A flexible load aggregation modeling method and device based on price response, a terminal device, and a storage medium

Abstract

This invention discloses a method, apparatus, terminal equipment, and storage medium for flexible load aggregation modeling based on price response, belonging to the field of load dispatching. The method includes: constructing a physical aggregation model and corresponding aggregation model constraints based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data; solving the model under constraints based on the acquired electricity purchase price sequence and predicted electricity purchase price, with the objective of minimizing electricity purchase cost; obtaining the optimal model parameters by inverse optimization of the physical aggregation model based on the solved initial aggregation response power trajectory, initial power allocation results, and measured response power data from the load data, with the objective of minimizing the mean square error between the observed response and the simulated response; and performing grid dispatching based on the target aggregation response trajectory, target power allocation results, and grid topology obtained from the updated physical aggregation model. By implementing this invention, the problem of prior art neglecting the dynamics and real-time rolling of flexible loads is solved.

Description

Technical Field

[0001] This invention relates to the field of flexible load dispatching technology, and in particular to a flexible load aggregation modeling method, apparatus, terminal equipment, and storage medium based on price response. Background Technology

[0002] With the widespread integration of renewable energy sources (RES) into the power system, the supply and demand balance of the power system faces significant uncertainties and intermittency issues. To alleviate power imbalances in the power system, demand-side flexible loads (FLs), such as interruptible loads, transferable loads, thermally controlled loads, distributed generation, and energy storage devices, can autonomously adjust their electricity consumption capacity in response to electricity price signals, becoming an important means to alleviate power imbalances and ensure the safe operation of the power system.

[0003] Currently, the price-response mechanism based on electricity price signals is a core foundation for accurately characterizing load response and identifying parameters, thus fully tapping the potential of flexible load regulation and supporting optimized power system dispatch. By collecting historical operational data of flexible loads, electricity price signals, and power response data, methods such as machine learning, statistical modeling, or inverse optimization are used to directly fit the price response characteristics of aggregated models. However, most methods only assume a static response mechanism, neglecting the dynamic decision-making and real-time rolling adjustment processes of flexible loads in actual systems, resulting in insufficient model universality and practicality. Summary of the Invention

[0004] This invention provides a method, apparatus, terminal device, and storage medium for flexible load aggregation modeling based on price response, which can effectively solve the problem that existing technologies neglect the dynamic decision-making and real-time rolling adjustment process of flexible loads in actual systems, resulting in insufficient model universality.

[0005] One embodiment of the present invention provides a flexible load aggregation modeling method based on price response, comprising: Obtain load data, energy storage operation status data, distributed generation capacity data, grid topology, electricity purchase price series, and electricity purchase price forecast for the distribution network to be optimized; Based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, a physical aggregation model is constructed to characterize fixed load, adjustable power, and energy storage aggregation. The aggregation model constraints corresponding to the physical aggregation model are also constructed. The aggregation model constraints include: fixed load uncertainty constraints, load power upper and lower bound constraints, ramping power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. Based on the electricity purchase price series and the predicted electricity purchase price, with the goal of minimizing the electricity purchase cost, the physical aggregation model is optimized for the electricity price scenario under the constraints of the aggregation model. The initial aggregation response power trajectory and initial power allocation results for each time period are obtained by solving the problem. With the goal of minimizing the mean square error between the observed response and the simulated response, the physical aggregation model is solved by inverse optimization in two layers based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data. The optimal model parameters of the physical aggregation model are obtained, and the physical aggregation model is updated based on the optimal model parameters to obtain the final physical aggregation model. The final physical aggregation model is re-solved to obtain the final target aggregation response trajectory and target power allocation result. Power grid scheduling is then carried out based on the target aggregation response trajectory, target power allocation result, and power grid topology.

[0006] Furthermore, the load data also includes: uncontrollable load data and adjustable load data; the uncontrollable load data includes: uncontrollable load forecast values ​​and uncontrollable load disturbance standard deviation; the adjustable load data includes: maximum adjustable capacity, adjustable ratio, earliest start-up / stop time, latest start-up / stop time, minimum continuous operating time, and response speed; the energy storage operation status data includes: initial energy storage capacity, upper and lower limits of energy storage capacity, upper and lower limits of charge / discharge power, and energy storage efficiency; the distributed generation capacity data includes: photovoltaic output forecast values, photovoltaic output forecast errors, and curtailment. Based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, a physical aggregation model is constructed to characterize stationary loads, adjustable power, and energy storage aggregation, including: The uncontrollable load prediction values ​​are indexed and assigned using a preset discrete-time index to generate a power trajectory vector. A Gaussian distribution is calculated based on the standard deviation of the uncontrollable load disturbance and the power trajectory vector to construct a fixed load model. The upper and lower limits of adjustable power load are determined based on the maximum adjustable capacity and adjustable ratio. The upper and lower limits of ramp power are determined based on the response speed. The adjustment time range is determined based on the earliest start-stop time, the latest start-stop time, and the minimum continuous running time. An adjustable power load model is constructed based on the upper and lower limits of adjustable power load, the upper and lower limits of ramp power, and the adjustment time range. Based on the initial energy storage capacity, upper and lower limits of energy storage capacity, upper and lower limits of charge and discharge power, and energy storage efficiency, an energy storage model is constructed to characterize the relationship between energy storage power and energy state. Based on the predicted photovoltaic output, the predicted photovoltaic output error, and the amount of curtailed solar power, the net load is calculated. Then, based on the fixed load model, the adjustable power load model, the energy storage model, and the net load, a physical aggregation model is constructed to characterize the aggregation of fixed load, adjustable power, and energy storage.

[0007] Furthermore, based on the electricity purchase price series and predicted electricity purchase prices, and with the goal of minimizing electricity purchase costs, the physical aggregation model is optimized for electricity price scenario response under the constraints of the aggregation model. The initial aggregation response power trajectory and initial power allocation results for each time period are obtained, including: Determine the corresponding electricity price response scenario based on the type of electricity purchase price series; In the case of time-of-use pricing for electricity price response, based on the fixed electricity price in the electricity purchase price sequence, with the goal of minimizing the total electricity purchase cost for all time periods within the preset time domain, the physical aggregation model is solved under the constraints of the aggregation model to obtain the initial aggregation response power trajectory and initial power allocation results for each time period. In the case of a real-time electricity price response scenario, starting from the current moment, a segment of the predicted electricity purchase price for a preset decision period is selected to generate a rolling price vector. Based on the rolling price vector, with the objective of minimizing the total electricity purchase cost within a preset rolling window, the physical aggregation model is solved under the constraints of the aggregation model to obtain the initial aggregation response power trajectory and initial power allocation results for each rolling window. The results of the first period of each rolling window are then spliced ​​together to obtain the initial aggregation response power trajectory and initial power allocation results for the entire period.

