Distributed resource aggregator multi-time scale collaborative decision optimization method, system and device and storage medium

Abstract

The invention relates to the technical field of power system scheduling optimization, in particular to a distributed resource aggregator multi-time scale collaborative decision optimization method, system and device and a storage medium. The method comprises the following steps: collecting real-time operation data of distributed resources, establishing a photovoltaic, energy storage and load prediction model and quantifying prediction errors, establishing monomer adjustment capability constraints for different resource characteristics, and forming an adjustable capacity boundary of an aggregation resource pool through complementary characteristics among resources; establishing a mapping relation of three time scales of day-ahead, day-intra and real-time, reserving an adjustment margin in a day-ahead decision, reserving a deviation buffer in a day-intra decision, obtaining each market price prediction value, and generating a collaborative optimization declaration scheme and a scheduling instruction; quantifying risk exposure of market price and resource output, evaluating extreme scene loss by adopting a conditional value-at-risk method and dynamically adjusting a decision scheme; and obtaining update information in each decision period, and performing rolling optimization update by taking the execution state of the previous period as an initial state.

Description

Technical Field

[0001] This invention relates to the field of power system dispatch optimization technology, and in particular to a method, system, device and storage medium for multi-timescale collaborative decision optimization of distributed resource aggregators. Background Technology

[0002] In the power system, distributed resource aggregators are key players connecting distributed resources with the electricity market. These aggregators primarily integrate various types of distributed resources, such as photovoltaic power generation, energy storage systems, and controllable loads, to form virtual power plants, mainly for participating in electricity market transactions. Currently, the existing electricity market is typically divided into three time scales: day-ahead market, intraday market, and real-time balancing market. The day-ahead market involves declaration and clearing one day before the operating day, with a time scale of hours. The intraday market involves rolling declaration and clearing on the operating day, with a time scale of 15 minutes to 1 hour. The real-time balancing market settles deviations in real time during operation, with a time scale of minutes. These different time scales of the market have different prices and assessment rules.

[0003] In existing technologies, decision support typically handles decisions at three time scales—day-ahead market, intraday market, and real-time balancing market—independently. When submitting day-ahead market reports, optimization is mainly based on day-ahead price and resource forecasts, without considering the costs of intraday adjustments and the risks of real-time deviations. When adjusting intraday market reports, only the deviations of day-ahead plans are corrected, lacking provisions for real-time deviations. During real-time scheduling, the approach is often reactive, making it difficult to control deviation assessment costs.

[0004] Furthermore, photovoltaic power generation is highly susceptible to weather fluctuations, energy storage systems are constrained by state of charge and charge / discharge efficiency, and controllable loads are affected by uncertainties in user behavior. Existing resource aggregation methods often employ simple linear superposition, failing to accurately characterize the complementary characteristics and coupling constraints between different types of resources. This results in inaccurate assessments of the aggregated adjustable capacity, affecting the rationality of market applications.

[0005] At the same time, multiple risk factors such as electricity market price fluctuations, uncertainties in distributed resource output, and uncertainties in user response are intertwined, and existing technologies lack a sound risk quantification assessment system and dynamic risk avoidance methods. Summary of the Invention

[0006] In view of the problems existing in the prior art, the present invention is proposed.

[0007] Therefore, the problems to be solved by this invention are how to solve the problem of disconnect between decision-making at multiple time scales, how to solve the problem of insufficient accuracy in distributed resource aggregation, and how to solve the problems of imperfect risk quantification and weak risk resistance.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a multi-timescale collaborative decision optimization method for distributed resource aggregators, which includes: collecting real-time operating data of distributed resources; establishing photovoltaic output prediction, energy storage status prediction and load response prediction based on the real-time operating data; quantifying the prediction errors of various resources; establishing individual resource adjustment capability constraints according to the operating characteristics of distributed resources; and forming an adjustable capacity boundary of the aggregated resource pool through the complementary characteristics and coupling constraints between resources. Establish a time mapping relationship between the day-ahead decision period, the intraday decision period, and the real-time scheduling period. By reserving adjustment margin in the day-ahead decision and deviation buffer in the intraday decision, the decision-making connection between different time scales is realized. Obtain the predicted values ​​of the day-ahead market price, the intraday market price, and the real-time equilibrium price. Combine the adjustable capacity boundary and the time mapping relationship to generate the day-ahead market reporting plan, the intraday adjustment plan, and the real-time scheduling instruction. The risk exposure of market price deviation from forecast value and the deviation risk of resource output from planned value are quantified. The conditional value at risk method is used to assess the loss in extreme scenarios, and the day-ahead market reporting plan is adjusted according to the risk level. In each decision cycle, updated resource forecast information and market price information are obtained. The execution status of the previous cycle is used as the initial status of the current cycle, and the day-ahead market reporting plan, intraday adjustment plan and real-time scheduling instructions are recalculated and updated.

[0009] As a preferred embodiment of the multi-timescale collaborative decision-making optimization method for distributed resource aggregators described in this invention, wherein: the establishment of individual resource adjustment capability constraints based on the operating characteristics of distributed resources includes, wherein the distributed resources include photovoltaic resources, energy storage resources, and controllable load resources; For photovoltaic resources, the range of available photovoltaic power is calculated based on weather forecast data and photovoltaic rated capacity, and the uncertainty range of photovoltaic output is determined based on the statistical characteristics of historical forecast errors. For energy storage resources, the rechargeable and dischargeable capacity of energy storage in the future period is calculated based on the current state of charge, charging and discharging power limits and charging and discharging efficiency. For controllable load resources, the load reduction and increase amounts are calculated based on the load baseline power and user historical response data, and the uncertainty of user response is quantified.

[0010] As a preferred embodiment of the distributed resource aggregator multi-timescale collaborative decision optimization method of the present invention, wherein: the step of forming an adjustable capacity boundary of the aggregated resource pool through the complementary characteristics and coupling constraints between resources includes analyzing the temporal complementary characteristics of photovoltaic power output fluctuations and load fluctuations, and evaluating the fluctuation offsetting effect brought about by natural complementarity; Analyze the coordination relationship between energy storage charging and discharging regulation and photovoltaic power output fluctuations, and evaluate the ability of energy storage to mitigate the uncertainty of photovoltaic power output. By considering the available power range of photovoltaic resources, the bidirectional regulation capability of energy storage resources, and the regulation potential of controllable load resources, the upward and downward regulation capacity of the aggregated resource pool in each time period is calculated.

[0011] As a preferred embodiment of the distributed resource aggregator multi-timescale collaborative decision optimization method of the present invention, the step of establishing the time mapping relationship between the day-ahead decision period, the intraday decision period and the real-time scheduling period includes mapping the 24-hour period of the day-ahead decision to the 96 15-minute periods of the intraday decision, and reserving an adjustment margin between the day-ahead declared power and the adjustable capacity boundary of the aggregated resource pool. The 15-minute timeframe for intraday decision-making is mapped to a 5-minute timeframe for real-time scheduling, reserving a buffer space for deviations based on the intraday planned power. In the day-ahead decision-making process, intraday adjustment costs and real-time deviation costs are considered. Potential costs are assessed by the probability distribution of forecast errors, and cost compensation is carried out in the day-ahead declaration.

[0012] The beneficial effects of this preferred technical solution are as follows: First, this preferred technical solution maps the 24-hour period of day-ahead decision-making to 96 15-minute periods of intraday decision-making. By reserving an adjustment margin between the day-ahead declared power and the adjustable capacity boundary of the aggregated resource pool, it reserves operational space for power adjustment in the intraday market, avoiding the problem of day-ahead declarations being too aggressive and thus unable to be adjusted intraday. Next, the 15-minute periods of intraday decision-making are further mapped to 5-minute periods of real-time scheduling, reserving a deviation buffer space on the basis of intraday planned power, leaving room for adjustment to cope with real-time output fluctuations. By proactively considering intraday adjustment costs and real-time deviation costs in the day-ahead decision-making stage, assessing potential costs based on the probability distribution of prediction errors, and compensating for costs in the day-ahead declaration, it achieves organic connection between decisions at different time scales.

[0013] As a preferred embodiment of the multi-timescale collaborative decision-making optimization method for distributed resource aggregators described in this invention, the step of obtaining the predicted values ​​of the day-ahead market price, intraday market price, and real-time equilibrium price, and combining them with the adjustable capacity boundary and time mapping relationship to generate the day-ahead market application plan, intraday adjustment plan, and real-time scheduling instruction includes optimizing the application power within the day-ahead decision period based on the day-ahead market price prediction and the adjustable capacity boundary of the aggregated resource pool, thereby maximizing the net profit after deducting the expected intraday adjustment cost and the expected real-time deviation cost from the day-ahead market revenue. Based on intraday market price forecasts and day-ahead reporting results, power is optimized and adjusted during the intraday decision-making period to correct day-ahead plan deviations and reduce real-time deviation risks. Based on real-time balanced prices and actual resource operating status, power instructions for each distributed resource are generated during the real-time scheduling period to track intraday adjustment plans.

