Direct-control and non-direct-control flexible resource multi-time-scale cooperative regulation and control method

Abstract

The invention provides a direct-control and non-direct-control flexible resource multi-time-scale cooperative regulation and control method, and the method comprises the steps: predicting the power of non-adjustable energy and load in 24 hours of the next day through historical data in a day-ahead stage, and determining the expected response amount of a non-direct-control flexible resource; in the intra-day rolling stage, the power of the direct-control flexible resources is optimized by predicting the power of short-term non-adjustable energy and load, and the optimization result of the first regulation and control moment serves as a regulation and control instruction of the next regulation and control time period; and in the real-time feedback stage, on-line correction is carried out on a regulation and control instruction of the direct-control flexible resource based on a power measurement value of the direct-control flexible resource and power prediction of ultra-short-term non-adjustable energy and load. According to the invention, the influence caused by the response deviation of the non-direct-control flexible resources can be made up by regulating the output of the direct-control flexible resources, and the regulation effect is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, in particular to a multi-time scale coordinated regulation method of direct control and non-direct control flexible resources. BACKGROUND

[0002] In recent years, flexible resources are integrated into distribution networks in various forms, and can participate in regulation and control through demand response and other behaviors, effectively achieving power and energy balance of distribution systems. Flexible resources are divided into direct control flexible resources and non-direct control flexible resources, wherein the direct control flexible resources include controllable distributed power sources, energy storage and the like. The direct control flexible resources directly receive regulation and control instructions issued by the power grid, and have the characteristics of fast response, so they can participate in regulation and control in multi-time scales such as day-ahead and intra-day. The non-direct control flexible resources include interruptible loads, shiftable loads and the like. They rely on price signals, incentive mechanisms or protocols for autonomous adjustment. The response behavior of the non-direct control flexible resources depends on price guidance and incentives, and the adjustment speed is limited, so they can only be invited in day-ahead. However, due to economic incentives, user energy use habits, information acquisition and other factors, there may be a deviation between the actual response amount and the expected response amount of the non-direct control flexible resources. In the intra-day regulation stage, the response deviation caused by the non-direct control flexible resources is likely to cause poor regulation and control effect and affect the feasibility of the regulation and control scheme. Therefore, how to solve the influence of the response deviation of the non-direct control flexible resources on the regulation and control effect is an important problem to be solved. SUMMARY

[0003] The purpose of the present application is to provide a multi-time scale coordinated regulation method of direct control and non-direct control flexible resources, which can optimize the regulation and control effect.

[0004] To achieve the above technical purpose, the present application adopts the following technical scheme:

[0005] In a first aspect, the present application provides a multi-time scale coordinated regulation method of direct control and non-direct control flexible resources, wherein the non-direct control flexible resources include the power of interruptible loads, and the method comprises:

[0006] S1, day-ahead optimization: performing day-ahead optimization according to a preset day-ahead optimization period; each day-ahead optimization comprises: predicting the power of non-adjustable energy sources and loads in the next day; inputting the predicted power of the non-adjustable energy sources and loads in the next day into a day-ahead optimization regulation and control model, solving the day-ahead optimization regulation and control model, and obtaining the optimization value of each adjustable resource; wherein the adjustable resource includes direct control flexible resources and non-direct control flexible resources; the non-direct control flexible resources include the power of interruptible loads;

[0007] S2, intra-day rolling optimization: performing intra-day rolling optimization according to a preset intra-day rolling optimization period; each intra-day rolling optimization comprises: predicting the power of non-adjustable energy sources and loads in the current day; inputting the predicted power of the non-adjustable energy sources and loads in the current day into an intra-day rolling optimization regulation and control model, solving the intra-day rolling optimization regulation and control model, and obtaining the optimization value of each adjustable resource; wherein the adjustable resource includes direct control flexible resources and non-direct control flexible resources; the non-direct control flexible resources include the power of interruptible loads; The power at each control point within the time period is predicted; the predicted unadjustable energy and load are then... The power at each control point within the time period and the power optimization value of interruptible load obtained from the day-ahead optimization are input into the intraday rolling optimization control model. The intraday rolling optimization control model is solved to obtain... The optimized values ​​of each directly controlled flexible resource at each control moment within the time period, and the values ​​at the first control moment in the future. The optimized value at a given time is used as the control value for the next control period; where, Indicates the current moment. The preset intraday rolling optimization cycle, The number of periods included in the time window optimized for intraday rolling. The time window length is optimized for intraday rolling. Less than 24 hours;

[0008] S3. Perform real-time calibration according to the preset real-time calibration cycle. Each real-time calibration includes: calibration of non-adjustable energy sources and loads. Power is predicted within a specific time period; the predicted unadjustable energy sources and loads are then used in... The power during the time period, the power optimization value of interruptible loads obtained from the day-ahead optimization, the optimization value of each directly controlled flexible resource obtained from the intraday rolling optimization, and the actual power measurement values ​​of each directly controlled flexible resource and interruptible load are input into the real-time correction model. The real-time correction model is solved to obtain the adjustment amount of each directly controlled flexible resource. Based on the obtained adjustment amounts of each directly controlled flexible resource, the control values ​​of each directly controlled flexible resource obtained in the intraday rolling optimization stage are corrected. The preset real-time correction period is set to... , .