[0008] Furthermore, with the goal of minimizing the mean square error between the observed and simulated responses, the physical aggregation model is solved using a two-level inverse optimization based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data from the load data. This yields the optimal model parameters for the physical aggregation model, including: The aggregated response power is obtained based on the initial aggregated response power trajectory and the initial power allocation results. The aggregated response power is used as the simulation response, and the measured response power data of the period corresponding to the initial aggregated response trajectory in the load data is used as the observed response. The aggregation model constraint is used as the inner constraint of the inverse optimization bi-level problem. The goal is to minimize the electricity purchase cost under the corresponding electricity price response scenario. The simulated response power and simulated electricity purchase cost under the current candidate parameters are obtained by solving the inner optimization. The current candidate parameters include: the power of the adjustable power load, the ramp power, the energy storage charging and discharging power, and the energy capacity. The outer objective function is constructed with the goal of minimizing the mean square error between the observed response and the simulated response; Based on the simulated response power and simulated electricity purchase cost obtained from the inner layer optimization solution, Bayesian optimization is performed on the outer layer objective function until the mean square error difference is less than a preset threshold. The current candidate parameter corresponding to the minimum mean square error is then obtained as the optimal model parameter.

[0009] Furthermore, it also includes: performing noise robustness modeling when constructing the outer objective function; the robustness modeling includes: The fixed load forecasting error is modeled as a first Gaussian noise with a preset fixed mean and a fixed covariance matrix. The observation error of the aggregate response is modeled as a second Gaussian noise with a preset aggregate mean and a covariance of the aggregate covariance matrix. The outer objective function is corrected based on the first Gaussian noise and the second Gaussian noise; wherein the corrected outer objective function is the sum of the mean square error of the observed response and the simulation response plus the noise influence term, and the noise influence term is the sum of the traces of the covariance matrices of the first Gaussian noise and the second Gaussian noise.

[0010] Furthermore, the target power allocation results include: the base power allocation for fixed loads, the adjustment power allocation for adjustable power loads at different times, the charging and discharging power allocation for energy storage systems, and the complementary power allocation between distributed generation and loads. Power grid dispatching is performed based on the target aggregated response trajectory, target power allocation results, and power grid topology, including: Based on the target aggregated response trajectory and combined with the real-time operating status of the distribution network, the active power supply and demand of each node are dynamically matched. Based on the grid topology and branch capacity limitations, verify whether the branch power flow corresponding to the target power allocation result exceeds the safety threshold. If it does, adjust the adjustment range of the adjustable power load and the energy storage charging and discharging power. Based on the target power allocation results, the scheduling strategy is dynamically adjusted by prioritizing the use of flexible resources with the lowest adjustment costs, combined with the electricity purchase price sequence.

[0011] As an improvement to the above solution, another embodiment of the present invention provides a flexible load aggregation modeling device based on price response, comprising: The data acquisition module is used to acquire load data, energy storage operation status data, distributed generation capacity data, grid topology, electricity purchase price series, and electricity purchase price forecast of the distribution network to be optimized. The initial aggregation model construction module is used to construct a physical aggregation model to characterize fixed load, adjustable power, and energy storage aggregation based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, and to construct the aggregation model constraints corresponding to the physical aggregation model. The aggregation model constraints include: fixed load uncertainty constraints, load power upper and lower bound constraints, ramping power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. The initial model solving module is used to optimize the physical aggregation model for electricity price scenario response under the constraints of the aggregation model, based on the electricity purchase price sequence and the predicted electricity purchase price, with the goal of minimizing the electricity purchase cost. It solves for the initial aggregation response power trajectory and the initial power allocation results for each time period. The model inverse optimization module is used to minimize the mean square error between the observed response and the simulated response. It performs a two-level inverse optimization solution on the physical aggregation model based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data to obtain the optimal model parameters of the physical aggregation model. The physical aggregation model is then updated based on the optimal model parameters to obtain the final physical aggregation model. The power grid dispatch module is used to re-solve the final physical aggregation model to obtain the final target aggregation response trajectory and target power allocation result, and to perform power grid dispatch based on the target aggregation response trajectory, target power allocation result and power grid topology.

[0012] Furthermore, the power grid dispatching module is used to perform power grid dispatching based on the target aggregated response trajectory, the target power allocation result, and the power grid topology, including: Based on the target aggregated response trajectory and combined with the real-time operating status of the distribution network, the active power supply and demand of each node are dynamically matched. Based on the grid topology and branch capacity limitations, verify whether the branch power flow corresponding to the target power allocation result exceeds the safety threshold. If it does, adjust the adjustment range of the adjustable power load and the energy storage charging and discharging power. Based on the target power allocation results, the scheduling strategy is dynamically adjusted by prioritizing the use of flexible resources with the lowest adjustment costs, combined with the electricity purchase price sequence.

[0013] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a price-response-based flexible load aggregation modeling method as described in the above embodiments.

[0014] Another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the price-response-based flexible load aggregation modeling method described in the above embodiment.

[0015] By implementing this invention, at least the following beneficial effects are achieved: This invention provides a price-response-based flexible load aggregation modeling method, apparatus, terminal equipment, and storage medium. The method acquires multi-dimensional data, including distribution network load data, energy storage operation status data, and distributed generation capacity data, providing accurate basic inputs for physical aggregation model construction. This avoids modeling bias caused by missing data and ensures model reliability. A unified physical aggregation model covering fixed loads, adjustable power loads, and energy storage systems is constructed, achieving integrated characterization of different types of flexible resources. This solves the problem of difficult collaborative modeling of multiple load types in existing technologies and improves model compatibility. Aggregation model constraints include fixed load uncertainty constraints, load power upper and lower bound constraints, ramp power constraints, energy storage charging and discharging power constraints, and energy capacity constraints, comprehensively covering the core physical characteristics of loads and energy storage. This ensures that the model output conforms to actual operating patterns and improves modeling accuracy. Then, with the goal of minimizing electricity purchase cost, electricity price scenario response optimization is performed under the aggregation model constraints. This yields an initial aggregation response power trajectory and power allocation result that meets economic operation requirements, providing a reliable benchmark for subsequent parameter identification. The optimal model parameters are then solved through a bi-level inverse optimization problem. By combining measured response power data with the mean square error between observed and simulated responses, the initial model parameter deviations are effectively corrected. The updated model better reflects actual load characteristics, resolving the insufficient adaptability issue caused by fixed parameters in traditional models. The final aggregated response trajectory and power allocation results can be directly used for grid dispatching, fully considering load characteristics, energy storage constraints, and electricity price impacts. This provides accurate and feasible decision-making basis for grid dispatching, contributing to improved safety and economy of power system operation. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a flexible load aggregation modeling method based on price response provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a flexible load aggregation modeling device based on price response provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1To address the problem of insufficient model versatility caused by neglecting the dynamic decision-making and real-time rolling adjustment process of flexible loads in actual systems in existing technologies, an embodiment of the present invention provides a flowchart of a flexible load aggregation modeling method based on price response, including: S1. Obtain load data, energy storage operation status data, distributed generation capacity data, grid topology, electricity purchase price series, and electricity purchase price forecast of the distribution network to be optimized; Preferably, the load data further includes: uncontrollable load data and adjustable load data; the uncontrollable load data includes: uncontrollable load forecast value and uncontrollable load disturbance standard deviation; the adjustable load data includes: maximum adjustable capacity, adjustable ratio, earliest start-up / stop time, latest start-up / stop time, minimum continuous running time, and response speed; the energy storage operation status data includes: initial energy storage capacity, upper and lower limits of energy storage capacity, upper and lower limits of charge / discharge power, and energy storage efficiency; the distributed generation capacity data includes: photovoltaic output forecast value, photovoltaic output forecast error, and curtailment amount.