[0014] As a preferred embodiment of the distributed resource aggregator multi-timescale collaborative decision optimization method of the present invention, the quantification of the risk exposure of market price deviation from the predicted value and the deviation risk of resource output from the planned value includes: statistically analyzing historical market price prediction errors, calculating the probability distribution of price deviation from the predicted value, and assessing the risk of revenue loss caused by price fluctuations. Statistically analyze historical resource output prediction errors, calculate the probability distribution of output deviation from planned values, and assess the risk of performance losses caused by output deviations; Set a confidence level and use the conditional value at risk method to calculate the expected loss beyond the confidence level, thus quantifying the maximum risk exposure in extreme scenarios; Set a risk threshold, and when the calculated risk exposure exceeds the threshold, reduce the day-ahead market reporting power or increase the adjustment margin.

[0015] The beneficial effects of this preferred technical solution are as follows: First, by statistically analyzing historical market price prediction errors, this preferred technical solution calculates the probability distribution of price deviations from predicted values, accurately assessing the impact of price fluctuations on returns; simultaneously, by statistically analyzing historical resource output prediction errors, it calculates the probability distribution of output deviations from planned values, quantifying the assessment risks brought about by resource uncertainty; based on this, a confidence level is set, and the conditional value at risk method is used to calculate the expected loss exceeding the confidence level, thereby quantifying the maximum risk exposure in extreme scenarios; finally, by setting a risk threshold, when the calculated risk exposure exceeds the threshold, protective adjustments are triggered, reducing the day-ahead market reporting power or increasing the adjustment margin, proactively avoiding high-risk scenarios.

[0016] As a preferred embodiment of the distributed resource aggregator multi-timescale collaborative decision-making optimization method of the present invention, the step of acquiring updated resource forecast information and market price information in each decision cycle, using the execution state of the previous cycle as the initial state of the current cycle, and recalculating and updating the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions includes: During the day-ahead decision-making cycle, obtain the photovoltaic output forecast, load forecast, market price forecast, and current state of charge of energy storage for the next 24 hours, and recalculate the day-ahead market application plan; During the intraday decision-making cycle, obtain updated forecast information for the next 4 hours, and recalculate the intraday adjustment plan based on the previous day's declaration results; During the real-time scheduling cycle, the actual operating status of resources and real-time market prices at the current moment are obtained, and the real-time scheduling instructions are recalculated with the intraday plan as the tracking target. Each time a decision is recalculated, the values ​​of the unexecuted portions of the decision variables from the previous period are used as the initial values ​​for the corresponding time period in the current period to maintain decision continuity.

[0017] Secondly, embodiments of the present invention provide a multi-timescale collaborative decision-making optimization system for distributed resource aggregators, which includes a data acquisition module that collects real-time operating data of distributed resources, establishes photovoltaic output prediction, energy storage status prediction and load response prediction based on the real-time operating data, quantifies the prediction error of various resources, establishes individual resource adjustment capability constraints for the operating characteristics of distributed resources, and forms an adjustable capacity boundary of the aggregated resource pool through the complementary characteristics and coupling constraints between resources. The time-series coordination module establishes a time mapping relationship between the day-ahead decision-making period, the intraday decision-making period, and the real-time scheduling period. By reserving adjustment margins in day-ahead decisions and deviation buffers in intraday decisions, it achieves decision-making connection between different time scales, obtains the predicted values ​​of day-ahead market prices, intraday market prices, and real-time equilibrium prices, and generates day-ahead market reporting plans, intraday adjustment plans, and real-time scheduling instructions by combining the adjustable capacity boundary and the time mapping relationship. The risk management module quantifies the risk exposure of market price deviations from forecast values ​​and the deviation risk of resource output from planned values, uses the conditional value at risk approach to assess losses in extreme scenarios, and adjusts the day-ahead market reporting plan according to the risk level. The rolling correction module acquires updated resource forecast information and market price information in each decision cycle, uses the execution status of the previous cycle as the initial status of the current cycle, and recalculates and updates the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions.

[0018] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the distributed resource aggregator multi-timescale collaborative decision optimization method as described in the first aspect of the present invention.

[0019] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the distributed resource aggregator multi-timescale collaborative decision optimization method as described in the first aspect of the present invention.

[0020] The beneficial effects of this invention are as follows: This invention collects real-time operational data of distributed resources and establishes prediction models for photovoltaics, energy storage, and load. It establishes individual resource regulation capacity constraints based on the operational characteristics of different resources, and forms an adjustable capacity boundary for the aggregated resource pool through complementary characteristics and coupling constraints between resources, accurately assessing the overall regulation capacity of the resource pool. It establishes a mapping relationship between three time scales: day-ahead, intraday, and real-time. By reserving regulation margins in day-ahead decisions and deviation buffers in intraday decisions, it achieves seamless decision-making across different time scales. In the day-ahead decision-making stage, it proactively considers intraday adjustment costs and real-time deviation costs, generating collaboratively optimized application schemes and scheduling instructions. Simultaneously, by statistically analyzing historical market price prediction errors and resource output prediction errors, it calculates the probability distribution of deviations from predicted values, quantifies the maximum risk exposure under extreme scenarios using the conditional value at risk method, and dynamically adjusts the day-ahead declared power or increases the regulation margin based on the risk level. Through a rolling correction framework, it obtains updated prediction information in each decision cycle, using the execution state of the previous cycle as the initial state of the current cycle for iterative updates, achieving continuous dynamic optimization. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart for a multi-timescale collaborative decision-making optimization method for distributed resource aggregators; Figure 2 A computer device diagram for a multi-timescale collaborative decision-making optimization method for distributed resource aggregators; Figure 3 Another flowchart for a multi-timescale collaborative decision-making optimization method for distributed resource aggregators. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0026] Example 1 Reference Figure 1 - Figure 2 As a first embodiment of the present invention, this embodiment provides a multi-timescale collaborative decision-making optimization method for distributed resource aggregators, including, S100: Collects real-time operating data of distributed resources, establishes photovoltaic output prediction, energy storage status prediction and load response prediction based on real-time operating data, quantifies the prediction error of various resources, establishes individual resource adjustment capacity constraints according to the operating characteristics of distributed resources, and forms the adjustable capacity boundary of the aggregated resource pool through the complementary characteristics and coupling constraints between resources.

[0027] S200: Establish the time mapping relationship between the day-ahead decision period, the intraday decision period, and the real-time scheduling period. By reserving adjustment margin in the day-ahead decision and deviation buffer in the intraday decision, the decision-making connection between different time scales is realized. The predicted values ​​of the day-ahead market price, the intraday market price, and the real-time equilibrium price are obtained. Combined with the adjustable capacity boundary and the time mapping relationship, the day-ahead market declaration plan, the intraday adjustment plan, and the real-time scheduling instruction are generated.

[0028] S300: Quantify the risk exposure of market price deviation from forecast value and the deviation risk of resource output from planned value, use the conditional value at risk method to assess losses in extreme scenarios, and adjust the day-ahead market reporting scheme according to the risk level.

[0029] S400: In each decision cycle, it acquires updated resource forecast information and market price information, uses the execution status of the previous cycle as the initial status of the current cycle, and recalculates and updates the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions.

[0030] It should be noted that day-ahead market submissions require a 24-hour advance forecast of resource output and market prices. However, photovoltaic power generation is subject to significant forecasting errors due to weather changes, and controllable loads are affected by uncertainties in user behavior, leading to fluctuations in response rates. The state of charge of energy storage systems is path-dependent due to prior decisions. Price fluctuations in the intraday and real-time markets further exacerbate the difficulty of decision-making. If day-ahead submissions are too aggressive, intraday adjustments will be costly; if they are too conservative, market opportunities will be lost.

[0031] Therefore, through steps S100 to S400, real-time operational data of distributed resources are collected and a predictive model is established. Adjustment capacity constraints are established for different characteristics of photovoltaic, energy storage, and controllable loads. The adjustable capacity boundary of the aggregated resource pool is evaluated through the complementary characteristics between resources. Then, a mapping relationship is established between three time scales: day-ahead, intraday, and real-time. Adjustment margin and deviation buffer are reserved in day-ahead decision-making. Combined with market price forecasts, collaboratively optimized application schemes and scheduling instructions are generated. At the same time, price risk and output deviation risk are quantified. The conditional value at risk method is used to assess losses in extreme scenarios and dynamically adjust decision schemes to reduce risk exposure. Finally, the predictive information and market status are continuously updated through a rolling correction framework to iteratively optimize decision schemes at each time scale.

[0032] Example 2 Reference Figure 1 - Figure 3 This is the second embodiment of the present invention.

