect above.

[0034] The aforementioned photovoltaic-storage system collaborative scheduling method, scheduling system, photovoltaic-storage system, and medium, by employing a multi-timescale collaborative optimization scheme, obtain a baseline scheduling strategy within the scheduling cycle based on the predicted photovoltaic power, the predicted load power, and constraints during the day-ahead phase. During the intraday phase, the baseline scheduling strategy for the day-ahead phase is corrected based on the deviation between real-time photovoltaic power data and predicted photovoltaic power within the rolling window, as well as the deviation between real-time load power data and predicted load power, thereby improving the long-term stability and accuracy of photovoltaic-storage system collaborative scheduling.

[0035] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0036] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0037] Figure 1 This is a hardware structure block diagram of a terminal device for a collaborative scheduling method of a photovoltaic and energy storage system in one embodiment;

[0038] Figure 2 This is a flowchart illustrating the collaborative scheduling method for a photovoltaic-storage system in one embodiment;

[0039] Figure 3This is a structural diagram of the collaborative scheduling system of the optical storage system in one embodiment;

[0040] Figure 4 This is a structural diagram of the photovoltaic-storage system collaborative scheduling system in a preferred embodiment. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0042] The method embodiments provided in this example can be executed on a terminal, computer, or similar computing device. For example, it can run on a terminal. Figure 1 This is a hardware structure block diagram of the terminal of the optical-storage system collaborative scheduling method in this embodiment. For example... Figure 1 As shown, a terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 and a memory 104 for storing data are also included. The processor 102 may be, but is not limited to, a microprocessor (MCU) or a programmable logic device (FPGA). The terminal may also include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that… Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal described above. For example, the terminal may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0043] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the optical storage system collaborative scheduling method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0044] The transmission device 106 is used to receive or send data via a network. This network includes a wireless network provided by the terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 can be a radio frequency (RF) module used for wireless communication with the Internet.

[0045] This application provides a method for collaborative scheduling of a photovoltaic-storage system, which can be applied to... Figure 1 Taking the terminal in the example of this, for example... Figure 2 As shown, the method includes the following steps:

[0046] Step S201: Based on historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the energy storage system, predict the photovoltaic power and load power during the scheduling period.

[0047] Obtain historical photovoltaic power and historical meteorological data, and use It means that, among them, Historical photovoltaic power data, Indicating a historical moment Actual photovoltaic power generation collected (unit: kW); Historical meteorological data, including but not limited to: solar irradiance (W / m²), ambient temperature (°C), relative humidity (%), cloud cover index, wind speed (m / s), and other key meteorological factors affecting photovoltaic power; historical data window. The timeframe is typically set to 7–30 days to cover typical weather patterns (sunny, cloudy, rainy) and load cycles (weekdays / weekends).

[0048] Obtain historical load data Indicates the user at a historical moment The actual power consumption (kW) is obtained from smart meters or energy management systems (EMS). Due to the strong time correlation (daily cycle, weekly cycle) and event sensitivity (such as holidays, equipment start-up and shutdown) of historical load data, a sufficient historical data window length needs to be retained to capture behavioral patterns.

[0049] Obtaining time-of-use electricity price information , For time period The grid purchase price of electricity (yuan / kWh); For time period Electricity price sold to the grid (yuan / kWh); electricity price series The corresponding scheduling cycle (e.g., a scheduling cycle of 24 hours, with a scheduling period of 15 minutes, totaling 96 points) is announced in advance by the power grid company.

[0050] Obtain the set of hard constraint parameters for the energy storage system, including: initial state of charge, total energy storage capacity, rated power of energy storage, charging efficiency, discharging efficiency, state of charge safety boundary under each operating mode, grid power limit, and feeder power limit.

[0051] Initial state of charge of energy storage: This indicates the state of charge of the energy storage system at the start of the scheduling cycle, which is reported in real time by the BMS (Battery Management System).

[0052] Total energy storage capacity: (kWh) refers to the nominal available energy capacity of the energy storage system, used for normalization in the calculation of State of Charge (SOC).

[0053] Rated energy storage capacity: (kW) refers to the maximum charging and discharging power of the energy storage converter (PCS), which is constrained. ; For energy storage discharge power, Power for charging energy storage.

[0054] Charging efficiency: A typical value of 0.90–0.98 indicates the efficiency of converting electrical energy into chemical energy.

[0055] Discharge efficiency: A typical value of 0.90–0.98 indicates the efficiency of converting chemical energy into electrical energy.