[0009] In one possible implementation, in S1, the optimized values ​​of each adjustable resource include the optimized values ​​of the charge and discharge states of energy storage and the optimized values ​​of the power of interruptible loads.

[0010] In S2, the predicted unadjustable energy and load are... The power at each control point within the time period and the optimized power value of interruptible load obtained from the previous day's optimization are input into the intraday rolling optimization control model to solve the intraday rolling optimization control model, including: inputting the predicted unadjustable energy and load at... The power at each control moment within the time period, the optimized values ​​of the charge and discharge state of energy storage obtained from the day-ahead optimization, and the optimized power values ​​of interruptible loads are input into the intraday rolling optimization control model to solve the intraday rolling optimization control model.

[0011] In S3, the predicted unadjustable energy and load are... The power during the time period, the power optimization value of interruptible loads obtained from day-ahead optimization, the optimization value of each directly controlled flexible resource obtained from intraday rolling optimization, and the actual power measurement values ​​of each directly controlled flexible resource and interruptible load are input into the real-time correction model to solve the real-time correction model, including: inputting the predicted unadjustable energy and load values ​​into the real-time correction model. The power during the time period, the optimized values ​​of the charge and discharge states of energy storage obtained by day-ahead optimization, the optimized values ​​of interruptible load power, the optimized values ​​of each directly controlled flexible resource obtained by intraday rolling optimization, and the actual power measurement values ​​of each directly controlled flexible resource and interruptible load are input into the real-time correction model to solve the real-time correction model.

[0012] In one possible implementation, S1 includes predicting the power of the non-adjustable energy source and load for the next day, including:

[0013] S1.1, Construct a sample dataset, which includes historical power data and meteorological data of non-adjustable energy sources and loads;

[0014] S1.2, Based on the sample dataset, train an LSTM-based day-ahead prediction model, enabling the day-ahead prediction model to predict the power of unadjustable energy sources and loads in the next 24 hours based on historical power data and meteorological data of unadjustable energy sources and loads in the previous day.

[0015] S1.3, based on historical power data and meteorological forecast data of unadjustable energy sources and loads in the near future, use a trained day-ahead prediction model to predict the power of unadjustable energy sources and loads in the next 24 hours.

[0016] In one possible implementation, the directly controlled flexible resources include the exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, and the charging and discharging power of energy storage.

[0017] In S1, the objective function of the daytime optimization control model is... The expression is as follows:

[0018] ;

[0019] In the formula: The operating cost of the distribution network system for the next day, for The cost of power exchange between the distribution network and the main grid at any given time; , , They represent Time of the first Taiwan's controllable distributed power generation cost, Taiwan's energy storage operating costs, The response cost of interruptible loads; This represents the total number of hours for the next day. , and These represent the number of nodes for controllable distributed power sources, energy storage, and interruptible loads, respectively.

[0020] in, , , , Based on respectively Power exchange between the distribution network and the main grid at all times , Time of the first Power generation capacity of Taiwan's controllable distributed power source , Time of the first Taiwan's energy storage charging and discharging power , Time of the first Power reduction for interruptible loads calculate;

[0021] The constraints of the optimized control model include power balance constraints, power generation constraints of controllable distributed power sources, ramp-up constraints of controllable distributed power sources, SOC and charge / discharge constraints of energy storage, and interruptible load constraints.

[0022] In one possible implementation, the objective function of the intraday rolling optimization control model in S2 is... The expression is as follows:

[0023] ;

[0024] In the formula: for Operating costs of the distribution network system during the time period For the current moment, , , intraday The cost of power exchange between the distribution network and the main grid at any time, the first Taiwan's controllable distributed power generation cost, intraday Time of the first Taiwan's energy storage operating costs.

[0025] The model aims to minimize the system operating cost.

[0026] The constraints of the intraday rolling optimization control model include power balance constraints, power generation constraints of controllable distributed power sources, ramping constraints of controllable distributed power sources, and SOC and charge / discharge constraints of energy storage.

[0027] In one possible implementation, the objective function of the real-time correction model in S3 is:

[0028] ;

[0029] In the formula: This refers to the amount of flexible resource adjustment that can be directly controlled at any given moment. For adjustable resources The adjustment amount; It represents the collection of all adjustable resources, including the exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, the power of energy storage, and the response power of interruptible loads; Available resources at the current moment The actual power measurement value; The optimized values ​​of adjustable resources obtained by the intraday rolling optimization model or the day-ahead optimization model, including the optimized exchange values ​​between the distribution network and the main grid. The optimized generation values ​​and optimized energy storage values ​​of controllable distributed power sources are obtained by solving an intraday rolling optimization model, while the optimized power value of interruptible loads is obtained by solving an intraday rolling optimization model. The solution was obtained through the optimization model.