[0033] In this embodiment, step S100 involves collecting real-time operational data of distributed resources. Based on this data, photovoltaic output prediction, energy storage status prediction, and load response prediction are established. The prediction errors for each resource are quantified. Individual resource adjustment capacity constraints are established based on the operational characteristics of distributed resources. Through the complementary characteristics and coupling constraints between resources, an adjustable capacity boundary for the aggregated resource pool is formed. This includes the following steps A1-A3: A1: Collect real-time operational data of distributed resources, and based on the real-time operational data, establish photovoltaic output prediction, energy storage status prediction, and load response prediction respectively, and quantify the prediction error of various resources.

[0034] Specifically, the system first collects operational data from various resources in real time through local controllers, smart meters, and communication gateways for distributed resources. For photovoltaic (PV) resources, the collected data includes parameters such as PV power generation, irradiance, ambient temperature, and module temperature, with a sampling frequency of once per minute to capture the rapid fluctuations in PV output. For energy storage resources, the collected data includes parameters such as state of charge (SOC), charge / discharge power, battery temperature, and charge / discharge efficiency, also with a sampling frequency of once per minute to monitor the energy storage's operational status in real time.

[0035] The data integrity metric is defined as shown in formula (1): (1) In the formula, The data integrity coefficient. This represents the actual number of data packets received. The number of data packets to be received. If the data is found to be incomplete, the data completion program will be initiated.

[0036] The data integrity coefficient is equal to the ratio of the number of data packets actually received to the number of data packets that should have been received. When the data integrity coefficient is less than 0.95, the data is considered incomplete, and a data completion procedure is initiated.

[0037] After completing data collection and quality verification, the photovoltaic resources are assessed for their status, and the current available photovoltaic power is calculated according to formula (2): (2) In the formula, Let be the available photovoltaic power at time t. For photovoltaic rated capacity, This represents the current operating efficiency. Operating efficiency is determined by the ratio of measured power to power under standard test conditions.

[0038] The standard irradiance is set at 1000 watts per square meter. The temperature correction factor is calculated using a formula, which is equal to 1 minus the temperature coefficient multiplied by the difference between the actual temperature and the standard temperature of 25 degrees Celsius. The typical value of the temperature coefficient is 0.004 per degree Celsius.

[0039] The energy storage resources are assessed for their condition and available capacity is calculated. The maximum rechargeable power is calculated according to formula (3): (3) The maximum rechargeable power is equal to the minimum value of the difference between the rated power of the energy storage and the upper limit of the state of charge minus the current state of charge, multiplied by the energy storage capacity, divided by the time interval, and then divided by the charging efficiency.

[0040] The upper limit of the state of charge is usually set to 0.9 to protect the battery from overcharging.

[0041] Calculate the maximum discharge power according to formula (4): (4) In the formula, and These represent the maximum rechargeable power and the maximum dischargeable power at time t, respectively. This is the rated power of the energy storage. The current state of charge, and These are the upper and lower limits of the state of charge (usually taken as 0.9 and 0.1). For energy storage capacity, The time interval is 0.25 hours. and These represent the charge / discharge efficiency (typical value 0.95).

[0042] The maximum dischargeable power is equal to the minimum value of the result obtained by multiplying the difference between the rated power of the energy storage and the current state of charge minus the lower limit of the state of charge, multiplying by the energy storage capacity, dividing by the time interval, and then multiplying by the discharge efficiency.

[0043] The state of charge (SCC) limit is typically set to 0.1 to prevent over-discharge from affecting battery life. The time interval is set to 0.25 hours (15 minutes), corresponding to the time granularity of the intraday market. The typical charge / discharge efficiency is set to 0.95.

[0044] The response capability of the controllable load is assessed, and the maximum load that can be reduced is calculated according to formula (5): (5) In the formula, The maximum load that can be reduced at time t. For controllable load quantity, This is the reference power for the i-th load. The adjustable ratio of load i (determined according to load type, 0.3 for lighting load and 0.5 for air conditioning load). The user's responsive status is 1 (1 indicates responsive, 0 indicates unresponsive, determined by user response history and statistics of the current time period).

[0045] The maximum load reduction is equal to the sum of all controllable loads. The reduction amount for each load is equal to the load's base power multiplied by the adjustable ratio and then multiplied by the user's responsiveness.

[0046] The adjustable ratio is determined based on the load type. For lighting loads, a ratio of 0.3 indicates a 30% power reduction; for air conditioning loads, 0.5 indicates a 50% power reduction; and for industrial production loads, the ratio is between 0.2 and 0.4, depending on process requirements. The user's responsiveness status is a binary variable: 1 indicates the user can currently respond to the power reduction request, and 0 indicates the user cannot respond. This status is determined through historical user response data and statistical patterns of the current time period. For example, users are more willing to respond during peak electricity consumption periods and less willing to respond during off-peak periods.

[0047] After completing the current status assessment, resource output for future periods is predicted based on historical and real-time data. For photovoltaic output prediction, a combined prediction method combining numerical weather prediction (NWP) data and historical statistics is adopted. The predicted photovoltaic power value for future times is calculated according to formula (6): (6) In the formula, This represents the predicted photovoltaic power at time t+k. For predicted solar irradiance (obtained via NWP). This represents the current measured irradiance. This is the temperature correction factor, calculated as follows: ,in To predict temperature.

[0048] The predicted photovoltaic power output equals the rated photovoltaic capacity multiplied by the ratio of predicted irradiance to the current measured irradiance, then multiplied by a temperature correction factor and the current operating efficiency. Predicted irradiance is obtained through numerical weather prediction, providing hourly irradiance forecasts for the next 72 hours. The temperature correction factor is calculated as 1 minus the temperature coefficient multiplied by the difference between the predicted temperature and 25 degrees Celsius.

[0049] The uncertainty of photovoltaic forecasting is quantified. The variance of photovoltaic forecasting is calculated according to formula (7): (7) In the formula, Let N be the variance of photovoltaic forecasts, and N be the number of historical samples (taking similar weather data from the most recent 30 days). and These are the predicted and actual values ​​for the j-th sample, respectively.

[0050] The variance of photovoltaic predictions is equal to the sum of squared prediction errors over historical samples divided by the number of samples minus 1.

[0051] The historical sample size N consists of similar weather data from the most recent 30 days. A weather classification algorithm categorizes the historical data into types such as sunny, cloudy, overcast, and rainy. Historical data of the same category are selected for statistical analysis when predicting the weather for the current day. The prediction error is the difference between the predicted and actual values.

[0052] To predict the future state of energy storage, the predicted state of charge at future times is calculated according to formula (8): (8) In the formula, The predicted state of charge at time t+k and Time periods The planned charge and discharge power, these values ​​are derived from the results of the third step of optimization decision-making.

[0053] The state of charge at a future time is equal to the current state of charge plus the sum of the charging power multiplied by the charging efficiency during each period from the current time to the predicted time, minus the cumulative sum of the discharge power divided by the discharge efficiency, and then divided by the energy storage capacity.

[0054] The planned charging and discharging power for each time period is derived from the results of the day-ahead market optimization decision in the fourth step. This recursive calculation method can predict the state-of-charge trajectory of energy storage after executing the predetermined plan, ensuring that the energy storage will not be depleted or overcharged throughout the entire operating cycle.

[0055] To predict user response behavior under controllable load, a user response probability model is established based on formula (9): (9) In the formula, Let be the probability of user i's response at time t. The incentive price is provided, and h(t) is a time-period characteristic variable (1 for weekdays and 0 for weekends). The model parameters are obtained through logistic regression fitting of historical response data, with typical values ​​being... .

[0056] The probability of user i responding at time t is calculated using a logistic regression model. This probability equals 1 divided by 1 plus the negative power of the natural exponent, where the exponent term is the sum of the parameters. Model parameters include the intercept term, the coefficient of the incentive price, and the coefficients of the time-period characteristic variable. The incentive price is the load reduction compensation price offered to users; the higher the price, the greater the probability of user response. The time-period characteristic variable takes the value 1 on weekdays and 0 on weekends, because users are more sensitive to incentives on weekdays. The model parameters are obtained by logistic regression fitting of historical response data, with typical parameter values ​​of -2 for the intercept, 5 for the incentive price coefficient, and 0.8 for the time-period characteristic coefficient.

[0057] A2: Based on the operational characteristics of distributed resources, constraints on the adjustment capabilities of individual resources are established, including photovoltaic resources, energy storage resources, and controllable load resources; For photovoltaic resources, the range of available photovoltaic power is calculated based on weather forecast data and photovoltaic rated capacity, and the uncertainty range of photovoltaic output is determined based on the statistical characteristics of historical forecast errors. For energy storage resources, the rechargeable and dischargeable capacity of energy storage in the future period is calculated based on the current state of charge, charging and discharging power limits and charging and discharging efficiency. For controllable load resources, the load reduction and increase amounts are calculated based on the load baseline power and user historical response data, and the uncertainty of user response is quantified.

[0058] Specifically, for photovoltaic resources, their regulation capability is mainly reflected in the controllable reduction capability of output. Photovoltaic resources themselves do not have the ability to increase output, but the output can be reduced by controlling the inverter. Based on the available photovoltaic power calculated by formula (2), the upper limit of photovoltaic output in each time period is determined. At the same time, considering the uncertainty of photovoltaic forecasting, the uncertainty range of photovoltaic output is determined according to the forecast variance calculated by formula (7). This uncertainty range is defined as ±1.96 times the standard deviation of the forecast value, corresponding to a 95% confidence level.