[0019] Specifically, the distribution network to be optimized refers to a distribution network system that requires flexible load dispatch optimization, covering various scenarios such as industrial distribution networks, park distribution networks, and urban distribution networks. Load data includes fixed load data that is insensitive to electricity prices (uncontrollable load data, i.e., load-related data that is insensitive to electricity prices and cannot be adjusted), adjustable load data that responds to electricity price adjustments (i.e., load-related data that responds to electricity price adjustments), and measured response power data during load operation. Energy storage operation status data characterizes the operating characteristics of energy storage devices, including initial energy storage capacity. Energy storage capacity limit and lower limit Upper limit of charging and discharging power and lower limit Energy storage efficiency Distributed generation capacity data refers to output-related data of distributed power sources (such as photovoltaic and wind power), including output forecasts. Prediction error (Deviation between predicted and actual photovoltaic output), curtailment of solar power (The deviation between the predicted and actual photovoltaic output).

[0020] S2. Based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, construct a physical aggregation model to characterize fixed load, adjustable power, and energy storage aggregation, and construct the aggregation model constraints corresponding to the physical aggregation model; the aggregation model constraints include: fixed load uncertainty constraints, load power upper and lower bound constraints, ramping power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. Specifically, the physical aggregation model is a mathematical model used to uniformly characterize the aggregation response characteristics of fixed loads, adjustable power loads, and energy storage systems. Its core expression is: ,in For fixed load component collection, For adjustable power component collection, It is a collection of energy storage components, and all three are collections of component indices. For fixed load models, For adjustable power load models, For the energy storage model, the power trajectory vectors of the three types of components are given. The aggregation model constraints are the constraints that ensure the physical aggregation model conforms to actual operating laws, including: fixed load uncertainty constraints, load power upper and lower bound constraints, ramp power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. Fixed load uncertainty constraints are probabilistic constraints characterizing the fixed load prediction error, corresponding to the formula... The fixed load follows a Gaussian distribution. This represents the actual power value of the i-th fixed load at time t. Indicates a Gaussian distribution. This represents the predicted average power of the i-th fixed load. This represents the standard deviation of the predicted power for the i-th fixed load; the upper and lower bound constraints on load power are the limits on the operating power range of adjustable loads, corresponding to the formula... , This represents the lower power limit of the j-th adjustable load. This represents the upper limit of the power of the j-th adjustable load. This represents the power value of the j-th adjustable load at time t, where T represents the time index set; the ramp power constraint is the limit on the rate of change of adjustable load power, corresponding to the formula... , This represents the lower limit of the power decrease rate (negative ramp rate) for the j-th adjustable load. This represents the upper limit of the power ramp rate (positive ramp rate) for the j-th adjustable load. This represents the power value of the j-th adjustable load at time t-1. Indicates the time step (the interval between adjacent moments); the energy storage charging and discharging power constraint is the limitation on the charging and discharging power range of the energy storage device, corresponding to the formula... , This represents the charging and discharging power value of the i-th energy storage device at time t; the energy capacity constraint is the energy state range limit of the energy storage device, corresponding to the formula... , This represents the energy state of the i-th energy storage device at time t.

[0021] Schematic diagram: Based on uncontrollable load data, a fixed load model is constructed. The uncontrollable load prediction values ​​can be indexed using a discrete-time index set T={1,2,…,|T|} to generate a fixed load power trajectory vector (fixed load model). The sample variance is calculated by combining the standard deviation of the disturbance. N is the sample size. For the power prediction residual (the difference between the actual value and the predicted value) of the nth sample of the i-th fixed load at time t, establish a Gaussian distribution probability model. Based on adjustable load data, an adjustable power load model is constructed. The upper and lower bounds of power can be determined by the maximum adjustable capacity and the adjustable ratio, and the upper and lower bounds of the ramp rate can be determined by the response speed. Combined with the adjustment time range, a time series model containing power constraints and ramp constraints is established. ; ; ; ; ;in For dimension A column vector of all 1s.

[0022] in, This is the adjustable load power trajectory vector (adjustable power load model). This indicates that the j-th adjustable load occurs at time t (t=1,2,…). The power value of ) This represents the total number of moments within a time period; the vector's dimension is the same as the number of time points. The dimension is The real vector space indicates that the power vector is a A dimensional real column vector, This represents the cumulative ramp lower limit vector for the j-th adjustable load, consisting of a single ramp rate lower limit. Multiplying by the all-1 vector S yields the result. This represents the cumulative ramp rate upper limit vector for the j-th adjustable load, consisting of a single ramp rate upper limit. Multiplying by the all-1 vector S yields the result. Let represent the cumulative power change vector of the j-th adjustable load, and let represent the vector composed of power changes at multiple consecutive moments.

[0023] Based on energy storage operation status data, an energy storage model can be constructed to establish the dynamic relationship between energy storage power and energy state. The energy update equation is as follows: And apply charging and discharging power constraints and energy capacity constraints. Based on distributed generation data, net load (the difference between load demand and distributed generation output) is calculated. A unified physical aggregation model is then constructed by combining three sub-models. And integrate all constraints to form an aggregated model constraint set: .

[0024] in, A projection / transpose vector representing the initial energy state, used to convert the scalar initial energy into a form that matches the time vector; This represents the energy accumulation matrix of the i-th energy storage device, which is usually a lower triangular matrix used to accumulate the energy corresponding to the power change at each time step.

[0025] Preferably, based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, a physical aggregation model is constructed to characterize fixed loads, adjustable power, and energy storage aggregation, including: The uncontrollable load prediction values ​​are indexed and assigned using a preset discrete-time index to generate a power trajectory vector. A Gaussian distribution is calculated based on the standard deviation of the uncontrollable load disturbance and the power trajectory vector to construct a fixed load model. The upper and lower limits of adjustable power load are determined based on the maximum adjustable capacity and adjustable ratio. The upper and lower limits of ramp power are determined based on the response speed. The adjustment time range is determined based on the earliest start-stop time, the latest start-stop time, and the minimum continuous running time. An adjustable power load model is constructed based on the upper and lower limits of adjustable power load, the upper and lower limits of ramp power, and the adjustment time range. Based on the initial energy storage capacity, upper and lower limits of energy storage capacity, upper and lower limits of charge and discharge power, and energy storage efficiency, an energy storage model is constructed to characterize the relationship between energy storage power and energy state. Based on the predicted photovoltaic output, the predicted photovoltaic output error, and the amount of curtailed solar power, the net load is calculated. Then, based on the fixed load model, the adjustable power load model, the energy storage model, and the net load, a physical aggregation model is constructed to characterize the aggregation of fixed load, adjustable power, and energy storage.

[0026] Specifically, net load is the load value after subtracting the output of distributed generation from the total load demand of the distribution network, that is, net load = total load demand - photovoltaic output forecast + curtailment.

[0027] To illustrate, the data reported by the distribution network is used to filter out electricity-price-insensitive equipment that does not participate in regulation (such as data center base loads and security lighting). Equipment with similar operating characteristics is aggregated into a fixed load component set, and historical power measurements from multiple days are collected. For each fixed load component i, the average power of each time period t is calculated according to the daily cycle: Generate typical fixed load curves ; Calculate the prediction residuals for each time period Solve for the residual sample variance Construct a Gaussian distribution model .