[0030] The constraints of the real-time correction model include: requiring the power of the directly controlled flexible resources to be corrected. It satisfies power balance constraints, power generation constraints of controllable distributed power sources, ramping constraints of controllable distributed power sources, and SOC and charge / discharge constraints of energy storage.

[0031] In a second aspect, this application provides an electronic device, including: a memory and a processor;

[0032] The memory is used to store computer programs;

[0033] The processor is used to invoke the computer program to execute the method described above.

[0034] Thirdly, this application provides a computer-readable storage medium storing a computer program that, when run on an electronic device, causes the electronic device to perform the method described above.

[0035] Fourthly, this application provides a computer program product, including a computer program that, when run on an electronic device, causes the electronic device to perform the method described above.

[0036] The specific implementation methods of the second to fourth aspects of this application can refer to the implementation methods of the first aspect, and will not be elaborated here.

[0037] Beneficial effects:

[0038] 1. To address the issue of discrepancies between the actual and expected responses of non-directly controlled flexible resources, which affect the intraday control effect, direct-controlled and non-directly controlled flexible resources are coordinated. By correcting the control commands of direct-controlled flexible resources online, the impact of response deviations of non-directly controlled flexible resources is compensated, thereby achieving power balance in the distribution network.

[0039] 2. The constructed three-stage control framework fully considers multiple time scales, including day-ahead, intraday, and real-time, and multi-period power prediction can effectively reduce the prediction error of wind and solar power. Furthermore, correction is introduced in the real-time stage to dynamically modify the control commands of directly controlled flexible resources with the goal of minimizing the adjustment amount. In the ultra-short term, the power of directly controlled flexible resources is continuously corrected, thereby compensating for the response deviation of non-directly controlled flexible resources and optimizing the control effect. Attached Figure Description

[0040] Figure 1 This is a flowchart of an embodiment of this application. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present application, the technical solution of the present application will be further described in detail below with reference to the embodiments and accompanying drawings.

[0042] To address the issue that response deviations caused by non-directly controlled flexible resources may lead to poor control performance, this application proposes a multi-timescale collaborative control method for both directly controlled and non-directly controlled flexible resources. The method constructs a control framework comprising three stages: day-ahead optimization, intraday rolling optimization, and real-time correction. In the day-ahead stage, historical data is used to predict the power of unadjustable energy sources and loads for the next 24 hours, determining the expected response of non-directly controlled flexible resources. In the intraday rolling stage, the power of directly controlled flexible resources is optimized by predicting the power of short-term (e.g., 4 hours) unadjustable energy sources and loads. The optimization result at the first control time is used as the control command for the next control period (e.g., 15 minutes), shifting the time window for rolling optimization. In the real-time feedback stage, based on the power measurements of directly controlled flexible resources and the power predictions of ultra-short-term unadjustable energy sources and loads, the control commands for directly controlled flexible resources are corrected online. By adjusting the output of directly controlled flexible resources, the impact of response deviations in non-directly controlled flexible resources is compensated, achieving power balance in the distribution network. Compared to traditional strategies, this invention enables collaborative control of both directly controlled and non-directly controlled flexible resources across multiple timescales.

[0043] The following will refer to Figure 1 A specific implementation method according to this application is described.

[0044] Example 1:

[0045] This application provides a method for coordinated regulation of flexible resources across multiple time scales using both direct and indirect control, including:

[0046] S1. Day-ahead optimization: Day-ahead optimization is performed according to the preset day-ahead optimization cycle. Each day-ahead optimization includes: predicting the power of non-adjustable energy sources and loads on the next day; inputting the predicted power of non-adjustable energy sources and loads on the next day into the day-ahead optimization control model, solving the day-ahead optimization control model, and obtaining the optimized value of each adjustable resource; wherein, the optimized value of each adjustable resource includes the optimized power value of interruptible loads.

[0047] Preferably, the preset daily optimization period is 24 hours; and the power of the next day is the power of the next 24 hours.

[0048] The day-ahead optimization, with a 24-hour optimization cycle, means that a day-ahead optimization is performed every 24 hours. Each day-ahead optimization makes optimization decisions regarding the status and power of flexible resources in the following 24 hours.

[0049] Non-adjustable energy sources include distributed photovoltaic (PV) and wind power. Because the output of distributed PV and wind power depends primarily on weather conditions and natural resource availability, and is difficult to artificially regulate according to the needs of the power system, they are considered non-adjustable energy sources.

[0050] Non-adjustable loads refer to those electrical loads in a power system that cannot or are difficult to adjust according to system demand. These loads typically have rigid demand characteristics, with relatively fixed electricity consumption times and amounts. They do not respond to price signals or dispatch instructions, creating a "hard constraint" on the operation of the power system, such as residential loads.

[0051] This step predicts the power of unadjustable energy sources and unadjustable loads on the next day, which will be used as known quantities in the subsequent power balance equations.