[0059] For energy storage resources, their regulation capability is reflected in two dimensions: bidirectional power regulation and energy storage. Based on formulas (3) and (4), the maximum rechargeable power and maximum dischargeable power of the energy storage in each time period are calculated. These two values ​​are not only limited by the rated power of the energy storage but also constrained by the current state of charge (SOC). When the SOC approaches the upper limit, the rechargeable power decreases until it reaches zero; when the SOC approaches the lower limit, the dischargeable power decreases until it reaches zero. Furthermore, energy storage is also subject to ramp rate constraints, meaning that the power change between adjacent time periods cannot exceed the maximum ramp rate, which is typically 20% of the rated power of the energy storage. For example, for a 1 MW energy storage system, the power change in a single time period cannot exceed 0.2 MW.

[0060] For controllable load resources, their adjustment capability is mainly reflected in the load reduction capability and the time-period transfer capability of part of the load. According to formula (5), the maximum load that can be reduced in each time period is calculated. However, the actual adjustment capability also needs to consider user comfort constraints and response frequency limits. User comfort constraints are reflected in the fact that the duration of a single reduction cannot be too long. For example, the duration of a single reduction of air conditioning load should not exceed 2 hours, otherwise the indoor temperature will exceed the comfort range. Response frequency limits are reflected in the fact that the number of responses in a single day should not be too many. For example, the number of responses for commercial users should not exceed 3 times a day to avoid frequent disturbances that affect user satisfaction.

[0061] A3: By leveraging the complementary characteristics and coupling constraints among resources, an adjustable capacity boundary for the aggregated resource pool is formed, including analyzing the temporal complementary characteristics of photovoltaic power output fluctuations and load fluctuations, and evaluating the fluctuation offsetting effect brought about by natural complementarity. Analyze the coordination relationship between energy storage charging and discharging regulation and photovoltaic power output fluctuations, and evaluate the ability of energy storage to mitigate the uncertainty of photovoltaic power output. By considering the available power range of photovoltaic resources, the bidirectional regulation capability of energy storage resources, and the regulation potential of controllable load resources, the upward and downward regulation capacity of the aggregated resource pool in each time period is calculated.

[0062] Specifically, after completing the individual resource modeling, an aggregated resource pool model is constructed based on the state assessment in the first step and the prediction results in the second step. Through reasonable configuration and coordinated scheduling, the overall performance of the aggregated resource pool is superior to the simple superposition of individual resources. First, the temporal complementary characteristics of photovoltaic power output fluctuations and load fluctuations are analyzed. During peak photovoltaic power output periods in the daytime, which are also often peak periods for commercial loads, photovoltaic power can directly supply the load, reducing the need to purchase electricity from the grid. During the evening hours when photovoltaic power output decreases, the load also decreases accordingly; this natural complementary characteristic can offset some of the fluctuations.

[0063] Energy storage can smooth out short-term fluctuations in photovoltaic output. When the actual photovoltaic output is higher than the predicted value, the energy storage can charge to absorb the excess power; when the actual photovoltaic output is lower than the predicted value, the energy storage can discharge to supplement the insufficient power. According to the energy storage state prediction model of formula (8), the charging and discharging sequence of energy storage can be planned in advance to ensure that the energy storage has sufficient available capacity when adjustment is needed. For example, when submitting the application for the day before, if it is predicted that the photovoltaic output will fluctuate greatly the next morning, the energy storage SOC will be adjusted to an intermediate level, such as 50%, the night before so that it can both charge and discharge the next day, maximizing the bidirectional adjustment capability of the energy storage.

[0064] Specifically, the total available capacity of the aggregated resource pool in each time period is calculated according to formula (10): (10) In the formula, To aggregate and increase capacity, The power output is used as the photovoltaic benchmark (80% of the actual output is taken as a conservative estimate).

[0065] The total increased capacity equals the maximum rechargeable power of energy storage plus the difference between the photovoltaic benchmark output and the actual declared output of photovoltaic power, plus the maximum load that can be reduced.

[0066] Specifically, the photovoltaic baseline output is conservatively estimated at 80% of the actual output forecast, with a 20% margin reserved for upward adjustment. This conservative strategy can address situations where photovoltaic forecasts are overestimated.

[0067] The total available downsizing capacity is calculated according to formula (11): (11) In the formula, To reduce capacity for aggregation, To increase the load (some controllable loads have time-shifting capabilities).

[0068] The total reduced capacity equals the maximum discharge power of energy storage plus the difference between the actual declared output of photovoltaic power and the rated capacity of photovoltaic power, plus the load that can be increased.

[0069] Here, "expandable load" refers to the ability of a portion of controllable load to shift during specific time periods, enabling it to respond when additional load is needed. These two formulas allow for the quantification of the bidirectional adjustment capability boundary of the aggregated resource pool across different time periods.

[0070] Considering the uncertainty of resource response, a confidence coefficient is introduced to correct the aggregation capacity. The adjusted capacity considering the confidence coefficient is calculated according to formula (12): (12) In the formula, To account for the increased capacity of confidence level, The confidence level coefficient (1.96 for 95% confidence level) is used. The aggregated resource variation coefficient is used. The corrected upward adjustment capacity is equal to the original upward adjustment capacity minus the confidence level coefficient multiplied by the product of the original upward adjustment capacity and the aggregated resource variation coefficient.

[0071] The confidence level coefficient is taken as 1.96 at a 95% confidence level, and the coefficient of variation of aggregated resources is calculated according to formula (13): (13) In the formula, For the load response variance, through The coefficient of variation is calculated to be the square root of the sum of the photovoltaic forecast variance and the load response variance, divided by the original increased capacity. The load response variance is obtained through statistical analysis of historical user response data and reflects the uncertainty of user response.

[0072] For example, if historical data shows an average user response rate of 70% and a standard deviation of 15%, then the load response variance is 0.15 squared. Through confidence level correction, the resulting aggregate capacity is more conservative and reliable, and can meet actual adjustment needs with a higher probability.

[0073] The continuous adjustment capability of the aggregated resource pool also needs to be assessed, and the longest sustainable adjustment time is calculated according to formula (14): (14) In the formula, For sustainable time adjustment, This represents the total transferable electricity (determined based on user comfort constraints, typically 5% to 10% of the daily load).

[0074] The duration of operation is equal to the sum of the available energy storage capacity and the total load transferable amount, divided by the regulation power. The available energy storage capacity is calculated based on the current SOC and upper and lower limits, while the total load transferable amount is determined based on user comfort constraints, typically ranging from 5% to 10% of the daily load.

[0075] For example, if the total daily load is 100 MWh, the transferable electricity is approximately 5 to 10 MWh. If continuous regulation at 1 MWh is required, and the available energy storage capacity is 0.5 MWh, the transferable load is 5 MWh, resulting in a maximum sustainable regulation time of 5.5 hours. This metric is crucial for assessing the ability of aggregated resources to cope with persistent deviations.

[0076] In this embodiment, step S200 establishes a time mapping relationship between the day-ahead decision-making period, the intraday decision-making period, and the real-time scheduling period. By reserving adjustment margins in day-ahead decisions and deviation buffers in intraday decisions, decision-making connections between different time scales are achieved. The predicted values ​​of day-ahead market prices, intraday market prices, and real-time equilibrium prices are obtained. Combined with the adjustable capacity boundary and the time mapping relationship, a day-ahead market reporting plan, an intraday adjustment plan, and a real-time scheduling instruction are generated, including the following steps B1-B2: B1: Establish the time mapping relationship between the day-ahead decision period, the intraday decision period, and the real-time scheduling period, including mapping the 24-hour period of the day-ahead decision to the 96 15-minute periods of the intraday decision, and reserving an adjustment margin between the day-ahead declared power and the adjustable capacity boundary of the aggregated resource pool; The 15-minute timeframe for intraday decision-making is mapped to a 5-minute timeframe for real-time scheduling, reserving a buffer space for deviations based on the intraday planned power. In the day-ahead decision-making process, intraday adjustment costs and real-time deviation costs are considered. Potential costs are assessed by the probability distribution of forecast errors, and cost compensation is carried out in the day-ahead declaration.

[0077] Specifically, the electricity market is typically divided into three time scales: the day-ahead market, which involves declaration and clearing one day before the operating day, with a time dimension of 1 hour and a total of 24 time periods; the intraday market, which operates on a rolling basis on the operating day, with a time dimension of 15 minutes and a total of 96 time periods; and the real-time balancing market, which settles in real time during operation, with a time granularity of 5 minutes and a total of 288 time periods. First, the 24-hour time periods for day-ahead decisions are mapped to the 96 15-minute time periods for intraday decisions. Specifically, the mapping relationship is as follows: the h-th hour of the day-ahead market corresponds to the 4h-3 to 4h 15-minute time periods of the intraday market. For example, the 10th hour of the day-ahead market (9:00-10:00) corresponds to the 37th to 40th time periods of the intraday market (9:00-9:15, 9:15-9:30, 9:30-9:45, 9:45-10:00).