[0028] Based on the maximum adjustable capacity and adjustable ratio, calculate the minimum and maximum operating power of the adjustable power load; based on the response speed, determine the fastest allowable power reduction rate (minimum ramp rate is usually negative) and the fastest allowable power increase rate (maximum ramp rate is positive) per unit time; based on the earliest start / stop time, latest start / stop time, and minimum continuous running time, clarify the feasible adjustment period for the adjustable load, such as industrial production lines can only be adjusted from 8:00 to 22:00, and must run continuously for at least 2 hours after adjustment; integrate the upper and lower limits of power, ramp power, and adjustment time range to establish a time-series response model for the adjustable power load. and impose constraints and .

[0029] Based on the initial energy storage capacity and energy storage efficiency, an energy storage state update equation is established. By imposing charging and discharging power constraints and energy capacity constraints, an energy storage model is constructed. .

[0030] Based on the predicted photovoltaic output, the prediction error of photovoltaic output, and the amount of curtailed solar power, the total net load demand for each period is calculated; the fixed load model, adjustable power load model, energy storage model and net load are combined and substituted into the aggregation model expression to construct a complete physical aggregation model.

[0031] By integrating the net load calculation of distributed generation, the impact of distributed power sources on flexible load response is considered, improving the practicality of the model; the constraints are highly consistent with the physical mechanisms, ensuring the physical rationality and interpretability of the model.

[0032] S3. Based on the electricity purchase price sequence and the predicted electricity purchase price, with the goal of minimizing the electricity purchase cost, the physical aggregation model is optimized for the electricity price scenario under the constraints of the aggregation model, and the initial aggregation response power trajectory and initial power allocation results for each time period are obtained. Specifically, the electricity price scenario response optimization means selecting the corresponding optimization strategy according to the electricity price type (time-of-use pricing, real-time pricing) and solving for the optimal response power trajectory of flexible loads.

[0033] Schematic illustration: The electricity price response scenario can be determined based on the type of electricity purchase price sequence, classifying it as a time-of-use pricing scenario (fixed price period) or a real-time pricing scenario (prices fluctuate in real time). Optimization of the time-of-use pricing scenario aims to minimize the total electricity purchase cost over a preset time period (e.g., 24 hours), with the objective function being: ,in, Indicates the preset time domain The set of constraints in the aggregate model includes all physical operation constraints such as power upper and lower bounds, ramping constraints, and energy storage charging and discharging constraints. θ is the parameter of the constraint set (such as the characteristic parameters of load and energy storage). This represents the energy state of the i-th energy storage device at the end of the preset time domain. This represents the initial energy state of the i-th energy storage device. Under the constraints of the aggregation model, the initial aggregated response power trajectory and initial power allocation results for each time period are obtained. Real-time electricity price scenario optimization starts from the current time and selects the predicted electricity purchase price value for a preset decision period (e.g., 1 hour) to generate a rolling price vector. The objective function is to minimize the electricity purchase cost within the rolling window. , The average predicted electricity price at the current time t is used to estimate the long-term electricity price level outside the rolling window. The initial results for each rolling window are obtained by solving the problem, and the results of the first time period are spliced ​​together to obtain the initial trajectory and allocation results for the entire time period.

[0034] Preferably, based on the electricity purchase price sequence and the predicted electricity purchase price, with the objective of minimizing the electricity purchase cost, the physical aggregation model is optimized for the electricity price scenario response under the constraints of the aggregation model. The initial aggregation response power trajectory and initial power allocation results for each time period are obtained, including: Determine the corresponding electricity price response scenario based on the type of electricity purchase price series; In the case of time-of-use pricing for electricity price response, based on the fixed electricity price in the electricity purchase price sequence, with the goal of minimizing the total electricity purchase cost for all time periods within the preset time domain, the physical aggregation model is solved under the constraints of the aggregation model to obtain the initial aggregation response power trajectory and initial power allocation results for each time period. In the case of a real-time electricity price response scenario, starting from the current moment, a segment of the predicted electricity purchase price for a preset decision period is selected to generate a rolling price vector. Based on the rolling price vector, with the objective of minimizing the total electricity purchase cost within a preset rolling window, the physical aggregation model is solved under the constraints of the aggregation model to obtain the initial aggregation response power trajectory and initial power allocation results for each rolling window. The results of the first period of each rolling window are then spliced ​​together to obtain the initial aggregation response power trajectory and initial power allocation results for the entire period.

[0035] Specifically, the type of electricity purchase price sequence refers to the pricing mechanism type of electricity prices, including time-of-use (TOU) pricing and real-time (RT) pricing. TOU pricing has a fixed price period (e.g., peak, flat, and valley periods), while RT pricing fluctuates in real time. TOU pricing scenarios involve prices divided into preset time periods (e.g., peak, flat, and valley periods), with fixed prices for each period. RT pricing scenarios involve prices adjusted in real time according to the supply and demand relationship in the electricity market, resulting in frequent price fluctuations. The rolling price vector, in the RT pricing scenario, is a vector formed by extracting the predicted electricity purchase price value for a preset decision-making period (e.g., 1 hour, 30 minutes) starting from the current moment. The scrolling window is the optimization time window in the real-time electricity price scenario. The window length is the preset decision cycle and it keeps scrolling forward as time goes by. For example, if the current time is t, the window is [t, t+1h], and the next time is t+15min, the window is [t+15min, t+1h+15min].

[0036] Schematic analysis of the characteristics of electricity purchase price sequences: if prices are divided into fixed time periods and remain constant across periods, it is classified as a time-of-use (TOU) pricing scenario; if prices change in real time without a fixed period, it is classified as a real-time (RTU) pricing scenario. TOU pricing scenario optimization aims to minimize the total electricity purchase cost across all time periods within a preset time domain (e.g., 24 hours), with the objective function being: ,in Let t be the fixed electricity price for time period t, and Δt be the time step (e.g., 1 hour). Substitute the set of constraints of the aggregation model (fixed load uncertainty constraints, upper and lower bound constraints of load power, ramp power constraints, energy storage charging and discharging power constraints, and energy capacity constraints) into the optimization problem, and solve it using linear programming or quadratic programming methods to obtain the initial aggregation response power trajectory for each time period. And the initial power allocation results. Real-time electricity price scenario optimization starts from the current time t, selecting a preset decision period. (e.g., 1 hour) Electricity purchase price forecast, generating a rolling price vector. The objective function is to minimize the total electricity purchase cost within the rolling window. Solving under the constraints of the aggregation model yields the initial aggregate response power trajectory and initial power allocation results for the rolling window. Only the optimization results for the first time period (time period t) of the current rolling window are retained, while the results for other time periods are discarded. The results for the first time period of each rolling window are then concatenated in chronological order to obtain the initial aggregate response power trajectory for the entire time period. And the initial power allocation results.

[0037] Full-cycle optimization in time-of-use pricing scenarios can fully leverage price differences across different time periods to maximize cost savings; rolling window optimization in real-time pricing scenarios can quickly respond to price fluctuations, improving the timeliness and accuracy of optimization decisions; the method of stitching together the results of the first time period in the rolling window balances optimization accuracy and computational efficiency, making it suitable for large-scale real-time scheduling scenarios.