[0052] The adjustable resources include directly controlled flexible resources and non-directly controlled flexible resources. Directly controlled flexible resources include the generation capacity of controllable distributed power sources and the power of energy storage. Controllable distributed power sources include micro-generators that can be manually controlled. This application also includes the exchange power between the distribution network and the main grid in the optimization of directly controlled flexible resources. Non-directly controlled flexible resources include the power of interruptible loads.

[0053] In some embodiments, the optimized values ​​for each adjustable resource may also include optimized values ​​for the charge / discharge state of energy storage.

[0054] The optimized values ​​of the charge / discharge state of energy storage and the optimized values ​​of the interruptible load obtained in this step can be passed to the intraday rolling optimization stage for use in intraday rolling optimization.

[0055] Determining the energy storage status during the day-ahead optimization phase can provide a benchmark for subsequent regulation: during the intraday rolling optimization phase, the energy storage status determined during the day-ahead optimization phase can be used as the initial condition and adjustment benchmark; at the same time, determining the energy storage status during the day-ahead optimization phase can reduce the number of optimization variables in the intraday rolling optimization phase, achieve dimensionality reduction solution, and optimize computational efficiency.

[0056] S2. Intraday Rolling Optimization: Intraday rolling optimization is performed according to a preset intraday rolling optimization cycle; each intraday rolling optimization includes: optimizing non-adjustable energy sources and loads within... The power at each control point within the time period is predicted; the predicted unadjustable energy and load are then... The power at each control point within the time period and the power optimization value of interruptible load obtained from the day-ahead optimization are input into the intraday rolling optimization control model. The intraday rolling optimization control model is solved to obtain... The optimized values ​​of each directly controlled flexible resource at each control moment within the time period, and the values ​​at the first control moment in the future. The optimized value at any given time is used as the control value (planned issuance value) for the next control period; among which, Indicates the current moment. The preset intraday rolling optimization cycle, The number of periods included in the time window optimized for intraday rolling. The time window length is optimized for intraday rolling. Less than 24 hours;

[0057] The historical data used for intraday rolling optimization can be the power of non-adjustable energy sources and loads during the same period in the historical dates that are the same as the period to be predicted.

[0058] Intraday rolling optimization, with The intraday rolling optimization cycle refers to every [period]. The time is used for intraday rolling optimization. Each intraday rolling optimization refers to the current moment... Constantly monitoring non-adjustable energy sources and loads Short-term power forecasts are made for each control moment within the time period to enable short-term optimization decisions for flexible resources.

[0059] In some embodiments, the optimized state of charge and discharge of energy storage obtained by day-ahead optimization can be used as a fixed value to input into the intraday rolling optimization control model, and then the intraday rolling optimization control model can be solved.

[0060] In some embodiments, It can be set to 2-6 hours.

[0061] Preferably, , , This allows for a balance between optimization effectiveness and efficiency.

[0062] S3. Perform real-time calibration according to the preset real-time calibration cycle. Each real-time calibration includes: calibration of non-adjustable energy sources and loads. Power is predicted within a specific time period; the predicted unadjustable energy sources and loads are then used in... The power during the time period, the optimized state of charge / discharge values ​​of energy storage obtained from day-ahead optimization, the optimized power values ​​of interruptible loads, the optimized values ​​of each directly controlled flexible resource obtained from intraday rolling optimization, and the actual power measurements of each directly controlled flexible resource and interruptible load are input into the real-time correction model. The real-time correction model is solved to obtain the adjustment amount for each directly controlled flexible resource. Based on the obtained adjustment amounts for each directly controlled flexible resource, the control values ​​of each directly controlled flexible resource obtained during the intraday rolling optimization phase are corrected (feedback correction). The preset real-time correction period is set to... , .

[0063] The real-time correction, The real-time correction period refers to every [period]. The time is adjusted in real time. Each real-time adjustment refers to the adjustment made at the current moment. Constantly monitoring non-adjustable energy sources and loads Power is predicted in the ultra-short term and the actual power measurements of each directly controlled flexible resource and interruptible load are obtained. The control values ​​of each directly controlled flexible resource obtained in the intraday rolling optimization phase are then corrected.

[0064] In some embodiments, the optimized state of charge and discharge values ​​of the energy storage obtained from the previous day's optimization can be used as fixed values ​​to input into the real-time correction model, and then the real-time correction model can be solved.

[0065] It should be understood that every Perform an intraday rolling optimization; every Perform a real-time correction until the end of the overall control cycle.

[0066] In some embodiments, in S1, predicting the power of unadjustable energy sources and loads for the next day includes:

[0067] S1.1, Construct a sample dataset, which includes historical power data and meteorological data of non-adjustable energy sources and loads;

[0068] In this step, the sample dataset can be divided, with 80% of the sample data used as the training set and the remaining 20% ​​used as the test set.

[0069] Considering seasonal similarity, the historical power data used for intraday forecasts can be power data of non-adjustable energy sources and loads from at least 2 years of historical data for the same period (e.g., ±15 days).