[0078] Establish aggregation capacity constraints according to formula (17): (17) The declared power must meet the requirement that the revised downward capacity is less than or equal to the declared power, which is less than or equal to the revised upward capacity. However, to allow for intraday adjustments, the capacity boundary will not be fully utilized in actual declarations; instead, a certain margin will be reserved within the boundary. The size of the margin will be dynamically adjusted based on the forecast uncertainty. If the photovoltaic forecast variance is large or the user response uncertainty is high, a larger margin will be reserved. The economic cost of reserving the margin is calculated using formula (16): (16) In the formula, The margin coefficient, ranging from 0.1 to 0.2, represents the opportunity cost weight of reserved capacity. The margin cost is equal to the sum of the margin coefficient multiplied by the product of the difference between the day-ahead market price and the declared power and the revised upward capacity over the entire day. The margin coefficient, ranging from 0.1 to 0.2, represents the opportunity cost weight of reserved capacity.

[0079] The mapping relationship is as follows: the q-th 15-minute period in the intraday market corresponds to the 3q-2 to 3q-5-minute periods in real-time scheduling. For example, the 40th period in the intraday market (9:45-10:00) corresponds to the 118th to 120th periods in real-time scheduling (9:45-9:50, 9:50-9:55, 9:55-10:00). Based on the planned intraday power output, a deviation buffer is also required to cope with output fluctuations in the real-time phase. The size of the deviation buffer is determined based on the statistical characteristics of the real-time forecast error.

[0080] The potential costs are assessed by evaluating the probability distribution of the prediction error, and cost compensation is made in the day-ahead filing. The objective function for day-ahead market filing optimization is established based on formula (15): (15) In the formula, This represents the total market return to date. The market price forecast for time period t is given before the current day. This refers to the power output reported by the market recently. To reserve the opportunity cost for adjustment margin.

[0081] Total revenue equals the sum of the day-ahead market price multiplied by the order power for all time periods, minus the opportunity cost of the reserved adjustment margin. However, this objective function needs to be further expanded to consider the expected intraday adjustment cost and real-time deviation cost.

[0082] B2: Obtain the predicted values ​​of the day-ahead market price, intraday market price, and real-time equilibrium price, and combine them with the adjustable capacity boundary and time mapping relationship to generate the day-ahead market application plan, intraday adjustment plan, and real-time scheduling instructions. This includes optimizing the application power during the day-ahead decision period based on the day-ahead market price prediction and the adjustable capacity boundary of the aggregated resource pool, and maximizing the net profit after deducting the expected intraday adjustment cost and expected real-time deviation cost from the day-ahead market revenue. Based on intraday market price forecasts and day-ahead reporting results, power is optimized and adjusted during the intraday decision-making period to correct day-ahead plan deviations and reduce real-time deviation risks. Based on real-time balanced prices and actual resource operating status, power instructions for each distributed resource are generated during the real-time scheduling period to track intraday adjustment plans.

[0083] Specifically, the first step is to obtain market price forecasts for various time scales. Day-ahead market price forecasts are based on historical price data, load forecasts, and renewable energy output forecasts, using time series models. Price forecasts not only need to provide point forecast values ​​but also quantify the uncertainty of the forecasts and provide forecast ranges. Intraday market prices typically fluctuate to some extent based on day-ahead prices, with a typical fluctuation range of plus or minus 20%.

[0084] During the day-ahead decision period, optimize the declared power to maximize net income. Establish the objective function according to formula (15), with constraints including capacity constraints of formula (17), energy storage SOC constraints of formula (18), and energy storage intraday balance constraints of formula (19).

[0085] Formula (18) indicates that the SOC after the plan is executed must meet the following requirements: (18) In the formula, The SOC after the execution of the day-ahead plan is calculated using formula (8). The minimum SOC is less than or equal to the SOC after the execution of the day-ahead plan, which is less than or equal to the maximum SOC. The SOC after the execution of the day-ahead plan is calculated recursively using formula (8).

[0086] The intraday equilibrium constraint of formula (19) is expressed as follows: (19) This constraint ensures that energy storage maintains a balance throughout the day and prevents energy depletion.

[0087] The net charge / discharge of energy storage within a 24-hour cycle is zero, meaning the sum of the charging power multiplied by the charging efficiency and the discharging power divided by the discharging efficiency for all time periods equals zero. This constraint ensures that energy storage achieves energy balance within the day, avoiding energy depletion or overcharging. The optimization problem is solved using a mixed-integer linear programming (MILP) method with commercial solvers such as Gurobi or CPLEX. The resulting day-ahead reporting scheme includes the reporting power values ​​for 24 hours and is submitted to the electricity market.

[0088] After the recent market clearing, the intraday market enters a rolling optimization phase. The intraday market opens once per hour and closes 15 minutes in advance. During each intraday decision-making cycle, rolling optimization is performed based on the latest resource forecast information and market price information. The optimization objective for intraday adjustments is established according to formula (20): (20) In the formula, Adjusting earnings for intraday trading For the current moment, This is the intraday market price. To adjust power within the day, This represents the actual executable power (updated based on the latest resource status). The deviation penalty price is 1.5 times the previous day's price.

[0089] Intraday adjusted profit equals the sum of intraday market prices multiplied by intraday adjusted power for all periods after the current moment, minus the difference between actual executable power and the previous day plus intraday declared power multiplied by the deviation penalty price.

[0090] The constraints include the intraday power adjustment limit in formula (21) and the energy storage SOC update constraint in formula (22). Formula (21) means: (twenty one) In the formula, and The aggregate capacity, updated based on the latest data, is obtained by re-executing steps one through three.

[0091] The power declared recently plus the power adjusted within the day must meet the latest adjustable capacity range.

[0092] Formula (22) means: (twenty two) In the formula, To take into account the intraday adjusted SOC, Adjusting power for energy storage in the intraday market.

[0093] The SOC adjusted intraday is calculated recursively, similar to formula (8) but with the addition of an intraday power adjustment term. The intraday optimization problem is solved to obtain the adjustment scheme, and the updated total declared power is calculated according to formula (23): (twenty three) The total reporting power The target power for real-time scheduling is passed to step six. The total declared power equals the power declared the day before plus the power adjusted within the day.

[0094] During the real-time operation phase, a rolling optimization is performed every 5 minutes to generate refined control instructions for each distributed resource. The optimization objective for real-time scheduling is established based on formula (24): (twenty four) In the formula, To reduce real-time scheduling costs, For the total number of resources, For resources i Actual output To contribute to the plan, For resources i The adjustment cost coefficient (0.05 yuan / kWh for energy storage, 0.1 yuan / kWh for load) is calculated. The total deviation penalty coefficient is set at 0.5 yuan / kWh.

[0095] The real-time scheduling cost equals the sum of the absolute values ​​of the differences between the actual output and planned output of all resources multiplied by the adjustment cost coefficient of that resource, plus the absolute value of the total deviation multiplied by the total deviation penalty coefficient. Constraints include the resource output constraint in formula (25), the power balance constraint in formula (26), and the energy storage ramping constraint in formula (27). Formula (25) states: (25) In the formula, and The upper and lower limits of the output of resource i are obtained from the state assessment in the first step.

[0096] The actual output of each resource must be within the upper and lower limits of that resource's output.

[0097] Formula (26) means: (26) The sum of the actual output of all resources must equal the total declared power of the day before and the day after, to ensure that the aggregated resource pool tracks the declaration plan as a whole. Formula (27) means: (27) In the formula, The maximum ramp rate for energy storage is 20% of the rated power.

[0098] The power variation of the energy storage in adjacent time periods shall not exceed the maximum ramp rate, which is taken as 20% of the rated power of the energy storage. The control commands for each resource are obtained by solving the real-time optimization problem, and the commands are sent to the local controller of each resource for execution through the communication network.

[0099] In this embodiment, step S300 quantifies the risk exposure of market price deviation from the forecast value and the deviation risk of resource output from the planned value, uses the conditional value at risk method to assess losses in extreme scenarios, and adjusts the day-ahead market reporting plan according to the risk level, including the following steps C1-C2: C1: Quantify the risk exposure of market price deviation from forecast value and the deviation risk of resource output from planned value, including statistical analysis of historical market price forecast errors, calculation of the probability distribution of price deviation from forecast value, and assessment of the risk of revenue loss caused by price fluctuations; Statistically analyze historical resource output prediction errors, calculate the probability distribution of output deviation from planned values, and assess the risk of performance losses caused by output deviations; Set a confidence level and use the conditional value at risk method to calculate the expected loss beyond the confidence level, thus quantifying the maximum risk exposure in extreme scenarios; Set a risk threshold, and when the calculated risk exposure exceeds the threshold, reduce the day-ahead market reporting power or increase the adjustment margin.