[0056] SOC security boundaries for each operating mode:

[0057] SOC security boundary for working mode 0 (self-use):

[0058] SOC safety boundary for operating mode 1 (grid feed):

[0059] SOC safety boundary for operating mode 2 (grid backup power):

[0060] The SOC safety boundary for each operating mode is provided by the energy storage device, ensuring battery life and safety under different operating strategies.

[0061] Power limitations of energy storage devices feeding into the grid: (kW) refers to the maximum power that can be fed to the grid as permitted by the grid connection agreement, which is usually a percentage of the rated power of photovoltaic or energy storage (e.g., 50%).

[0062] Power supply limitations for energy storage devices: (kW) refers to the maximum power output of the energy storage system when it is configured to supply power, which may be less than the rated power of the energy storage system. .

[0063] The aforementioned set of physical parameters is configured once during system initialization or dynamically read through the device communication protocol, serving as hard constraint parameters for the optimization model to ensure that the generated scheduling instructions are physically executable.

[0064] This application, after acquiring historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the energy storage system, inputs these data into a prediction model, outputting a photovoltaic predicted power sequence and a load predicted power sequence within the scheduling period. , Represents the future time period The expected photovoltaic power (kW) ranges from 100 to 1000. Load forecast power sequence , Represents the future time period The expected user electricity demand (kW) is calculated. Before inputting the data into the prediction model, the raw data is cleaned (outliers are removed and missing values ​​are imputed), normalized (Min-Max or Z-score), and time-frequency aligned (uniform sampling rate) to ensure the quality of the data input. If meteorological data for a certain period is missing, nearest neighbor interpolation or climate average value is used as a substitute.

[0065] Step S202: In the day-ahead phase, based on the photovoltaic predicted power, the load predicted power, and the constraints, a baseline scheduling strategy for the scheduling period is obtained.

[0066] The day-ahead phase involves performing optimization calculations 24 hours in advance to formulate the optimal operating plan for the next day (scheduling cycle). In this application, based on the predicted photovoltaic power, the predicted load power, and constraints, the scheduling strategy for the next day (scheduling cycle) is solved, serving as the baseline scheduling strategy for the intra-day layer. The baseline scheduling strategy includes the operating mode and grid interaction power. The timing of mode switching and grid interaction power for the next day are determined through the baseline scheduling strategy.

[0067] In some embodiments, the constraints include at least: power balance constraints, energy storage dynamic constraints, power limit constraints, mode-dependent constraints, and state-of-charge safety boundary constraints.

[0068] Constraint 1: Power balance constraint (user power priority principle).

[0069]

[0070] The left side of the equation represents the total power supply. Photovoltaic power generation capacity, For energy storage discharge power, The power purchased from the grid is the total power consumption; the right side of the equation represents the total electricity consumption. For user load power, For energy storage charging power, The power output sold to the power grid.

[0071] Core principles: User load must be 100% met without interruption; energy storage charging and discharging are mutually exclusive.

[0072] Constraint 2: Dynamic constraints on energy storage.

[0073]

[0074] Key improvement: Introducing total energy storage capacity Normalize the energy term to make the SOC unit %.

[0075] Efficiency modeling: Energy loss during charging is The energy requirement during discharge is It accurately reflects energy conversion losses; initial conditions: (Provided by input parameters); This equation ensures the conservation of stored energy, taking into account charging and discharging losses.

[0076] Constraint 3: Power limit constraint (equipment physical boundary).

[0077]

[0078] This refers to the rated power of the energy storage (input parameter). To limit the power of the feeder grid, thus constraining the charge and discharge rates;

[0079]

[0080]

[0081] Real-time grid acceptance capability;

[0082] Rated energy storage capacity.

[0083] Constraint 4: Schema-dependent constraints (logical mutual exclusion).

[0084] Mode 0 (Self-initiated and self-used): → Operate offline as much as possible;

[0085] Mode 1 (Power Supply): →Only out, no in;

[0086] Mode 2 (Backup Power): →Only in, no out.

[0087] This constraint is implemented in MILP using the Big M method or binary variables.

[0088] Constraint 5: SOC security boundary constraint (mode coupling dynamic boundary).

[0089]

[0090] The boundaries are provided by the input parameters, for example:

[0091] Mode 0: , (Conservative approach);

[0092] Mode 1: , (Deep discharge electricity sales);

[0093] Mode 2: , (Maximize charging);

[0094] This constraint dynamically adjusts the safety margin, releasing scheduling flexibility while ensuring equipment lifespan.