[0070] The data sampling interval can be 15 minutes.

[0071] S1.2, Based on the sample dataset, train an LSTM-based day-ahead prediction model, enabling the day-ahead prediction model to predict the power of unadjustable energy sources and loads in the next 24 hours based on historical power data and meteorological data of unadjustable energy sources and loads in the past few days (e.g., the past 3 days).

[0072] LSTM stands for Long Short-Term Memory.

[0073] It should be noted that the day-ahead forecasting model only needs to be retrained based on historical data when a new weather season arrives or when the error is detected to be greater than the preset threshold; otherwise, the day-ahead forecasting model is incrementally trained every day (e.g., at 23:00).

[0074] S1.3, based on the historical power data and meteorological forecast data (forecast data for the next 24 hours) of unadjustable energy and load in the near future (e.g., the last 3 days), use a trained day-ahead prediction model to predict the power of unadjustable energy and load in the next 24 hours.

[0075] By solving the day-ahead optimization control model for the predicted unadjustable energy and load power input in the next 24 hours, the optimized charging and discharging state values ​​of energy storage and the optimized power values ​​of interruptible loads can be obtained.

[0076] In some embodiments, in step S1, a day-ahead optimization control model is constructed with the goal of minimizing the operating cost of the distribution network system; the objective function of the day-ahead optimization control model is... The expression is as follows:

[0077]

[0078] In the formula: The operating cost of the distribution network system for the next day, for The cost of power exchange between the distribution network and the main grid at any given time; , , They represent Time of the first Taiwan's controllable distributed power generation cost, Taiwan's energy storage operating costs, The response cost of interruptible loads; This represents the total number of hours for the next day; the next day is 24 hours; for example, if calculated at fixed time intervals... Discretization results in a total of 24 time periods, corresponding to 24 time points; if calculated at fixed time intervals... Discretization results in a total of 96 time periods, corresponding to 96 time points. , and These represent the number of nodes for controllable distributed power sources, energy storage, and interruptible loads, respectively.

[0079] The formulas for calculating each cost are as follows:

[0080] 1) Cost of power exchange between the distribution network and the main grid:

[0081]

[0082] In the formula: for The power exchanged between the distribution network and the main grid at all times; for The electricity price at any given moment.

[0083] 2) Controllable distributed power generation cost:

[0084]

[0085] In the formula: for Time of the first The power generation capacity of the controllable distributed power source; Cost of distributed power generation.

[0086] 3) Energy storage operating costs:

[0087]

[0088] In the formula: for Time of the first Taiwan's energy storage charging and discharging power; , and These are the unit capacity installation cost of energy storage, the capital recovery factor, and the capacity factor; and These are the annual operating hours of energy storage and the operation and maintenance cost coefficient, respectively.

[0089] 4) Response cost of interruptible loads:

[0090]

[0091] In the formula: for Time of the first The power reduction (response quantity) that can interrupt the load. This is a compensation price for interruptible loads.

[0092] The constraints for optimizing the control model include:

[0093] 1) Power balance constraints:

[0094]

[0095] In the formula: and They represent Time of the first The discharge and charging power of the energy storage device; and The predicted results are as follows: The power of energy sources and loads that are not adjustable at all times; since load uncertainty is not considered, The value is constant; for Responding to the first The power of interruptible loads can be obtained by forecasting data from past time periods that were not involved in demand response.

[0096] 2) Power generation constraints of controllable distributed power sources:

[0097]

[0098] In the formula: , The first The minimum and maximum output power of the controllable distributed power source.

[0099] 3) Ramp-up constraints for controllable distributed power sources:

[0100]

[0101] In the formula: , The first The ramp and landslide rates of a controllable distributed power source. The time interval for regulation.

[0102] 4) Energy storage SOC and charge / discharge constraints:

[0103]

[0104] In the formula: , It is a binary variable representing the charging and discharging state (1 and 0 represent the charging and discharging states, respectively). , These represent the maximum discharge and charging power of the energy storage, respectively. , These represent discharge efficiency and charging efficiency, respectively. Indicates energy storage capacity; express Time of the first State of charge of energy storage; , These represent the minimum and maximum states of charge, respectively.

[0105] 5) Interruptible load constraints:

[0106]

[0107] In the formula: for Time of the first The maximum response amount that can be interrupted by the load.

[0108] The variables to be optimized in the objective function mainly include four types of key decision variables: 1) the exchange power between the distribution network and the main grid, with the corresponding variable being... The corresponding constraints are implicit in the power balance constraints; 2) The output of the controllable distributed power source, the corresponding variable is The corresponding constraints are the output range and ramping constraints of equations (7)-(8); 3) the charging and discharging power and charging and discharging state constraints of energy storage, and the corresponding variables are The corresponding constraints are the SOC and charge / discharge state constraints in equation (9); 4) the power reduction of interruptible loads, the corresponding variables are The corresponding constraint is the maximum response constraint of equation (10).