[0100] Specifically, the first step is to assess price risk. Price risk stems from the volatility and unpredictability of market prices. Price risk exposure is calculated using formula (28): (28) In the formula, For price risk exposure, denoted as the standard deviation of prices for period t, calculated using historical price data.

[0101] Price risk exposure equals the sum of the day-ahead reported power over all periods multiplied by the price standard deviation for that period. The price standard deviation is obtained through statistical analysis of historical price data and reflects the degree of price volatility during that period. For example, if the historical price standard deviation for a certain period is 0.1 yuan per kilowatt-hour, and the day-ahead reported power is 1000 kilowatts, then the price risk exposure for that period is 100 yuan. Price risk exposure quantifies the impact of price volatility on returns; a larger value indicates a more sensitive decision-making approach to price fluctuations.

[0102] Further assess the output deviation risk. Output deviation risk arises from the difference between actual and planned output, primarily caused by photovoltaic forecasting errors and load response uncertainties. The Conditional Value at Risk (CVaR) method is used for assessment. CVaR is defined according to formula (29): (29) In the formula, Confidence level The conditional value at risk (value) is given, where Loss is the loss caused by the deviation. Value at risk.

[0103] At confidence level α, the conditional value at risk equals the conditional expectation of the portion of the loss exceeding the value at risk (VaR).

[0104] The confidence level α is typically set to 0.95, representing a focus on the worst-case 5% scenario. To calculate CVaR, 1000 random scenarios are generated using Monte Carlo simulation. The bias loss for each scenario is calculated according to formula (30): (30) In the formula, For the deviation loss of scene s, The actual executable power in scenario s is generated by superimposing prediction errors on photovoltaics and load.

[0105] The deviation loss for scenario s is equal to the sum of the sum of the real-time deviation penalty prices for all time periods multiplied by the absolute values ​​of the differences between the actual executable power and the day-ahead reported power for that scenario. The actual executable power for a scenario is generated by superimposing random prediction errors on photovoltaic and load data. The prediction errors follow a normal distribution with a mean of zero and a standard deviation based on historical statistics. After generating 1000 scenarios, the deviation losses are sorted from smallest to largest. The loss value of the 950th scenario is taken as VaR, and the average of the losses from the 951st to the 1000th scenarios is taken as CVaR. The CVaR value quantifies the potential loss under extremely unfavorable conditions.

[0106] The user response risk is also assessed. User response risk stems from the uncertainty of response from users under controllable load, meaning the actual user response volume may be lower than expected. The user response risk is calculated using formula (31): (31) In the formula, Potential losses caused by user non-response.

[0107] The potential loss caused by a user's non-response is equal to the sum of 1 minus the user's response probability multiplied by the user's planned power reduction multiplied by the real-time deviation penalty price for all time periods and all users. The user response probability is given by the logistic regression model of formula (9). For example, if a user's response probability is 0.7, the planned power reduction is 100 kW, the real-time deviation penalty price is 0.6 yuan per kilowatt-hour, and the running time is 1 hour, then the user's non-response risk is 30 kW multiplied by 0.6 yuan per kilowatt-hour multiplied by 1 hour, which equals 18 yuan. The total user response risk is obtained by accumulating the losses for all users and time periods.

[0108] Taking into account multiple risk factors, a comprehensive risk index is calculated. The comprehensive risk index is defined according to formula (32): (32) In the formula, For the overall risk score, the weighting coefficients are... Based on the aggregator's risk appetite, conservative aggregators take... radical .

[0109] The total risk score equals the price risk weight multiplied by the normalized price risk exposure, plus the output deviation risk weight multiplied by the normalized CVaR value, plus the user response risk weight multiplied by the normalized user response risk. The weighting coefficients are determined based on the aggregator's risk preference. For conservative aggregators, the three weights are 0.3, 0.5, and 0.2 respectively, focusing more on output deviation risk; for aggressive aggregators, the weights are 0.5, 0.3, and 0.2, focusing more on price opportunities. Normalization involves dividing each risk indicator by its historical maximum value to unify their dimensions.

[0110] C2: Use the conditional value at risk approach to assess losses in extreme scenarios and adjust the day-ahead market reporting scheme according to the risk level.

[0111] Specifically, a risk control threshold is defined, and protective adjustments are triggered when the overall risk index exceeds the threshold. The risk control threshold is set according to the aggregator's risk tolerance: 0.6 for conservative aggregators and 0.8 for aggressive aggregators. When the overall risk index is detected to exceed the threshold, risk-driven decision adjustments are initiated.

[0112] First, adjust the day-ahead reporting strategy to reduce risk exposure. Calculate the adjusted day-ahead reporting power according to formula (33): (33) In the formula, The adjusted day-ahead declared power. The risk response coefficient (taken as 0.3) is used, and this formula takes effect in the day-ahead report of the next day.

[0113] The adjusted day-ahead declared power is equal to the original day-ahead declared power minus the risk response coefficient multiplied by the product of the original day-ahead declared power and the comprehensive risk index minus the risk threshold. The risk response coefficient is set to 0.3, meaning that when the risk exceeds the limit, the declared power is reduced by 30% of the excess. For example, if the original declared power is 1000 kW, the comprehensive risk index is 0.85, and the risk threshold is 0.6, then the declared power is adjusted to 1000 minus 0.3 multiplied by 1000 multiplied by 0.25, which equals 925 kW. By reducing the declared power, price risk exposure and output deviation risk are reduced, but some market opportunities are sacrificed.

[0114] Simultaneously, the reserve margin for real-time scheduling is adjusted to enhance response capabilities. The adjusted reserve capacity is calculated according to formula (34): (34) when hour( To minimize the reserved capacity, take 10% of the adjustable capacity, and reduce daily reporting.

[0115] The adjusted reserved capacity equals the original reserved capacity plus the difference between the comprehensive risk index and the risk threshold, multiplied by the adjustable capacity. However, the reserved capacity cannot be lower than the minimum reserved capacity, which is 10% of the adjustable capacity. For example, if the adjustable capacity is 200 kW, the comprehensive risk index is 0.85, and the risk threshold is 0.6, then the increased reserved capacity is 0.25 multiplied by 200, equaling 50 kW. By increasing the reserved capacity, there is greater adjustment leeway to cope with output deviations during real-time dispatch.

[0116] When the risk level is high, the intraday market reporting power will also be reduced. According to formula (35). (35) When the comprehensive risk index exceeds the risk threshold plus 0.1, the daily declared power is reduced by the risk response coefficient multiplied by the original declared power. This additional protection measure ensures sufficient safety margin in high-risk situations. At the same time, the energy storage dispatch strategy is adjusted. The target SOC level is adjusted according to formula (36): (36) In the formula, The target SOC level is determined by the risk level; the higher the risk, the higher the target SOC.

[0117] The target SOC level equals 0.5 plus the difference between the comprehensive risk index and the risk threshold. The higher the risk, the higher the target SOC. For example, if the comprehensive risk index is 0.85 and the risk threshold is 0.6, then the target SOC is 0.5 plus 0.25, which equals 0.75. A higher SOC level allows energy storage to retain more dischargeable energy, enhancing its ability to cope with insufficient photovoltaic output. These dynamic adjustment measures constitute a complete risk response, capable of adaptively adjusting decision-making strategies based on real-time risk levels, achieving a balance between returns and risks.

[0118] In this embodiment, step S400 involves acquiring updated resource forecast information and market price information in each decision cycle, using the execution status of the previous cycle as the initial status of the current cycle, and recalculating and updating the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions, including the following step D1: D1: During the day-ahead decision-making cycle, obtain the photovoltaic output forecast, load forecast, market price forecast, and current state of charge of energy storage for the next 24 hours, and recalculate the day-ahead market application plan; During the intraday decision-making cycle, obtain updated forecast information for the next 4 hours, and recalculate the intraday adjustment plan based on the previous day's declaration results; During the real-time scheduling cycle, the actual operating status of resources and real-time market prices at the current moment are obtained, and the real-time scheduling instructions are recalculated with the intraday plan as the tracking target. Each time a decision is recalculated, the values ​​of the unexecuted portions of the decision variables from the previous period are used as the initial values ​​for the corresponding time period in the current period to maintain decision continuity.

[0119] Specifically, a complete rolling optimization framework is established to achieve dynamic updates of decisions. The time window for rolling optimization is defined according to formula (41): (41) In the formula, For the optimized window during the k-th scroll, For the current moment, To optimize the time domain (the day before is 24 hours, the daytime is the remaining time period, and the real time is the next hour).

[0120] The optimization window for the k-th rolling iteration equals the time interval from the current moment to the current moment plus the optimization time domain. The optimization time domain varies depending on the decision-making level: for decisions made before the day, the optimization time domain is 24 hours, covering the entire next day; for intraday decisions, the optimization time domain is the remaining runtime, for example, when intraday optimization is performed at 10:00, the optimization window is from 10:00 to 24:00, a total of 14 hours; for real-time scheduling, the optimization time domain is the next hour, divided into 12 time periods with a granularity of 5 minutes.

