Virtual Power Plant Regulation Method and System Based on Dynamic Parameter Identification and Hierarchical Collaboration

Abstract

The present invention discloses a virtual power plant control method and system based on dynamic parameter identification and hierarchical collaboration, which belongs to the technical field of dynamic control of virtual power plants. The method and system include: collecting multidimensional data of building distributed resources, wherein the building distributed resources include rooftop photovoltaic equipment, EV charging piles and energy storage equipment, and integrating their multidimensional data through federated learning; aggregating the building distributed resources according to their power generation characteristics and power generation economy, generating virtual power plant external characteristic parameters and constructing a virtual power plant energy and operation reserve optimization model; online updating of the power generation parameters of the model through the RLS algorithm optimization model, considering the uncertainty of the virtual power plant operation, constructing a hierarchical timing collaboration framework to dynamically adjust the virtual power plant energy and operation reserve optimization model, generating control instructions and sending them to each building distributed resource equipment to control the operating status and power output of the equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of dynamic control of virtual power plants, and more specifically, relates to a virtual power plant control method and system based on dynamic parameter identification and hierarchical collaboration. Background Art

[0002] With the widespread adoption of distributed energy resources, such as rooftop photovoltaic systems, EV charging stations, and energy storage systems deployed in buildings, traditional power system regulation faces numerous challenges. Virtual power plants (VPPs), as a means of effectively integrating distributed energy resources, have garnered widespread attention.

[0003] However, existing technologies have certain shortcomings in the regulation and control of virtual power plants. On the one hand, traditional methods often find it difficult to effectively integrate the multi-dimensional data processing of distributed building resources while ensuring data privacy. The data format, acquisition frequency and data volume of different types of equipment vary greatly. If a centralized data processing method is adopted, it will involve a large amount of data transmission and storage, which not only increases costs but also poses a risk of data leakage. Existing technologies lack effective distributed data integration methods. On the other hand, in the process of virtual power plant resource aggregation, in the past, only a single factor was considered, such as aggregation based solely on power generation characteristics or power generation economics, without comprehensive consideration. As a result, the generated virtual power plant external characteristic parameters are not accurate enough, and the constructed optimization model is difficult to adapt to the complex and changeable actual operating conditions.

[0004] Furthermore, virtual power plant operations are plagued by numerous uncertainties, such as fluctuations in photovoltaic power generation due to changes in sunlight intensity and the randomness of user charging demands. Existing control methods, mostly based on static models, lack effective mechanisms for addressing these uncertainties. They are unable to dynamically adjust optimization models based on real-time changes in power generation parameters, resulting in significant deviations between control instructions and actual operating conditions, making it difficult to achieve precise control of equipment operating status and power output. Furthermore, existing technologies do not fully utilize feedback information during the control process, failing to make timely and effective adjustments to optimization models based on actual operating data, thus impacting the overall operational efficiency and stability of the virtual power plant. Summary of the Invention

[0005] In order to solve the deficiencies in the prior art, the present invention provides a virtual power plant control method and system based on dynamic parameter identification and hierarchical collaboration.

[0006] The present invention adopts the following technical solutions.

[0007] A first aspect of the present invention provides a virtual power plant control method based on dynamic parameter identification and hierarchical collaboration, comprising the following steps:

[0008] Collect multi-dimensional data on distributed building resources, including rooftop photovoltaic equipment, EV charging stations, and energy storage equipment, and integrate their multi-dimensional data through federated learning;

[0009] Aggregate building distributed resources based on their power generation characteristics and power generation economics to generate virtual power plant external characteristic parameters and build a virtual power plant energy and operation reserve optimization model. Use the RLS algorithm to update the power generation parameters of the virtual power plant energy and operation reserve optimization model online. Considering the uncertainty of virtual power plant operation, build a hierarchical time series coordination framework, and dynamically adjust the virtual power plant energy and operation reserve optimization model based on the real-time power generation parameters updated by the RLS algorithm.

[0010] Based on the dynamically adjusted virtual power plant energy and operating reserve optimization model within the hierarchical timing coordination framework, control instructions are generated and sent to the distributed resource equipment in each building to control the operating status and power output of the equipment. The operating status of the virtual power plant is monitored in real time, and the virtual power plant energy and operating reserve optimization model is adjusted based on feedback from actual operating data.