[0109] In some embodiments, the non-adjustable energy source and load in S2 are... The power at each control moment within the time period is predicted, including:

[0110] S2.1, Construct a sample dataset, which includes historical power data and meteorological data of non-adjustable energy sources and loads;

[0111] In this step, the sample dataset can be divided, with 80% of the sample data used as the training set and the remaining 20% ​​used as the test set.

[0112] Considering seasonal similarity, the historical power data used for intraday rolling optimization can be the power data of non-adjustable energy sources and loads from the same historical period (e.g., ±15 days).

[0113] The data sampling interval can be 15 minutes.

[0114] The sample dataset can be the sample dataset from step S1.1 plus the latest collected data.

[0115] S2.2, Based on the sample dataset, train an intraday rolling forecast model based on LSTM, enabling the intraday rolling forecast model to predict the status of unadjustable energy sources and loads based on historical power data and weather forecast data from the past few hours (e.g., the past 4 hours) of unadjustable energy sources and loads. The function of power at each control moment within the time period;

[0116] It should be noted that the intraday rolling forecast model only needs to be fine-tuned weekly or updated when the forecast error is detected to be greater than the preset threshold.

[0117] S2.3, based on historical power data and weather forecast data (future) of unadjustable energy sources and loads in the near future (e.g., the last 3 days). (Using time-period meteorological forecast data) a trained intraday rolling forecasting model to predict uncontrollable energy sources and loads. Power at each control moment within the time period.

[0118] The predicted unadjustable energy and load are in The power at each control point within the time period is input into the intraday rolling optimization control model. Solving the intraday rolling optimization control model yields the following results. The exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, and the power of energy storage are measured at each control moment within the time period, and the value at the first control moment is used as the control value for the next control period.

[0119] In some embodiments, the intraday rolling optimization control model still aims to minimize the operating cost of the distribution network system. The objective function of the intraday rolling optimization control model is... The expression is as follows:

[0120]

[0121] In the formula: for Operating costs of the distribution network system during the time period For the current moment, , , intraday The cost of power exchange between the distribution network and the main grid at any time, the first Taiwan's controllable distributed power generation cost, intraday Time of the first Taiwan's energy storage operating costs.

[0122] The model aims to minimize the system operating cost.

[0123] Since the optimal values ​​for the charge / discharge state of energy storage and the optimal values ​​for the power of interruptible loads have been determined during the day-ahead regulation phase, i.e., within the intraday rolling optimization regulation model... Fixed value , For fixed intraday rolling optimization control models, interruptible load constraints are not required. Other constraints (including power balance constraints, power generation constraints of controllable distributed sources, ramping constraints of controllable distributed sources, and SOC and charge / discharge constraints of energy storage) remain consistent with those of the day-ahead optimization control phase.

[0124] The regulation values ​​for the exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, and the power of energy storage are calculated as follows:

[0125]

[0126]

[0127]

[0128] In the formula: , , They are respectively The control values ​​(planned values) for the exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, and the charging and discharging power of energy storage. , , To obtain the solution Moment, that is The optimized values ​​of the exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, and the charging and discharging power of energy storage at the first control time after the time point.

[0129] In some embodiments, in S3, the non-adjustable energy source and load are... Power prediction within a time period includes:

[0130] S3.1, Construct a sample dataset, which includes historical power data and meteorological data of non-adjustable energy sources and loads;

[0131] In this step, the sample dataset can be divided, with 80% of the sample data used as the training set and the remaining 20% ​​used as the test set.

[0132] Considering seasonal similarity, the historical power data used in the real-time correction phase can be the power data of non-adjustable energy sources and loads from the same historical period (e.g., ±15 days).

[0133] The data is high-frequency sampled data; for example, the sampling interval can be 5 to 15 seconds.

[0134] S3.2, Based on the sample dataset, train an LSTM-based real-time prediction model, enabling the model to predict unadjustable energy sources and loads based on historical power data and weather forecast data from the past few minutes (e.g., the past 5 minutes) of unadjustable energy sources and loads. The function of power within a time period;

[0135] It should be noted that the real-time prediction model only needs to be fine-tuned hourly (with minor adjustments to the weights), updated weekly (with network architecture optimization), or updated when the prediction error is detected to be greater than a preset threshold.

[0136] S3.3, based on historical power data and weather forecast data (future) of unadjustable energy sources and loads over the past few minutes (e.g., the past 5 minutes). (Using time-period meteorological forecast data) to predict uncontrollable energy sources and loads using a trained real-time forecasting model. Power during the time period.

[0137] In some embodiments, real-time correction period The timeframe is 5-15 seconds. Therefore, predictions at the second level can be achieved.