[0095] Among them, working mode variables Mode 0 (Self-consumption): Under the premise of meeting local load, the photovoltaic + energy storage system prioritizes photovoltaic power supply, suitable for periods of stable electricity prices; Mode 1 (Grid Feed): After meeting local load, excess electricity (PV + energy storage discharge) is sold to the grid, suitable for periods of high electricity prices; Mode 2 (Grid Backup): When PV is insufficient and the energy storage SOC is low, electricity is purchased from the grid to supplement the load or charge the energy storage, suitable for periods of low electricity prices or high load. The choice of operating mode not only affects grid interaction but also dynamically constrains the SOC safety boundary, reflecting the "mode-state" coupling. Grid interaction power includes: power sold to the grid. The unit is kW, and the power purchased from the grid. The unit is kW.

[0096] Step S203: During the intraday phase, based on the deviation between real-time photovoltaic power data and photovoltaic predicted power within the rolling window, and the deviation between real-time load power data and load predicted power, the baseline scheduling strategy for the current scheduling period is corrected to obtain the optimal scheduling strategy for the current scheduling period; and, based on the optimal scheduling strategy, an execution instruction is generated and the execution instruction is sent to the bidirectional energy storage converter.

[0097] The intraday phase refers to the dynamic adjustment of the day-ahead plan based on real-time collected photovoltaic power and load power data during the day's operation to address deviations and uncertainties in actual operation. Since day-ahead forecasts contain errors, high-frequency corrections are needed for grid interaction power in the dispatch strategy. However, full-time re-optimization involves large computational loads; therefore, this application adopts a rolling window + local optimization strategy. Rolling window length... (Each window contains 8 time steps, each time step is 15 minutes, and the total window duration is 2 hours), balancing foresight and computational burden; rolling step size. Minutes, consistent with the scheduling granularity.

[0098] The operating mode variable and the grid interaction power are optimization variables for the intraday stage. This application corrects the baseline scheduling strategy for the current scheduling period based on the deviation between the real-time photovoltaic power data and the photovoltaic predicted power in the rolling window, as well as the deviation between the real-time load power data and the load predicted power. At the same time, in each scheduling period t, the mode that maximizes the immediate benefit is selected as the optimal operating mode. In each scheduling period t, the corrected grid interaction power and operating mode are sent to the bidirectional energy storage converter. The bidirectional energy storage converter automatically allocates the power flow of photovoltaic, energy storage and grid according to the grid interaction power and operating mode without manual intervention.

[0099] This application adopts a multi-timescale collaborative optimization scheme. In the day-ahead phase, a baseline scheduling strategy is obtained based on the photovoltaic predicted power, the load predicted power, and the constraints. In the intraday phase, the baseline scheduling strategy in the day-ahead phase is corrected based on the deviation between the real-time photovoltaic power data and the photovoltaic predicted power in the rolling window, as well as the deviation between the real-time load power data and the load predicted power, thereby improving the long-term stability and accuracy of the collaborative scheduling of the photovoltaic and energy storage system.

[0100] In one embodiment, obtaining the baseline scheduling strategy within the scheduling period based on the photovoltaic predicted power, the load predicted power, and constraints includes the following steps:

[0101] Step S301: Based on explicit electricity sales revenue, explicit electricity purchase expenditure, and implicit electricity savings, construct an objective function to maximize the overall revenue within the scheduling cycle.

[0102] The objective function is designed as the sum of three parts of revenue: explicit electricity sales revenue, explicit electricity purchase expenditure, and implicit electricity savings, comprehensively reflecting the economic value. The expression of the objective function is as follows:

[0103]

[0104] in, This is explicit electricity sales revenue, which comes directly from cash revenue generated by the power grid; Explicit electricity purchase expenditure refers to the electricity fee paid to the power grid. Savings on hidden electricity consumption, specifically the cost savings for users using self-generated electricity (PV + energy storage) instead of purchasing electricity from the grid, are equivalent to avoiding the expense of purchasing electricity. For example, if the load is 10kW, and 6kW is met by PV, then the savings are... This design avoids underestimating the value of self-use using traditional methods, making the optimization closer to the real economic logic. The unit price of electricity sold to the grid during time period t. For time period Electricity sold to the grid For time period The unit price of electricity purchased from the grid, For time period Power purchased from the power grid, For time period Forecasted load power For time period The predicted photovoltaic power, This represents the energy storage discharge power during time period t. Time granularity. For the scheduling period length (e.g., 15 minutes), multiply all power by Converted to electricity volume (kWh), matching the electricity price unit.