[0038] S4. With the goal of minimizing the mean square error between the observed response and the simulated response, the physical aggregation model is solved by inverse optimization in two layers based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data. The optimal model parameters of the physical aggregation model are obtained, and the physical aggregation model is updated based on the optimal model parameters to obtain the final physical aggregation model. Specifically, the inverse optimization bi-level problem represents a nested optimization problem where the outer layer aims to minimize the mean square error between the observed and simulated responses, and the inner layer aims to minimize the electricity purchase cost. The corresponding equation is... and Mode The optimal model parameters are the physical parameters obtained through inverse optimization that best match the simulated response to the actual observed response. These include the upper and lower limits of adjustable load power and ramp power, and the upper and lower limits of energy storage charging and discharging power, energy capacity, and efficiency, corresponding to the parameter set. .

[0039] Schematic, we define the observed response and the simulated response, where the observed response is the measured response power data from the load data. The simulation response is the initial aggregate response power trajectory. Construct a two-level inverse optimization problem, with the inner constraint being the set of constraints for the aggregation model. The inner objective is to minimize the electricity purchase cost under the corresponding electricity price scenario, by solving for the simulated response power and simulated electricity purchase cost under the current candidate parameters θ; the outer objective is to minimize the mean square error between the observed response and the simulated response, with the objective function being... Bayesian optimization solves the problem by initializing the parameter sample set, constructing a Gaussian process surrogate model, selecting sampling points based on the expected improvement (EI) criterion, iteratively updating the training set until the mean squared error difference is less than a preset threshold, and then outputting the optimal model parameters. Finally, the optimal model parameters are substituted into the physical aggregation model to obtain the final physical aggregation model.

[0040] Preferably, with the goal of minimizing the mean square error between the observed response and the simulated response, the physical aggregation model is solved by inverse optimization in a two-level solution based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data, to obtain the optimal model parameters of the physical aggregation model, including: The aggregated response power is obtained based on the initial aggregated response power trajectory and the initial power allocation results. The aggregated response power is used as the simulation response, and the measured response power data of the period corresponding to the initial aggregated response trajectory in the load data is used as the observed response. The aggregation model constraint is used as the inner constraint of the inverse optimization bi-level problem. The goal is to minimize the electricity purchase cost under the corresponding electricity price response scenario. The simulated response power and simulated electricity purchase cost under the current candidate parameters are obtained by solving the inner optimization. The current candidate parameters include: the power of the adjustable power load, the ramp power, the energy storage charging and discharging power, and the energy capacity. The outer objective function is constructed with the goal of minimizing the mean square error between the observed response and the simulated response; Based on the simulated response power and simulated electricity purchase cost obtained from the inner layer optimization solution, Bayesian optimization is performed on the outer layer objective function until the mean square error difference is less than a preset threshold. The current candidate parameter corresponding to the minimum mean square error is then obtained as the optimal model parameter.

[0041] Specifically, the observed response refers to the actual operational response data of the flexible load, that is, the measured response power data of the period corresponding to the initial aggregate response trajectory in the load data. Simulated response refers to the response power data calculated based on the physical aggregation model and candidate parameters, i.e., the initial aggregated response power trajectory, corresponding to... The inverse optimization bi-level problem represents a nested optimization problem consisting of an outer optimization layer and an inner optimization layer. The outer optimization layer aims to minimize the mean square error between the observed response and the simulated response, while the inner optimization layer aims to minimize the electricity purchase cost. The current candidate parameters are the physical aggregation model parameters to be identified, including the upper and lower bounds of the adjustable power load, the ramp power, the upper and lower limits of the energy storage charging and discharging power, and the energy capacity, corresponding to θ. Bayesian optimization is a global optimization algorithm based on a Gaussian process (GP) surrogate model. Through iterative sampling and surrogate modeling, it efficiently explores the optimal parameters under finite sample conditions.

[0042] Schematic, the Bayesian optimization solution randomly selects parameters within the feasible region. Five initial candidate parameters are selected. The simulation response power and mean square error of each parameter group are calculated through inner-layer optimization to form the initial training dataset. Based on this initial training dataset, a Gaussian process surrogate model is constructed. The Matern kernel function is used to characterize the correlation between parameters. The signal variance α, scale length β, and noise term ϵ of the kernel function are optimized by marginal likelihood maximization. The Matern kernel function is as follows: ,in The expected improvement (EI) criterion is used to select the next candidate parameter point. The EI function is: ,in This represents the minimum objective function value of the current proxy model. , and The posterior mean and variance of the surrogate model are obtained; new candidate parameters are substituted into the inner layer optimization to obtain the simulation response power and mean square error, the training dataset is expanded, and the Gaussian process surrogate model is updated; the above steps are repeated until the difference in mean square error between two adjacent iterations is less than a preset threshold, at which point the iteration stops; the candidate parameters with the smallest mean square error in the training dataset are selected as the optimal model parameters. .

[0043] The inverse optimization two-layer structure takes into account both physical constraints and data fitting, ensuring the physical rationality and interpretability of the identification parameters; Bayesian optimization, based on the Gaussian process surrogate model, can efficiently explore the global optimum in the high-dimensional parameter space and avoid getting trapped in local optima; the expectation improvement criterion balances the exploration and utilization of the parameter space and improves the identification convergence speed.

[0044] Preferably, it further includes: performing noise robustness modeling when constructing the outer objective function; the robustness modeling includes: The fixed load forecasting error is modeled as a first Gaussian noise with a preset fixed mean and a fixed covariance matrix. The observation error of the aggregate response is modeled as a second Gaussian noise with a preset aggregate mean and a covariance of the aggregate covariance matrix. The outer objective function is corrected based on the first Gaussian noise and the second Gaussian noise; wherein the corrected outer objective function is the sum of the mean square error of the observed response and the simulation response plus the noise influence term, and the noise influence term is the sum of the traces of the covariance matrices of the first Gaussian noise and the second Gaussian noise.

[0045] Specifically, noise robustness modeling represents the process of improving the anti-interference capability of parameter identification results by modeling the prediction error and observation noise in the data; the first Gaussian noise characterizes the Gaussian noise of the fixed load prediction error, denoted as... The mean is a preset fixed mean. The covariance is a fixed covariance matrix. The multivariate Gaussian distribution, i.e. The second Gaussian noise characterizes the Gaussian noise of the aggregate response observation error, denoted as . The mean is the preset aggregate mean. The covariance is the aggregated covariance matrix. The multivariate Gaussian distribution; the noise influence term is the combined effect of the first Gaussian noise and the second Gaussian noise on parameter identification, which is the sum of the traces of their covariance matrices, i.e., tr(Σ fix )+tr(Σ agg), where tr() is the trace of the matrix (the sum of the elements on the main diagonal); the modified outer objective function incorporates the noise-affected outer objective function to improve the robustness of parameter identification, corresponding to .

[0046] The modified objective function can suppress noise interference, making the parameter identification results closer to the true values ​​and improving the reliability of the model. The introduction of the noise covariance matrix provides a basis for robustness analysis, which facilitates the assessment of the risk of parameter identification in engineering applications. It is applicable to real-world scenarios with poor data quality and high noise, thus expanding the scope of application of the method.