[0138] In some embodiments, a real-time correction model is established with the objective of minimizing the adjustment amount of directly controlled flexible resources at the current moment; the objective function of the real-time correction model is:

[0139]

[0140] In the formula: For adjustable resources The adjustment amount is the solution variable of this model. Here, the adjustment amount of the adjustable resources to be solved are the adjustment amounts of each directly controlled flexible resource that can respond in real time, including the exchange adjustment amount between the distribution network and the main grid. Controllable distributed power generation adjustment Adjustment amount of energy storage ,Right now For interruptible loads whose adjustment speed is limited and cannot respond in real time, set their adjustment amount to 0. This represents the set of adjustable resources, including the generating capacity of controllable distributed power sources. Energy storage power Response power of interruptible load And the exchange power between the distribution network and the main grid. Also included in the optimization, that is ; Available resources at the current moment The actual power measurement value; The optimized values ​​of adjustable resources obtained by the intraday rolling optimization model or the day-ahead optimization model include the optimized exchange values ​​between the distribution network and the main grid. Optimal generation value of controllable distributed power sources Optimal value of energy storage Optimal value of interruptible load ,Right now Among them, the optimized exchange value between the distribution network and the main network Optimal generation value of controllable distributed power sources And the optimal value of energy storage The power optimization value for interruptible loads was obtained by solving the intraday rolling optimization model. The solution was obtained through the optimization model.

[0141] In this model, the power of the corrected direct-control flexible resources is required. Each of the constraints must be satisfied, namely, power balance constraint, power generation constraint of controllable distributed power source, ramping constraint of controllable distributed power source, and SOC and charge / discharge constraint of energy storage, i.e., formulas (7)-(9).

[0142] The corrected power of each directly controlled flexible resource is used as the new control value.

[0143] It should be understood that the above numbering, such as S1 to S3, is only used to distinguish and facilitate the expression of different steps or stages, and does not necessarily constitute a restriction on the execution order between the steps or stages.

[0144] Example 2:

[0145] This embodiment provides an electronic device, including: a memory and a processor;

[0146] The memory is used to store computer programs;

[0147] The processor is configured to invoke the computer program to execute the method as described in Embodiment 1.

[0148] Example 3:

[0149] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is run on an electronic device, it causes the electronic device to perform the method described in Embodiment 1.

[0150] Example 4:

[0151] This embodiment provides a computer program product, including a computer program that, when run on an electronic device, causes the electronic device to perform the method described in Embodiment 1.

[0152] The specific implementation of the system, electronic device, computer-readable storage medium, and computer program product provided in this application can be referred to the specific embodiments of the above methods, and will not be repeated here.

[0153] The technical content of the above embodiments can be referred to each other. For the same or similar technical features, appropriate omissions have been made in some embodiments to avoid repeated descriptions.

[0154] Obviously, those skilled in the art should understand that the various units or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps into a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0155] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for flexible multi-timescale coordinated regulation of resources using both direct and indirect control, characterized in that, The non-directly controlled flexible resources include the power of interruptible loads, and the method includes: S1. Day-ahead optimization: Day-ahead optimization is performed according to a preset day-ahead optimization cycle; each day-ahead optimization includes: predicting the power of non-adjustable energy sources and loads on the next day; inputting the predicted power of non-adjustable energy sources and loads on the next day into the day-ahead optimization control model, solving the day-ahead optimization control model, and obtaining the optimized value of each adjustable resource; wherein, the adjustable resources include directly controlled flexible resources and non-directly controlled flexible resources; non-directly controlled flexible resources include the power of interruptible loads; S2. Intraday Rolling Optimization: Intraday rolling optimization is performed according to a preset intraday rolling optimization cycle; each intraday rolling optimization includes: optimizing non-adjustable energy sources and loads within... The power at each control point within the time period is predicted; the predicted unadjustable energy and load are then... The power at each control point within the time period and the power optimization value of interruptible load obtained from the day-ahead optimization are input into the intraday rolling optimization control model. The intraday rolling optimization control model is solved to obtain... The optimized values ​​of each directly controlled flexible resource at each control moment within the time period, and... The optimized value at a given time is used as the control value for the next control period; where, Indicates the current moment. The preset intraday rolling optimization cycle, The number of periods included in the time window optimized for intraday rolling. The time window length is optimized for intraday rolling. Less than 24 hours; S3. Perform real-time calibration according to the preset real-time calibration cycle. Each real-time calibration includes: calibration of non-adjustable energy sources and loads. Power is predicted within a specific time period; the predicted unadjustable energy sources and loads are then used in... The power during the time period, the power optimization value of interruptible loads obtained from the day-ahead optimization, the optimization value of each directly controlled flexible resource obtained from the intraday rolling optimization, and the actual power measurement values ​​of each directly controlled flexible resource and interruptible load are input into the real-time correction model. The real-time correction model is solved to obtain the adjustment amount of each directly controlled flexible resource. Based on the obtained adjustment amounts of each directly controlled flexible resource, the control values ​​of each directly controlled flexible resource obtained in the intraday rolling optimization stage are corrected. The preset real-time correction period is set to... , .