[0105] Step S302: Input the photovoltaic predicted power, the load predicted power, and the constraints into the mixed integer linear programming model, solve the objective function, and obtain the baseline scheduling strategy within the scheduling period.

[0106] The modeling details of the mixed-integer linear programming model are as follows:

[0107] Objective function linearization: To represent a non-linear term, an auxiliary variable is introduced. Indicate self-consumption of electricity, and add constraints:

[0108]

[0109] Under the goal of maximization Automatic equals .

[0110] Pattern variable encoding: Introducing three sets of binary variables ,satisfy:

[0111]

[0112] Pattern constraints are implemented using the Big M method, for example:

[0113]

[0114] in For a sufficiently large constant (such as) ).

[0115] SOC boundary coupling: Similarly, the upper limit is handled.

[0116] Efficiency and Capacity Integration: In the Evolution Equation of SOC Embedded as constant parameters, ensuring physical accuracy of the model.

[0117] At the current stage, the solution process using global optimization through mixed integer linear programming (MILP) is as follows:

[0118] Input preparation: Load the day-ahead forecast 24-hour electricity price and all physical parameters ( (SOC boundaries for each mode, power limits, etc.)

[0119] Model building: generating models containing A binary variable, One continuous variable A mixed-integer linear (MILP) model with several constraints.

[0120] Solver call: Use solvers such as PuLP / OR-Tools.

[0121] Output: Global optimal solution This serves as the baseline strategy for the intraday layer.

[0122] In one embodiment, the process of correcting the baseline scheduling strategy for the current scheduling period based on the deviation between real-time photovoltaic power data and photovoltaic predicted power within the rolling window, and the deviation between real-time load power data and load predicted power, to obtain the optimal scheduling strategy for the current scheduling period includes the following steps:

[0123] Step S301: Based on the cumulative deviation between the real-time photovoltaic power data and the photovoltaic predicted power in the rolling window, the photovoltaic predicted power for the current scheduling period is corrected to obtain the photovoltaic predicted power correction value; and based on the cumulative deviation between the real-time load power data and the load predicted power in the rolling window, the load predicted power for the current scheduling period is corrected to obtain the load predicted power correction value.

[0124] This step improves short-term scheduling accuracy by dynamically correcting the predictive model output through real-time monitoring of deviations between actual operation and plan. Its core is the exponential smoothing compensation algorithm, which is simple, efficient, and suitable for embedded deployment.

[0125] Photovoltaic prediction error:

[0126]

[0127] Load forecasting error:

[0128]

[0129] in, This refers to real-time photovoltaic power data for time period t. Let t be the predicted photovoltaic power for time period t. This refers to the real-time load power data for time period t. This represents the predicted load power for time period t. (Real-time data) and Data is collected in real time by on-site sensors (electricity meters, power meters).

[0130] Error characteristics are usually time-dependent (e.g., consecutive cloudy days lead to a continuous negative deviation), and dynamic tracking is required.

[0131] This application employs first-order exponential smoothing, updating subsequent predicted values ​​based on accumulated errors:

[0132]

[0133]

[0134] in, for Correction values ​​for photovoltaic power forecasts for the specified period. for Load forecast power correction values ​​for different time periods for Forecasted photovoltaic power for the period for Forecasted load power for the time period and For smoothing coefficients, For scrolling windows The cumulative deviation between real-time photovoltaic power data and photovoltaic power prediction data for the specified time period and previous periods. For scrolling windows The cumulative deviation between real-time load power data and predicted load power for the specified time period and previous periods.

[0135] Smoothing coefficient , Control the photovoltaic error correction rate, with a typical value of 0.2–0.5 (higher values ​​are used during sudden weather changes); Control the load error correction rate, typical value 0.1–0.3 (lower value for relatively stable loads); smoothing coefficient It can be automatically updated based on the cumulative revenue within the statistical time window.

[0136] Compensated predicted value Need to be cut to the physical area To prevent invalid values, the compensated predicted value As an input for intraday optimization, the power grid interaction is corrected to form a closed loop of "execution → monitoring → compensation → re-optimization", which significantly improves the robustness of intraday dispatch.