[0047] S5. Resolve the final physical aggregation model to obtain the final target aggregation response trajectory and target power allocation result, and perform grid scheduling based on the target aggregation response trajectory, target power allocation result and grid topology.

[0048] Specifically, the target aggregated response trajectory is a power time-series trajectory that reflects the optimal response characteristics of flexible loads, obtained by resolving the optimal model parameters; the target power allocation result is the power allocation scheme of the three types of components in each time period.

[0049] Schematic, based on the final physical aggregation model, the target aggregation response trajectory and target power allocation results are obtained through resolving the problem. Scheduling is then performed in conjunction with the power grid topology (branch capacity, node connectivity): dynamically matching active power supply and demand at each node, verifying branch power flow security, and prioritizing the use of low-cost flexible resources to ensure the safe and economical operation of the power grid. It is compatible with both static and dynamic response mechanisms, applicable to various market environments such as time-of-use pricing and real-time pricing, solving the problem of poor adaptability of existing technologies. The physical aggregation model uniformly represents three types of flexible resources, integrating multi-dimensional constraints and improving the model's integrity and physical interpretability. The inverse optimization two-layer structure combined with Bayesian optimization achieves efficient global identification of high-dimensional parameters, significantly improving parameter identification accuracy and robustness. The scheduling process combines the power grid topology and power allocation results, balancing safety and economy to ensure the stable operation of the power system.

[0050] Preferably, the target power allocation results include: the base power allocation of fixed loads, the adjustment power allocation of adjustable power loads at different times, the charging and discharging power allocation of energy storage systems, and the cooperative power complementarity allocation between distributed generation and loads; Power grid dispatching is performed based on the target aggregated response trajectory, target power allocation results, and power grid topology, including: Based on the target aggregated response trajectory and combined with the real-time operating status of the distribution network, the active power supply and demand of each node are dynamically matched. Based on the grid topology and branch capacity limitations, verify whether the branch power flow corresponding to the target power allocation result exceeds the safety threshold. If it does, adjust the adjustment range of the adjustable power load and the energy storage charging and discharging power. Based on the target power allocation results, the scheduling strategy is dynamically adjusted by prioritizing the use of flexible resources with the lowest adjustment costs, combined with the electricity purchase price sequence.

[0051] Specifically, the base power allocation for fixed loads is the operating power of the fixed load in each time period (non-adjustable), that is... The adjustable power load's power allocation for each time period is the operating power of the adjustable load in each time period (including the adjusted power), i.e. The charging and discharging power allocation of the energy storage system is the charging and discharging power of the energy storage system at each time period (charging is negative, discharging is positive), that is... The complementary power allocation between distributed generation and load is the complementary power between distributed generation output and load demand, that is, the load power offset by distributed generation in the net load.

[0052] Indicatively, the system monitors the real-time operating status (voltage, frequency, power supply and demand) of each node in the distribution network. Based on the target aggregated response trajectory, it adjusts the power output of adjustable loads and energy storage systems to dynamically match the active power supply and demand of each node, eliminating power imbalance deviations. For example, if node A has a power deficit of 50kW, it dispatches 50kW of energy storage discharge to supplement it. Based on the grid topology (node ​​connection relationships, branch parameters), it calculates the power flow of each branch corresponding to the target power allocation result. It verifies whether the branch power flow exceeds the safety threshold: if the power flow of a certain branch exceeds the threshold, such as the maximum allowable power flow of the branch being 300kW and the current power flow being 320kW, it adjusts the adjustment range of the adjustable power load, such as reducing the adjustable load within the power supply range of the branch by 20kW or the energy storage charging and discharging power, such as reducing the energy storage discharge power connected to the branch by 20kW, until the branch power flow drops below the safety threshold. Calculate the adjustment costs of various flexible resources, such as the production loss cost of adjustable loads and the charging and discharging loss cost of energy storage; based on the target power allocation results, prioritize the use of flexible resources with the lowest adjustment costs, such as prioritizing the dispatch of energy storage if the charging and discharging cost of energy storage is lower than the reduction cost of adjustable loads; dynamically adjust the dispatch strategy in conjunction with the electricity purchase price sequence, such as increasing the adjustment range of flexible resources when the electricity price rises and decreasing the adjustment range when the electricity price falls, so as to minimize the total sum of electricity purchase cost and adjustment cost over the entire dispatch cycle.

[0053] In a preferred embodiment of the present invention, taking a power distribution network in an industrial park as an example, the power distribution network includes fixed loads (data center basic loads), adjustable loads (industrial production line loads, air conditioning loads), energy storage systems (lithium battery energy storage power stations), and distributed photovoltaics. The system collects 24-hour uncontrollable load forecasts (mean 500kW, standard deviation of disturbance 20kW), the maximum adjustable capacity of the adjustable loads is 300kW, the initial energy storage capacity is 800kWh, the upper and lower limits of energy storage capacity are 200-1000kWh, and the upper and lower limits of charging and discharging power are... The system is designed with 400kW (charging) / 400kW (discharging), energy storage efficiency of 0.95, predicted photovoltaic output (peak 600kW), and time-of-use pricing (1.2 yuan / kWh during peak hours, 0.8 yuan / kWh during normal hours, and 0.4 yuan / kWh during off-peak hours). A Gaussian distribution model for the fixed load is constructed (mean 500kW, standard deviation 20kW), with adjustable load power limits of 100-400kW and a ramp rate of ±50kW / h. Under the time-of-use pricing scenario, the initial aggregated response power trajectory (off-peak energy storage charging, peak energy storage discharging + adjustable load reduction) is obtained with the objective of minimizing 24-hour electricity purchase cost. The inverse optimization problem is solved using Bayesian optimization to obtain the optimal adjustable load ramp rate of ±48kW / h and energy storage efficiency of 0.945. Based on the optimal model parameters, energy storage is scheduled to charge 300kW during off-peak hours and discharge 300kW during peak hours, while adjustable load is reduced by 200kW during peak hours, ensuring grid power balance and reducing daily electricity purchase cost by 15%.

[0054] By implementing this embodiment, multi-dimensional data such as distribution network load data, energy storage operation status data, and distributed generation capacity data are acquired, providing accurate basic inputs for the construction of the physical aggregation model. This avoids modeling bias caused by missing data and ensures the reliability of the model. A unified physical aggregation model covering fixed loads, adjustable power loads, and energy storage systems is constructed, achieving integrated characterization of different types of flexible resources. This solves the problem of difficult collaborative modeling of multiple types of loads in existing technologies and improves model compatibility. The aggregation model constraints include uncertainties for fixed loads, upper and lower bounds for load power, ramp power constraints, energy storage charging and discharging power constraints, and energy capacity constraints, comprehensively covering the core physical characteristics of loads and energy storage. This ensures that the model output conforms to actual operating patterns and improves modeling accuracy. Then, with the goal of minimizing electricity purchase costs, the electricity price scenario response is optimized under the constraints of the aggregation model. This yields an initial aggregation response power trajectory and power allocation result that meets economic operation requirements, providing a reliable benchmark for subsequent parameter identification. The optimal model parameters are then solved through a bi-level inverse optimization problem. By combining measured response power data with the mean square error between observed and simulated responses, the initial model parameter deviations are effectively corrected. The updated model better reflects actual load characteristics, resolving the insufficient adaptability issue caused by fixed parameters in traditional models. The final aggregated response trajectory and power allocation results can be directly used for grid dispatching, fully considering load characteristics, energy storage constraints, and electricity price impacts. This provides accurate and feasible decision-making basis for grid dispatching, contributing to improved safety and economy of power system operation.