2. The method according to claim 1, characterized in that, In S1, the optimized values ​​of each adjustable resource include the optimized values ​​of the charge and discharge states of energy storage and the optimized values ​​of the power of interruptible loads. In S2, the predicted unadjustable energy and load are... The power at each control point within the time period and the optimized power value of interruptible load obtained from the previous day's optimization are input into the intraday rolling optimization control model to solve the intraday rolling optimization control model, including: inputting the predicted unadjustable energy and load at... The power at each control moment within the time period, the optimized values ​​of the charge and discharge state of energy storage obtained from the day-ahead optimization, and the optimized power values ​​of interruptible loads are input into the intraday rolling optimization control model to solve the intraday rolling optimization control model. In S3, the predicted unadjustable energy and load are... The power during the time period, the power optimization value of interruptible loads obtained from day-ahead optimization, the optimization value of each directly controlled flexible resource obtained from intraday rolling optimization, and the actual power measurement values ​​of each directly controlled flexible resource and interruptible load are input into the real-time correction model to solve the real-time correction model, including: inputting the predicted unadjustable energy and load values ​​into the real-time correction model. The power during the time period, the optimized values ​​of the charge and discharge states of energy storage obtained by day-ahead optimization, the optimized values ​​of interruptible load power, the optimized values ​​of each directly controlled flexible resource obtained by intraday rolling optimization, and the actual power measurement values ​​of each directly controlled flexible resource and interruptible load are input into the real-time correction model to solve the real-time correction model.

3. The method according to claim 1, characterized in that, In step S1, the prediction of the power of non-adjustable energy sources and loads for the next day includes: S1.1, Construct a sample dataset, which includes historical power data and meteorological data of non-adjustable energy sources and loads; S1.2, Based on the sample dataset, train an LSTM-based day-ahead prediction model, enabling the day-ahead prediction model to predict the power of unadjustable energy sources and loads in the next 24 hours based on historical power data and meteorological data of unadjustable energy sources and loads in the previous day. S1.3, based on historical power data and meteorological forecast data of unadjustable energy sources and loads in the near future, use a trained day-ahead prediction model to predict the power of unadjustable energy sources and loads in the next 24 hours.

4. The method according to claim 1, characterized in that, The directly controlled flexible resources include the exchange power between the distribution network and the main grid, the power generation of controllable distributed power sources, and the charging and discharging power of energy storage. In S1, the objective function of the daytime optimization control model is... The expression is as follows: ； In the formula: The operating cost of the distribution network system for the next day, for The cost of power exchange between the distribution network and the main grid at any given time; , , They represent Time of the first Taiwan's controllable distributed power generation cost, Taiwan's energy storage operating costs, The response cost of interruptible loads; This represents the total number of hours for the next day. , and These represent the number of nodes for controllable distributed power sources, energy storage, and interruptible loads, respectively. in, , , , Based on respectively Power exchange between the distribution network and the main grid at all times , Time of the first Power generation capacity of Taiwan's controllable distributed power source , Time of the first Taiwan's energy storage charging and discharging power , Time of the first Power reduction for interruptible loads calculate; The constraints of the optimized control model include power balance constraints, power generation constraints of controllable distributed power sources, ramp-up constraints of controllable distributed power sources, SOC and charge / discharge constraints of energy storage, and interruptible load constraints.

5. The method according to claim 1, characterized in that, In S2, the objective function of the intraday rolling optimization control model is... The expression is as follows: ； In the formula: for Operating costs of the distribution network system during the time period For the current moment, , , intraday The cost of power exchange between the distribution network and the main grid at any time, the first Taiwan's controllable distributed power generation cost, intraday Time of the first Taiwan's energy storage operating costs; The constraints of the intraday rolling optimization control model include power balance constraints, power generation constraints of controllable distributed power sources, ramping constraints of controllable distributed power sources, and SOC and charge / discharge constraints of energy storage.

6. The method according to claim 1, characterized in that, In S3, the objective function of the real-time correction model is: ； In the formula: This refers to the amount of flexible resource adjustment that can be directly controlled at any given moment. For adjustable resources The adjustment amount; It represents the collection of all adjustable resources, including the exchange power between the distribution network and the main grid, the generation power of controllable distributed power sources, the power of energy storage, and the response power of interruptible loads; Available resources at the current moment The actual power measurement value; The optimized values ​​of adjustable resources are obtained by solving the intraday rolling optimization model or the day-ahead optimization model. Among them, the optimized values ​​of the exchange between the distribution network and the main grid, the optimized values ​​of the generation of controllable distributed power sources, and the optimized values ​​of energy storage are obtained by solving the intraday rolling optimization model, and the optimized values ​​of the power of interruptible loads are obtained. The solution was obtained from the current optimization model. The constraints of the real-time correction model include: requiring the power of the directly controlled flexible resources to be corrected. It satisfies power balance constraints, power generation constraints of controllable distributed power sources, ramping constraints of controllable distributed power sources, and SOC and charge / discharge constraints of energy storage.

7. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer programs; The processor is configured to invoke the computer program to perform the method as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed on an electronic device, causes the electronic device to perform the method as described in any one of claims 1 to 6.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is run on an electronic device, it causes the electronic device to perform the method as described in any one of claims 1 to 6.

Images