[0137] This application only corrects forecasts within the intraday rolling window (e.g., the next 2 hours), and does not affect day-ahead forecasts. If actual photovoltaic output is lower than the forecast ( If the forecast is positive, then the subsequent forecast value should be lowered, the electricity sales plan should be reduced, and excessive discharge of energy storage should be avoided; conversely, if the forecast is negative, then the subsequent forecast value should be lowered, the electricity sales plan should be reduced, and excessive discharge of energy storage should be avoided.

[0138] Step S302: Based on the photovoltaic power forecast correction value, load power forecast correction value, real-time photovoltaic power data, and real-time load power data for the current scheduling period, the baseline scheduling strategy for the current scheduling period is corrected to obtain the optimal grid interaction power for the current scheduling period. The calculation formula for the optimal grid interaction power for the current rolling step is as follows:

[0139]

[0140] in, for Time-corrected grid interaction power for Uncorrected grid interaction power during the time period for Net power deviation during the time period.

[0141] Specifically, the formulas for calculating the revised power sold to the grid and the revised power purchased from the grid are as follows:

[0142]

[0143] in, for Time-adjusted power output to the grid for Uncorrected power sales to the grid during the specified time period for Time-adjusted power purchase from the grid for Uncorrected power purchased from the grid during the specified time period.

[0144]

[0145] in, for Real-time photovoltaic power data for the time period for Correction values ​​for photovoltaic power forecasts for the specified period. for Real-time load power data for the time period for The power forecast correction value for the time period. The power forecast correction value for photovoltaic power and the power forecast correction value for load are calculated by the method described in step S301.

[0146] Step S303: Based on the energy storage state of charge during the current scheduling period and the state of charge safety boundary under each working mode, determine all feasible working modes, and select the working mode with the greatest immediate benefit as the optimal working mode for the current scheduling period.

[0147] Specifically, in each scheduling period From all feasible modes Choose the mode that maximizes immediate benefits:

[0148]

[0149] Feasible pattern set Determined by the current SOC and the mode boundary of the input:

[0150]

[0151] For example, if And the input parameters are If so, then mode 1 is feasible.

[0152] For each feasible pattern Assuming this model is adopted, the calculation is instantaneous. :

[0153]

[0154] in, The power balance constraint equations are solved in mode m and are subject to... Restrictions, etc.

[0155] Then, the working mode that yields the greatest immediate benefit is selected as the optimal working mode for the current scheduling period.

[0156] Step S304: Based on the optimal power grid interaction power and the optimal operating mode, obtain the optimal scheduling strategy for the current scheduling period.

[0157] This application avoids the local optimum of a single threshold scheduling scheme by using a joint decision-making mechanism of multi-mode and execution power, and directly ensures the maximization of global benefits by controlling the power purchased and sold.

[0158] In some embodiments, day-ahead rescheduling can be triggered in the event of a major event, such as a detected sudden change in electricity prices (e.g., a surge in spot market prices), equipment failure (e.g., energy storage outage), or an extreme weather warning. This includes interrupting the current day-ahead schedule; rerunning the mixed-integer linear programming model (MILP) to solve the problem and generate a new strategy; and notifying the user or the superior dispatch center via a communication interface.

[0159] In one embodiment, the method further includes: constructing a reward function based on the cumulative revenue within a statistical time window, and updating the state-of-charge safety boundary for each operating mode according to the feedback of the reward function.

[0160] This application uses the cumulative revenue within a statistical time window as the optimization objective and automatically adjusts key threshold parameters. The updated key threshold parameters include the SOC safety boundary (…). ) and error compensation coefficient ( Traditionally, using a fixed SOC safety boundary (e.g., 20%–90%) may be too conservative; this application dynamically adjusts the boundary based on historical SOC distribution and equipment aging data. For example, if historical data shows that the equipment failure rate increases when SOC < 15%, then the boundary is increased. If returns are significant during periods of high SOC, then the restrictions can be appropriately relaxed. Adjust the SOC boundary to comply with the equipment's hard constraints. Error compensation coefficient. Adjust the smoothing strength based on the autocorrelation of historical prediction errors; for example, increase the smoothing strength if the errors consistently move in the same direction. Accelerate the correction.

[0161] This application employs an update mechanism based on Online Gradient Descent (OGD) / Proximal Policy Optimization (PPO):

[0162] Reward function: Set a rolling statistical time window (e.g., the three months prior to the update) to calculate the long-term cumulative returns within the statistical time window. As a reward;

[0163] Strategy parameters: ;

[0164] Updated formula:

[0165]

[0166] Gradient estimation: through perturbation parameters And observe the changes in returns; learning rate The initial value is 0.01, which decreases with each iteration to ensure convergence. For gradient.