[0121] At the start of each rolling optimization, the initial state of the optimization model needs to be updated. The initial state set for the k-th optimization is defined according to formula (42): (42) In the formula, This is the initial set of states for the k-th optimization, including the current energy storage SOC, the output of each resource, and the total declared power.

[0122] The initial state set includes the current energy storage SOC, the actual output of each resource, and the total declared power. This state information is obtained from real-time monitoring data and reflects the actual state after implementing the decision-making scheme of the previous cycle. For example, when performing intraday rolling optimization, the current measured SOC value of the energy storage is read. This value may deviate from the SOC predicted at the time of the day-ahead decision because the photovoltaic output and load may deviate from the prediction during the actual execution.

[0123] In rolling optimization, decision variable inheritance is used to maintain the continuity of decisions. According to formula (43): (43) In the formula, Contribute to the new cycle's plan and inherit the optimization results from the previous cycle. Time period value.

[0124] The planned output of the new cycle inherits the values ​​of the unexecuted periods from the optimization results of the previous cycle. Specifically, if the current time is t0, and the planned output obtained from the optimization of the previous cycle covers periods t0 to tN, but only periods t0 to t1 are executed, then the planned output of periods t1 to tN is inherited into the new cycle as the initial value.

[0125] During the day-ahead decision-making cycle, a rolling day-ahead optimization is performed once daily at a specified time. This involves obtaining the latest PV output forecast, load forecast, and market price forecast for the next 24 hours. PV output forecast updates are based on the latest numerical weather forecast data; if the weather forecast changes, the PV forecast curve will be adjusted accordingly. Load forecast updates are based on historical load patterns and special circumstances for the following day, such as holidays. Market price forecast updates are based on the latest market information and competitor bidding activities. Simultaneously, the current state of charge (SOC) of the energy storage is read, reflecting the actual energy level of the storage after the day's operation. Based on this updated information, steps A1 to B2 of the modeling and optimization process are re-executed to generate a new day-ahead market bidding scheme. Compared to the previous day's scheme, the new scheme more accurately reflects the resource conditions and market conditions of the following day.

[0126] During the real-time scheduling cycle, real-time rolling optimization is performed every 5 minutes. Real-time scheduling faces the greatest uncertainty because the instantaneous fluctuations in resource output are the most dramatic. The system acquires the actual operating status of each resource at the current moment, including real-time photovoltaic power, real-time SOC and charging / discharging power of energy storage, and real-time load power. Simultaneously, it acquires real-time market price signals. Using the planned daily power as the tracking target, it calculates the amount of power adjustment required for each resource. The goal of real-time optimization is to minimize the adjustment costs of each resource and the penalty cost of the total deviation while maintaining overall power balance. By rapidly solving the optimization problem, new control commands are generated and issued to each resource within 1 minute.

[0127] Furthermore, a closed-loop feedback mechanism for decision execution and deviation monitoring is implemented, and the power execution deviation is calculated according to formula (37): (37) In the formula, This represents the relative deviation in power.

[0128] The relative power deviation is equal to the difference between the actual total output and the planned total output, divided by the absolute value of the planned total output. This indicator reflects the degree of relative deviation between actual performance and the plan.

[0129] Calculate the economic loss caused by the deviation according to formula (38): (38) Deviation loss equals the real-time deviation penalty price multiplied by the absolute value of the difference between the actual total output and the planned total output. According to formula (39): (39) In the formula, T represents the total number of time periods in the statistical period. The cumulative deviation loss within a week is used as an indicator to evaluate performance.

[0130] The contribution of various resources to the total deviation is analyzed according to formula (40): (40) In the formula, The contribution of the deviation to resource i.

[0131] The deviation contribution of resource i is equal to the difference between the actual output and the planned output of that resource divided by the total deviation. By analyzing the contribution, the main sources of deviation can be identified, and the prediction or scheduling strategies for this type of resource can be improved in a targeted manner.

[0132] Through comprehensive rolling optimization, dynamic decision-making updates are achieved across the entire process, from day-to-day to real-time. Each decision cycle is re-optimized based on the latest information, continuously correcting prediction errors and responding to unexpected changes.

[0133] In summary, a power output prediction model based on numerical weather forecasting is established for photovoltaic resources, and the uncertainty range is quantified through historical prediction error statistics. For energy storage resources, the rechargeable and dischargeable capacities for future periods are calculated, comprehensively considering state of charge and power constraints. For controllable loads, a logistic regression model based on historical user response data is established to predict user response probabilities. In the aggregation modeling stage, the upward and downward adjustment capacity of the aggregation resource pool is comprehensively calculated by analyzing the temporal complementarity characteristics of photovoltaic power output and load fluctuations, and the ability of energy storage to mitigate photovoltaic fluctuations, with reliability improved through confidence coefficient correction. In the multi-timescale optimization stage, a three-layer mapping relationship is established: day-ahead, intraday, and real-time. Adjustment margins are reserved and subsequent costs are considered in day-ahead decisions; rolling optimization is used to correct deviations in intraday decisions; and control commands are generated in real-time scheduling. In the risk management stage, the conditional value at risk method is used to assess losses in extreme scenarios, and protective adjustments are triggered through risk thresholds.

[0134] Example 3 This simulation example is based on operational data from a distributed resource aggregator in a certain region. The resources managed by this aggregator include: 5MW of distributed photovoltaic power, 2MWh / 1MW of energy storage systems, and 3MW of controllable load (including 2MW of industrial load and 1MW of commercial air conditioning load). The simulation period is 30 consecutive days, including various weather types such as sunny, cloudy, and rainy. Actual price data from a provincial electricity market are used, with day-ahead market prices ranging from 0.2 to 0.8 yuan / kWh, intraday market price fluctuations of ±20%, and the real-time deviation assessment price being 1.5 times the day-ahead price.

[0135] A simulation model was built on the MATLAB platform. Photovoltaic power output prediction used numerical weather prediction data, with a root mean square error (RMSE) of approximately 15% of the installed capacity. Load response model parameters were fitted using historical response data, with an average user response rate of approximately 70%. The initial state of energy storage (SOC) was set to 50%, with an efficiency of 95%. Simulations compared three decision-making methods: the traditional day-ahead single optimization method, the day-ahead-intraday two-stage optimization method, and the multi-timescale collaborative optimization method of this patent.

[0136] 2. Validity Verification Form Table 1: Comparison of Returns and Deviations

[0137] Although the current daytime returns of this patented method are slightly lower than those of traditional methods, the total returns are increased by 19.7% and the power deviation rate is reduced to 7.8% by reducing deviation losses through precise intraday adjustments and real-time refined scheduling.

[0138] Table 2: Comparison of Resource Utilization Efficiency

[0139] This patented method improves the overall resource utilization rate to 82.4% through refined aggregate modeling and multi-timescale collaborative optimization, with the most significant improvement in the utilization rate of energy storage systems, reaching 83.6%.

[0140] Table 3: Comparison of Risk Control Effectiveness

[0141] This patented method reduces the overall risk index by 55.3% through dynamic risk assessment and risk-based decision adjustment. In extreme price fluctuation scenarios, losses are controlled within 41,000 yuan, and risk-adjusted returns are increased by 31.7%.

[0142] Example 4 The above is an illustrative scheme of a distributed resource aggregator multi-timescale collaborative decision optimization method. It should be noted that the technical solution of this distributed resource aggregator multi-timescale collaborative decision optimization system and the technical solution of the aforementioned distributed resource aggregator multi-timescale collaborative decision optimization method belong to the same concept. Details not described in detail in this embodiment of the distributed resource aggregator multi-timescale collaborative decision optimization system can be found in the description of the aforementioned distributed resource aggregator multi-timescale collaborative decision optimization method.

[0143] This embodiment also provides a distributed resource aggregator multi-timescale collaborative decision-making optimization system, including: The data acquisition module collects real-time operating data of distributed resources. Based on the real-time operating data, it establishes photovoltaic output prediction, energy storage status prediction, and load response prediction respectively, quantifies the prediction error of various resources, establishes individual resource adjustment capacity constraints according to the operating characteristics of distributed resources, and forms the adjustable capacity boundary of the aggregated resource pool through the complementary characteristics and coupling constraints between resources. The time-series coordination module establishes a time mapping relationship between the day-ahead decision-making period, the intraday decision-making period, and the real-time scheduling period. By reserving adjustment margins in day-ahead decisions and deviation buffers in intraday decisions, it achieves decision-making connection between different time scales, obtains the predicted values ​​of day-ahead market prices, intraday market prices, and real-time equilibrium prices, and generates day-ahead market reporting plans, intraday adjustment plans, and real-time scheduling instructions by combining the adjustable capacity boundary and the time mapping relationship. The risk management module quantifies the risk exposure of market price deviations from forecast values ​​and the deviation risk of resource output from planned values, uses the conditional value at risk approach to assess losses in extreme scenarios, and adjusts the day-ahead market reporting plan according to the risk level. The rolling correction module acquires updated resource forecast information and market price information in each decision cycle, uses the execution status of the previous cycle as the initial status of the current cycle, and recalculates and updates the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions.