[0055] See Figure 2 This is a schematic diagram of a price-response-based flexible load aggregation modeling device according to an embodiment of the present invention, comprising: The data acquisition module is used to acquire load data, energy storage operation status data, distributed generation capacity data, grid topology, electricity purchase price series, and electricity purchase price forecast of the distribution network to be optimized. The initial aggregation model construction module is used to construct a physical aggregation model to characterize fixed load, adjustable power, and energy storage aggregation based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, and to construct the aggregation model constraints corresponding to the physical aggregation model. The aggregation model constraints include: fixed load uncertainty constraints, load power upper and lower bound constraints, ramping power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. The initial model solving module is used to optimize the physical aggregation model for electricity price scenario response under the constraints of the aggregation model, based on the electricity purchase price sequence and the predicted electricity purchase price, with the goal of minimizing the electricity purchase cost. It solves for the initial aggregation response power trajectory and the initial power allocation results for each time period. The model inverse optimization module is used to minimize the mean square error between the observed response and the simulated response. It performs a two-level inverse optimization solution on the physical aggregation model based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data to obtain the optimal model parameters of the physical aggregation model. The physical aggregation model is then updated based on the optimal model parameters to obtain the final physical aggregation model. The power grid dispatch module is used to re-solve the final physical aggregation model to obtain the final target aggregation response trajectory and target power allocation result, and to perform power grid dispatch based on the target aggregation response trajectory, target power allocation result and power grid topology.

[0056] Preferably, the power grid dispatching module is used to perform power grid dispatching based on the target aggregated response trajectory, the target power allocation result, and the power grid topology, including: Based on the target aggregated response trajectory and combined with the real-time operating status of the distribution network, the active power supply and demand of each node are dynamically matched. Based on the grid topology and branch capacity limitations, verify whether the branch power flow corresponding to the target power allocation result exceeds the safety threshold. If it does, adjust the adjustment range of the adjustable power load and the energy storage charging and discharging power. Based on the target power allocation results, the scheduling strategy is dynamically adjusted by prioritizing the use of flexible resources with the lowest adjustment costs, combined with the electricity purchase price sequence.

[0057] This invention provides a price-response-based flexible load aggregation modeling device. By acquiring multi-dimensional data such as distribution network load data, energy storage operation status data, and distributed generation capacity data, it provides accurate basic inputs for the construction of the physical aggregation model, avoiding modeling deviations caused by missing data and ensuring model reliability. A unified physical aggregation model covering fixed loads, adjustable power loads, and energy storage systems is constructed, achieving integrated characterization of different types of flexible resources. This solves the problem of difficult collaborative modeling of multiple types of loads in existing technologies and improves model compatibility. The aggregation model constraints include fixed load uncertainty constraints, load power upper and lower bound constraints, ramp power constraints, energy storage charging and discharging power constraints, and energy capacity constraints, comprehensively covering the core physical characteristics of loads and energy storage, ensuring that the model output conforms to actual operating patterns and improving modeling accuracy. Then, with the goal of minimizing electricity purchase cost, the electricity price scenario response is optimized under the constraints of the aggregation model, yielding an initial aggregation response power trajectory and power allocation results that meet economic operation requirements, providing a reliable benchmark for subsequent parameter identification. The optimal model parameters are then solved through a bi-level inverse optimization problem. By combining measured response power data with the mean square error between observed and simulated responses, the initial model parameter deviations are effectively corrected. The updated model better reflects actual load characteristics, resolving the insufficient adaptability issue caused by fixed parameters in traditional models. The final aggregated response trajectory and power allocation results can be directly used for grid dispatching, fully considering load characteristics, energy storage constraints, and electricity price impacts. This provides accurate and feasible decision-making basis for grid dispatching, contributing to improved safety and economy of power system operation.

[0058] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0059] Those skilled in the art will understand that, for convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0060] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a price-response-based flexible load aggregation modeling method as described in the above embodiments. The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0061] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0062] The memory can be used to store the computer program. The processor implements various functions of the terminal device by running or executing the computer program stored in the memory and calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device or other volatile solid-state storage device.

[0063] Another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the price-response-based flexible load aggregation modeling method described in the above embodiment.

[0064] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0065] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for modeling price-responsive flexible load aggregation, comprising: include: Obtain load data, energy storage operation status data, distributed generation capacity data, grid topology, electricity purchase price series, and electricity purchase price forecast for the distribution network to be optimized; Based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, a physical aggregation model is constructed to characterize fixed load, adjustable power, and energy storage aggregation. The aggregation model constraints corresponding to the physical aggregation model are also constructed. The aggregation model constraints include: fixed load uncertainty constraints, load power upper and lower bound constraints, ramping power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. Based on the electricity purchase price series and the predicted electricity purchase price, with the goal of minimizing the electricity purchase cost, the physical aggregation model is optimized for the electricity price scenario under the constraints of the aggregation model. The initial aggregation response power trajectory and initial power allocation results for each time period are obtained by solving the problem. With the goal of minimizing the mean square error between the observed response and the simulated response, the physical aggregation model is solved by inverse optimization in two layers based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data. The optimal model parameters of the physical aggregation model are obtained, and the physical aggregation model is updated based on the optimal model parameters to obtain the final physical aggregation model. The final physical aggregation model is re-solved to obtain the final target aggregation response trajectory and target power allocation result. Power grid scheduling is then carried out based on the target aggregation response trajectory, target power allocation result, and power grid topology.

2. The price response based flexible load aggregation modeling method of claim 1, wherein, The load data also includes: uncontrollable load data and adjustable load data; the uncontrollable load data includes: uncontrollable load forecast values ​​and uncontrollable load disturbance standard deviation; the adjustable load data includes: maximum adjustable capacity, adjustable ratio, earliest start-up and shutdown time, latest start-up and shutdown time, minimum continuous operating time, and response speed; the energy storage operation status data includes: initial energy storage capacity, upper and lower limits of energy storage capacity, upper and lower limits of charge and discharge power, and energy storage efficiency; the distributed generation capacity data includes: photovoltaic output forecast values, photovoltaic output forecast errors, and curtailment of photovoltaic power. Based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, a physical aggregation model is constructed to characterize stationary loads, adjustable power, and energy storage aggregation, including: The uncontrollable load prediction values ​​are indexed and assigned using a preset discrete-time index to generate a power trajectory vector. A Gaussian distribution is calculated based on the standard deviation of the uncontrollable load disturbance and the power trajectory vector to construct a fixed load model. The upper and lower limits of adjustable power load are determined based on the maximum adjustable capacity and adjustable ratio. The upper and lower limits of ramp power are determined based on the response speed. The adjustment time range is determined based on the earliest start-stop time, the latest start-stop time, and the minimum continuous running time. An adjustable power load model is constructed based on the upper and lower limits of adjustable power load, the upper and lower limits of ramp power, and the adjustment time range. Based on the initial energy storage capacity, upper and lower limits of energy storage capacity, upper and lower limits of charge and discharge power, and energy storage efficiency, an energy storage model is constructed to characterize the relationship between energy storage power and energy state. Based on the predicted photovoltaic output, the predicted photovoltaic output error, and the amount of curtailed solar power, the net load is calculated. Then, based on the fixed load model, the adjustable power load model, the energy storage model, and the net load, a physical aggregation model is constructed to characterize the aggregation of fixed load, adjustable power, and energy storage.