[0167] Deployment method: Run an update once a day at midnight, utilizing idle computing resources; after the update, the parameters are written to the configuration file for scheduling the next day.

[0168] Safety mechanism: Parameter updates are subject to hard boundary constraints (e.g.) This is to prevent learning from diverging and causing equipment damage.

[0169] Efficiency parameters It is usually considered a constant, but in reality it varies with SOC and temperature. In some embodiments, the effective efficiency is estimated online based on charge and discharge energy measurement data to optimize efficiency parameters.

[0170] The embodiments of this application continuously optimize the relevant scheduling strategy parameters based on historical operating data, thereby improving the accuracy of the scheduling strategy over time.

[0171] This application also provides a collaborative scheduling system for a photovoltaic and energy storage system, such as... Figure 3 As shown, the system includes: a prediction module 10, an acquisition module 20, a correction module 30, and an execution module 40.

[0172] The prediction module 10 is used to predict the photovoltaic power forecast and load forecast within the scheduling period based on historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the photovoltaic-storage system. The acquisition module 20 is used to obtain the baseline scheduling strategy within the scheduling period based on the photovoltaic power forecast, the load forecast, and the constraint conditions during the day-ahead phase. The correction module 30 is used to correct the baseline scheduling strategy for the current scheduling period based on the deviation between the real-time photovoltaic power data and the photovoltaic power forecast within the rolling window, and the deviation between the real-time load power data and the load forecast, during the intraday phase, to obtain the optimal scheduling strategy for the current scheduling period. The execution module 40 is used to generate an execution command based on the optimal scheduling strategy and send the execution command to the bidirectional energy storage converter.

[0173] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application, and are devices corresponding to the above-mentioned optical storage system collaborative scheduling method. All implementation methods in the above-mentioned method embodiments are applicable to the embodiments of this device. For details on its specific functions and the technical effects it brings, please refer to the method embodiments section, which will not be repeated here.

[0174] In a preferred embodiment, the structural block diagram of a photovoltaic-storage system collaborative scheduling system provided in this application is as follows: Figure 4 As shown, the system includes: a multi-source data input and prediction generation module, a multi-timescale collaborative optimization module, a multi-mode adaptive switching and execution power generation module, an execution deviation feedback and error compensation module, and a strategy self-updating module.

[0175] The aforementioned multi-source data input and prediction generation module serves as the information entry point and perception center of the entire scheduling system. It is responsible for the structured integration of raw information from different data sources and drives the prediction model to generate photovoltaic power generation and user load power sequences for future periods. Its core task is to ensure the integrity, timeliness, and accuracy of the input data, providing a high-quality predictive foundation for subsequent optimization decisions and providing the optimization model with complete system physical parameter boundaries. Input variables include: historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and hard constraint parameters of the energy storage system. Output variables include: photovoltaic predicted power sequence and load predicted power sequence.

[0176] The aforementioned multi-timescale collaborative optimization module is the core decision engine of the scheduling system. It is responsible for solving the optimal scheduling strategy based on predictive information and complete physical constraints (including charge / discharge efficiency, SOC boundary, power limits, etc.) at both the Day-Ahead (DA) and Intra-Day (ID) timescales. Its innovation lies in constructing a two-layer nested optimization framework. The upper layer pursues global economic efficiency, while the lower layer ensures real-time robustness. These two layers work closely together through hot start, mode freezing, and deviation feedback mechanisms. All physical parameters are rigorously modeled to ensure the feasibility of the solution. Optimization variables include: operating mode variables and grid interaction power. The solution algorithm of the multi-timescale collaborative optimization module is the core technology for maximizing overall benefits. This algorithm adopts a two-layer architecture of "day-Ahead global optimization + intra-day rolling correction," balancing robustness and real-time performance. Day-Ahead global optimization is achieved through mixed integer linear programming (MILP), and the day-Ahead output... Store in shared memory; read from this memory within the day, add real-time deviations, and correct using a rolling window and local optimization strategy before issuing instructions to the execution layer, forming a closed loop.