[0144] This embodiment also provides an electronic device suitable for multi-timescale collaborative decision optimization of distributed resource aggregators, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multi-timescale collaborative decision optimization method for distributed resource aggregators proposed in the above embodiment.

[0145] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the multi-timescale collaborative decision optimization method for distributed resource aggregators proposed in the above embodiments.

[0146] The storage medium proposed in this embodiment and the method for implementing multi-timescale collaborative decision optimization of distributed resource aggregators proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0147] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0148] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-timescale collaborative decision-making optimization method for distributed resource aggregators, characterized in that: This includes collecting real-time operational data of distributed resources, establishing photovoltaic output forecasts, energy storage status forecasts, and load response forecasts based on the real-time operational data, quantifying the forecast errors of various resources, establishing individual resource adjustment capacity constraints for the operational characteristics of distributed resources, and forming an adjustable capacity boundary for the aggregated resource pool through the complementary characteristics and coupling constraints between resources. Establish a time mapping relationship between the day-ahead decision period, the intraday decision period, and the real-time scheduling period. By reserving adjustment margin in the day-ahead decision and deviation buffer in the intraday decision, the decision-making connection between different time scales is realized. Obtain the predicted values ​​of the day-ahead market price, the intraday market price, and the real-time equilibrium price. Combine the adjustable capacity boundary and the time mapping relationship to generate the day-ahead market reporting plan, the intraday adjustment plan, and the real-time scheduling instruction. The risk exposure of market price deviation from forecast value and the deviation risk of resource output from planned value are quantified. The conditional value at risk method is used to assess the loss in extreme scenarios, and the day-ahead market reporting plan is adjusted according to the risk level. In each decision cycle, updated resource forecast information and market price information are obtained. The execution status of the previous cycle is used as the initial status of the current cycle, and the day-ahead market reporting plan, intraday adjustment plan and real-time scheduling instructions are recalculated and updated.

2. The distributed resource aggregator multi-timescale collaborative decision-making optimization method as described in claim 1, characterized in that: The aforementioned constraints on the adjustment capabilities of individual resources are established based on the operational characteristics of distributed resources, including photovoltaic resources, energy storage resources, and controllable load resources. For photovoltaic resources, the range of available photovoltaic power is calculated based on weather forecast data and photovoltaic rated capacity, and the uncertainty range of photovoltaic output is determined based on the statistical characteristics of historical forecast errors. For energy storage resources, the rechargeable and dischargeable capacity of energy storage in the future period is calculated based on the current state of charge, charging and discharging power limits and charging and discharging efficiency. For controllable load resources, the load reduction and increase amounts are calculated based on the load baseline power and user historical response data, and the uncertainty of user response is quantified.

3. The distributed resource aggregator multi-timescale collaborative decision-making optimization method as described in claim 2, characterized in that: The process of forming an adjustable capacity boundary for the aggregated resource pool through the complementary characteristics and coupling constraints between resources includes analyzing the temporal complementary characteristics of photovoltaic power output fluctuations and load fluctuations, and evaluating the fluctuation offsetting effect brought about by natural complementarity. Analyze the coordination relationship between energy storage charging and discharging regulation and photovoltaic power output fluctuations, and evaluate the ability of energy storage to mitigate the uncertainty of photovoltaic power output. By considering the available power range of photovoltaic resources, the bidirectional regulation capability of energy storage resources, and the regulation potential of controllable load resources, the upward and downward regulation capacity of the aggregated resource pool in each time period is calculated.

4. The distributed resource aggregator multi-timescale collaborative decision-making optimization method as described in claim 3, characterized in that: The establishment of the time mapping relationship between the day-ahead decision period, the intraday decision period, and the real-time scheduling period includes mapping the 24-hour period of the day-ahead decision to the 96 15-minute periods of the intraday decision, and reserving an adjustment margin between the day-ahead declared power and the adjustable capacity boundary of the aggregated resource pool. The 15-minute timeframe for intraday decision-making is mapped to a 5-minute timeframe for real-time scheduling, reserving a buffer space for deviations based on the intraday planned power. In the day-ahead decision-making process, intraday adjustment costs and real-time deviation costs are considered. Potential costs are assessed by the probability distribution of forecast errors, and cost compensation is carried out in the day-ahead declaration.

5. The distributed resource aggregator multi-timescale collaborative decision-making optimization method as described in claim 4, characterized in that: The process of obtaining the predicted values ​​of the day-ahead market price, intraday market price, and real-time equilibrium price, and combining them with the adjustable capacity boundary and time mapping relationship, generates a day-ahead market application plan, an intraday adjustment plan, and a real-time scheduling instruction. This includes optimizing the application power during the day-ahead decision period based on the day-ahead market price prediction and the adjustable capacity boundary of the aggregated resource pool, thereby maximizing the net profit after deducting the expected intraday adjustment cost and the expected real-time deviation cost from the day-ahead market profit. Based on intraday market price forecasts and day-ahead reporting results, power is optimized and adjusted during the intraday decision-making period to correct day-ahead plan deviations and reduce real-time deviation risks. Based on real-time balanced prices and actual resource operating status, power instructions for each distributed resource are generated during the real-time scheduling period to track intraday adjustment plans.

6. The distributed resource aggregator multi-timescale collaborative decision-making optimization method as described in claim 5, characterized in that: The risk exposure to deviations between the quantitative market price and the forecast, and the deviation risk between the resource output and the planned value, include: Statistically analyze historical market price forecast errors, calculate the probability distribution of price deviations from forecast values, and assess the risk of profit loss due to price fluctuations; Statistically analyze historical resource output prediction errors, calculate the probability distribution of output deviation from planned values, and assess the risk of performance losses caused by output deviations; Set a confidence level and use the conditional value at risk method to calculate the expected loss beyond the confidence level, thus quantifying the maximum risk exposure in extreme scenarios; Set a risk threshold, and when the calculated risk exposure exceeds the threshold, reduce the day-ahead market reporting power or increase the adjustment margin.

7. The distributed resource aggregator multi-timescale collaborative decision-making optimization method as described in claim 6, characterized in that: The process involves acquiring updated resource forecast information and market price information in each decision-making cycle, using the execution status of the previous cycle as the initial state of the current cycle, and recalculating and updating the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions. During the day-ahead decision-making cycle, obtain the photovoltaic output forecast, load forecast, market price forecast, and current state of charge of energy storage for the next 24 hours, and recalculate the day-ahead market application plan; During the intraday decision-making cycle, obtain updated forecast information for the next 4 hours, and recalculate the intraday adjustment plan based on the previous day's declaration results; During the real-time scheduling cycle, the actual operating status of resources and real-time market prices at the current moment are obtained, and the real-time scheduling instructions are recalculated with the intraday plan as the tracking target. Each time a decision is recalculated, the values ​​of the unexecuted portions of the decision variables from the previous period are used as the initial values ​​for the corresponding time period in the current period to maintain decision continuity.

8. A distributed resource aggregator multi-timescale collaborative decision-making optimization system, based on the distributed resource aggregator multi-timescale collaborative decision-making optimization method according to any one of claims 1 to 7, characterized in that: It also includes a data acquisition module, which collects real-time operating data of distributed resources, establishes photovoltaic output prediction, energy storage status prediction and load response prediction based on real-time operating data, quantifies the prediction error of various resources, establishes individual resource adjustment capacity constraints according to the operating characteristics of distributed resources, and forms an adjustable capacity boundary of the aggregated resource pool through the complementary characteristics and coupling constraints between resources. The time-series coordination module establishes a time mapping relationship between the day-ahead decision-making period, the intraday decision-making period, and the real-time scheduling period. By reserving adjustment margins in day-ahead decisions and deviation buffers in intraday decisions, it achieves decision-making connection between different time scales, obtains the predicted values ​​of day-ahead market prices, intraday market prices, and real-time equilibrium prices, and generates day-ahead market reporting plans, intraday adjustment plans, and real-time scheduling instructions by combining the adjustable capacity boundary and the time mapping relationship. The risk management module quantifies the risk exposure of market price deviations from forecast values ​​and the deviation risk of resource output from planned values, uses the conditional value at risk approach to assess losses in extreme scenarios, and adjusts the day-ahead market reporting plan according to the risk level. The rolling correction module acquires updated resource forecast information and market price information in each decision cycle, uses the execution status of the previous cycle as the initial status of the current cycle, and recalculates and updates the day-ahead market reporting plan, intraday adjustment plan, and real-time scheduling instructions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the distributed resource aggregator multi-timescale collaborative decision optimization method according to 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 the processor, it implements the steps of the distributed resource aggregator multi-timescale collaborative decision optimization method according to any one of claims 1 to 7.

Images