3. The price response based flexible load aggregation modeling method of claim 1, wherein, Based on the electricity purchase price series and predicted electricity purchase prices, and with the goal of minimizing electricity purchase costs, the physical aggregation model is optimized for electricity price scenario response under the constraints of the aggregation model. The initial aggregation response power trajectory and initial power allocation results for each time period are obtained, including: Determine the corresponding electricity price response scenario based on the type of electricity purchase price series; In the case of time-of-use pricing for electricity price response, based on the fixed electricity price in the electricity purchase price sequence, with the goal of minimizing the total electricity purchase cost for all time periods within the preset time domain, the physical aggregation model is solved under the constraints of the aggregation model to obtain the initial aggregation response power trajectory and initial power allocation results for each time period. In the case of a real-time electricity price response scenario, starting from the current moment, a segment of the predicted electricity purchase price for a preset decision period is selected to generate a rolling price vector. Based on the rolling price vector, with the objective of minimizing the total electricity purchase cost within a preset rolling window, the physical aggregation model is solved under the constraints of the aggregation model to obtain the initial aggregation response power trajectory and initial power allocation results for each rolling window. The results of the first period of each rolling window are then spliced ​​together to obtain the initial aggregation response power trajectory and initial power allocation results for the entire period.

4. The price response based flexible load aggregation modeling method of claim 1, wherein, With the goal of minimizing the mean square error between the observed and simulated responses, the physical aggregation model is solved inversely using a two-level solution based on the initial aggregated response power trajectory, initial power allocation results, and measured response power data from the load data. This yields the optimal model parameters for the physical aggregation model, including: The aggregated response power is obtained based on the initial aggregated response power trajectory and the initial power allocation results. The aggregated response power is used as the simulation response, and the measured response power data of the period corresponding to the initial aggregated response trajectory in the load data is used as the observed response. The aggregation model constraint is used as the inner constraint of the inverse optimization bi-level problem. The goal is to minimize the electricity purchase cost under the corresponding electricity price response scenario. The simulated response power and simulated electricity purchase cost under the current candidate parameters are obtained by solving the inner optimization. The current candidate parameters include: the power of the adjustable power load, the ramp power, the energy storage charging and discharging power, and the energy capacity. The outer objective function is constructed with the goal of minimizing the mean square error between the observed response and the simulated response; Based on the simulated response power and simulated electricity purchase cost obtained from the inner layer optimization solution, Bayesian optimization is performed on the outer layer objective function until the mean square error difference is less than a preset threshold. The current candidate parameter corresponding to the minimum mean square error is then obtained as the optimal model parameter.

5. A price response based flexible load aggregation modeling method as claimed in claim 4, wherein, Also includes: When constructing the outer objective function, noise robustness modeling is performed; The robustness modeling includes: The fixed load forecasting error is modeled as a first Gaussian noise with a preset fixed mean and a fixed covariance matrix. The observation error of the aggregate response is modeled as a second Gaussian noise with a preset aggregate mean and a covariance of the aggregate covariance matrix. The outer objective function is corrected based on the first Gaussian noise and the second Gaussian noise; wherein the corrected outer objective function is the sum of the mean square error of the observed response and the simulation response plus the noise influence term, and the noise influence term is the sum of the traces of the covariance matrices of the first Gaussian noise and the second Gaussian noise.

6. The flexible load aggregation modeling method based on price response as described in claim 1, characterized in that, The target power allocation results include: the base power allocation for fixed loads, the adjustment power allocation for adjustable power loads at different times, the charging and discharging power allocation for energy storage systems, and the complementary power allocation between distributed generation and loads. Power grid dispatch is performed based on the target aggregated response trajectory, target power allocation results, and power grid topology, including: Based on the target aggregated response trajectory and combined with the real-time operating status of the distribution network, the active power supply and demand of each node are dynamically matched. Based on the grid topology and branch capacity limitations, verify whether the branch power flow corresponding to the target power allocation result exceeds the safety threshold. If it does, adjust the adjustment range of the adjustable power load and the energy storage charging and discharging power. Based on the target power allocation results, the scheduling strategy is dynamically adjusted by prioritizing the use of flexible resources with the lowest adjustment costs, combined with the electricity purchase price sequence.

7. A flexible load aggregation modeling device based on price response, characterized in that, include: The data acquisition module is used to acquire load data, energy storage operation status data, distributed generation capacity data, grid topology, electricity purchase price series, and electricity purchase price forecast of the distribution network to be optimized. The initial aggregation model construction module is used to construct a physical aggregation model to characterize fixed load, adjustable power, and energy storage aggregation based on load data, energy storage operation status data, distributed generation capacity data, and grid topology data, and to construct the aggregation model constraints corresponding to the physical aggregation model. The aggregation model constraints include: fixed load uncertainty constraints, load power upper and lower bound constraints, ramping power constraints, energy storage charging and discharging power constraints, and energy capacity constraints. The initial model solving module is used to optimize the physical aggregation model for electricity price scenario response under the constraints of the aggregation model, based on the electricity purchase price sequence and the predicted electricity purchase price, with the goal of minimizing the electricity purchase cost. It solves for the initial aggregation response power trajectory and the initial power allocation results for each time period. The model inverse optimization module is used to minimize the mean square error between the observed response and the simulated response. It performs a two-level inverse optimization solution on the physical aggregation model based on the initial aggregated response power trajectory, the initial power allocation results, and the measured response power data in the load data to obtain the optimal model parameters of the physical aggregation model. The physical aggregation model is then updated based on the optimal model parameters to obtain the final physical aggregation model. The power grid dispatch module is used to re-solve the final physical aggregation model to obtain the final target aggregation response trajectory and target power allocation result, and to perform power grid dispatch based on the target aggregation response trajectory, target power allocation result and power grid topology.

8. The flexible load aggregation modeling device based on price response as described in claim 7, characterized in that, The power grid dispatching module is used to perform power grid dispatching based on the target aggregated response trajectory, target power allocation results, and power grid topology, including: Based on the target aggregated response trajectory and combined with the real-time operating status of the distribution network, the active power supply and demand of each node are dynamically matched. Based on the grid topology and branch capacity limitations, verify whether the branch power flow corresponding to the target power allocation result exceeds the safety threshold. If it does, adjust the adjustment range of the adjustable power load and the energy storage charging and discharging power. Based on the target power allocation results, the scheduling strategy is dynamically adjusted by prioritizing the use of flexible resources with the lowest adjustment costs, combined with the electricity purchase price sequence.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements a price-response-based flexible load aggregation modeling method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform a flexible load aggregation modeling method based on any one of claims 1 to 6.

Images