[0177] The daytime optimization runs during low-load nighttime hours, consuming server resources; the intraday optimization runs on edge controllers (such as EMS hosts) to ensure low latency. The coordination logic is as follows: the intraday layer does not change the long-term strategy direction of the daytime layer, only correcting short-term execution deviations; if a major event is detected intraday (such as sudden changes in electricity prices or equipment failure), daytime rescheduling can be triggered. The two-layer coordination mechanism is shown in Table 1.

[0178] Table 1

[0179]

[0180] The aforementioned multi-mode adaptive switching and power generation module transforms the optimization results into executable device control commands. Its core lies in the adaptability of mode selection and the accuracy of power commands. Its innovation lies in treating mode switching as a joint benefit-safety optimization problem, rather than a simple rule-based judgment, and strictly adhering to the physical constraints of the input. Switching logic: In each scheduling period... The system evaluates all feasible modes. Select the mode that maximizes immediate benefits. The module outputs a working mode instruction and an execution power instruction. (Working mode instruction) The power command is transmitted to the energy storage converter PCS and the photovoltaic inverter via the protocol; the power command is executed as follows:

[0181]

[0182] Execute power command The value is a scalar (kW), with a positive value indicating output to the grid, a negative value indicating input from the grid, and 0 indicating no interaction; each module... Updated every minute, synchronized with the scheduling granularity; the module verifies before the instruction is issued. Does it meet the requirements? , Input limits are set; if these limits are exceeded, the circuit is truncated to the boundary value. The energy storage converter (PCS) is based on... and It automatically allocates power flow between photovoltaics, energy storage, and the power grid without human intervention.

[0183] The aforementioned deviation feedback and error compensation module constitutes the "sensing-correction" link of the closed-loop control. By monitoring the deviation between actual operation and plan in real time, it dynamically corrects the output of the prediction model, improving short-term scheduling accuracy. Its core is the exponential smoothing compensation algorithm, which is simple, efficient, and suitable for embedded deployment.

[0184] Photovoltaic prediction error:

[0185]

[0186] Load forecasting error:

[0187]

[0188] in, This refers to real-time photovoltaic power data for time period t. Let t be the predicted photovoltaic power for time period t. This refers to the real-time load power data for time period t. This represents the predicted load power for time period t. Real-time data is acquired in real time by field sensors (electricity meters, power meters).

[0189] Subsequent forecasts are updated using first-order exponential smoothing:

[0190]

[0191]

[0192] in, for Correction values ​​for photovoltaic power forecasts for the specified period. for Load forecast power correction values ​​for different time periods for Forecasted photovoltaic power for the period for Forecasted load power for the time period and For smoothing coefficients, For scrolling windows The cumulative deviation between real-time photovoltaic power data and photovoltaic power prediction data for the specified time period and previous periods. For scrolling windows The cumulative deviation between real-time load power data and predicted load power for the specified time period and previous periods.

[0193] After compensation Need to be cut to the physical area To prevent invalid values, the prediction after compensation As input to the intraday optimization module, it forms a closed loop of "execution → monitoring → compensation → re-optimization", significantly improving the robustness of intraday scheduling.

[0194] The aforementioned strategy self-updating module enables the system to "learn over a long period," continuously optimizing scheduling strategy parameters based on historical operational data to improve system performance over time. Its core is an online learning framework that uses cumulative revenue as the optimization objective and automatically adjusts key thresholds, including the input SOC boundary and efficiency-related parameters.

[0195] The collaborative scheduling of the photovoltaic and energy storage system in this application adopts a closed-loop architecture of "prediction-optimization-decision-feedback". It integrates day-ahead global optimization and intraday rolling correction, directly outputs the optimal power purchase / sale, and realizes online updating of strategy parameters through dynamic compensation of deviation. Thus, it takes into account economy, safety, compliance and long-term operational robustness under multiple constraints and multiple objectives.

[0196] In one embodiment, an optical storage system is provided, including a controller for implementing the methods described in any of the above embodiments.

[0197] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps in any of the above embodiments of the optical storage system collaborative scheduling method.

[0198] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0199] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0200] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for coordinated scheduling of a photovoltaic-energy storage system, wherein the photovoltaic-energy storage system includes an energy storage system, characterized in that, The method includes: Based on historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the energy storage system, the photovoltaic power forecast and load forecast during the scheduling cycle are predicted. In the day-ahead phase, based on the photovoltaic predicted power, the load predicted power, and the constraints, a baseline scheduling strategy for the scheduling period is obtained. During the intraday phase, based on the deviation between real-time photovoltaic power data and photovoltaic predicted power within the rolling window, and the deviation between real-time load power data and load predicted power, the baseline scheduling strategy for the current scheduling period is corrected to obtain the optimal scheduling strategy for the current scheduling period; and, based on the optimal scheduling strategy, an execution instruction is generated and the execution instruction is sent to the bidirectional energy storage converter.

2. The method according to claim 1, characterized in that, The hard constraint parameters of the energy storage system include: initial state of charge, total energy storage capacity, rated energy storage power, charging efficiency, discharging efficiency, state of charge safety boundary under each operating mode, grid power limit, and feeder power limit.

3. The method according to claim 1, characterized in that, The constraints include: power balance constraints, energy storage dynamic constraints, power limit constraints, mode-dependent constraints, and state-of-charge safety boundary constraints.

4. The method according to claim 1, characterized in that, The baseline scheduling strategy for the scheduling period, based on the photovoltaic power forecast, the load forecast, and the constraints, includes: Based on explicit electricity sales revenue, explicit electricity purchase expenditure, and implicit electricity savings, an objective function for maximizing comprehensive revenue within the scheduling cycle is constructed. The photovoltaic predicted power, the load predicted power, and the constraints are input into a mixed-integer linear programming model to solve the objective function, thereby obtaining the baseline scheduling strategy within the scheduling cycle.

5. The method according to claim 1, characterized in that, The baseline scheduling strategy includes the operating mode and grid interaction power. The optimal scheduling strategy for the current scheduling period is obtained by correcting the baseline scheduling strategy based on the deviation between real-time photovoltaic power data and predicted photovoltaic power within the rolling window, and the deviation between real-time load power data and predicted load power. Based on the cumulative deviation between real-time photovoltaic power data and photovoltaic power prediction within the rolling window, the photovoltaic power prediction for the current scheduling period is corrected to obtain the photovoltaic power prediction correction value; and based on the cumulative deviation between real-time load power data and load power prediction within the rolling window, the load power prediction for the current scheduling period is corrected to obtain the load power prediction correction value. Based on the photovoltaic power forecast correction value, load power forecast correction value, real-time photovoltaic power data, and real-time load power data for the current scheduling period, the grid interaction power for the current scheduling period is corrected to obtain the optimal grid interaction power for the current scheduling period. Based on the energy storage state of charge during the current scheduling period and the state of charge safety boundary under each working mode, all feasible working modes are determined, and the working mode with the greatest immediate benefit is selected as the optimal working mode for the current scheduling period. Based on the optimal power grid interaction power and the optimal operating mode, the optimal scheduling strategy for the current scheduling period is obtained.

6. The method according to claim 5, characterized in that, The formula for calculating the optimal grid interaction power for the current scheduling period is as follows: The base scheduling strategy for the current scheduling period is corrected based on the photovoltaic power forecast correction value, load power forecast correction value, real-time photovoltaic power data, and real-time load power data during the current scheduling period. in, for Time-corrected grid interaction power for Uncorrected grid interaction power during the time period for Net power deviation over the time period; ； in, for Real-time photovoltaic power data for the time period for Correction values ​​for photovoltaic power forecasts for the specified period. for Real-time load power data for the time period for The power correction value for load forecasting during the time period.

7. The method according to claim 6, characterized in that, The method further includes: A reward function is constructed based on the cumulative revenue within the statistical time window, and the state-of-charge safety boundary for each working mode is updated based on the feedback of the reward function.

8. A collaborative scheduling system for a photovoltaic-storage system, characterized in that, The system includes: The prediction module is used to predict the photovoltaic power and load power during the scheduling period based on historical photovoltaic power data, historical meteorological data, historical load data, grid electricity price data, and the hard constraint parameters of the photovoltaic and energy storage system. The acquisition module is used to obtain a baseline scheduling strategy for the scheduling period based on the photovoltaic predicted power, the load predicted power, and constraints during the day-ahead phase. The correction module is used to correct the baseline scheduling strategy for the current scheduling period during the intraday phase based on the deviation between the real-time photovoltaic power data and the photovoltaic power forecast within the rolling window, as well as the deviation between the real-time load power data and the load forecast, so as to obtain the optimal scheduling strategy for the current scheduling period. The execution module is used to generate execution instructions based on the optimal scheduling strategy and send the execution instructions to the bidirectional energy storage converter.

9. A photovoltaic energy storage system, comprising a controller, characterized in that, The controller is used to implement the method of any one of claims 1 to 7.

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

Images