[0011] Optionally, aggregating building distributed resources according to their power generation characteristics and power generation economics, generating virtual power plant external characteristic parameters, and constructing a virtual power plant energy and operation reserve optimization model includes:

[0012] Each device is considered as a separate cluster, including rooftop photovoltaic devices, EV charging piles and energy storage devices;

[0013] Calculate the degree of consistency between the power generation characteristics of different devices based on the power generation characteristic curve of each device;

[0014] Aggregate devices based on the degree of compatibility between the power generation characteristics of each pair of devices to generate external characteristic parameters of the virtual power plant;

[0015] Based on the equipment aggregation results and the external characteristic parameters of the virtual power plant, a virtual power plant energy and operation reserve optimization model is constructed, including a baseline loss function, power balance constraints, energy storage system constraints, EV constraints, photovoltaic output constraints, and grid interaction power constraints. The baseline loss function is:

[0016]

[0017] in, is the grid interaction cost, For the The operating cost of an EV charging station, The penalty cost for abandoning light.

[0018] Optionally, the external characteristic parameters of the virtual power plant include rated power, economic power lower limit, economic power upper limit, unit control error dead zone, upward climbing rate, downward climbing rate, secondary frequency regulation upward limit, secondary frequency regulation downward limit, minimum time interval between unit shutdown and startup, and cost-capacity function.

[0019] Optionally, the online updating of power generation parameters of the virtual power plant energy and operating reserve optimization model by using the RLS algorithm includes:

[0020] Based on multi-dimensional data related to photovoltaic equipment, the RLS algorithm is used to dynamically correct the photovoltaic output prediction model and update the photovoltaic inverter efficiency online;

[0021] Based on the multi-dimensional data related to EV charging piles and energy storage devices, the RLS algorithm is used to update the energy storage SOC constraint and EV frequency modulation time constant online.

[0022] Optionally, considering the uncertainty of virtual power plant operation and building a hierarchical timing coordination framework includes:

[0023] The hierarchical time series coordination framework includes a day-ahead layer, an intraday layer, and a real-time layer;

[0024] At the day-ahead level, PV-EV joint scenarios are generated based on PV output, EV behavior, and historical grid and market data, along with PV output forecast models. High-risk scenarios that meet preset risk thresholds are selected and injected into the optimization model. The economic losses of these high-risk scenarios are quantified using conditional value-at-risk constraints, and a benchmark dispatch plan is developed.

[0025] In the intra-day layer, based on the periodically collected PV output forecast data, the dispatch plan is dynamically adjusted through rolling optimization, and the PV output forecast error distribution parameters are updated through Bayesian online learning.

[0026] In the real-time layer, based on the second-updated photovoltaic output measured data and the grid frequency regulation demand signal, event-triggered Bayesian online learning is used to correct the photovoltaic output forecast error distribution parameters in real time, dynamically adjust the backup capacity call strategy, and optimize the charging and discharging priorities of charging piles and energy storage equipment according to time-of-use electricity prices and real-time load demand.

[0027] Optionally, the day-ahead layer includes:

[0028] Based on the historical data of PV output, EV behavior, grid and market, and the PV output prediction model, a PV-EV joint scenario is generated. High-risk scenarios that meet the preset risk threshold are selected and injected into the optimization model. Cluster analysis is performed on the historical PV output data and EV behavior data. PV output and EV available capacity are used as feature vectors to generate Combined PV-EV scenarios;

[0029] Calculate the net load deviation of each scenario and filter out high-risk scenarios that meet the preset risk threshold;

[0030] The conditional risk value constraint is introduced into the objective function of the virtual power plant energy and operating reserve optimization model to minimize the conditional risk value. The conditional risk value constraint formula is:

[0031]

[0032] Add the conditional value-at-risk constraint to the benchmark loss function to generate a benchmark scheduling plan:

[0033]

[0034] in, is the quantile threshold; is the confidence level; is the expectation operator; is the baseline loss function.

[0035] Optionally, the day inner layer comprises:

[0036] Based on the periodically collected photovoltaic output forecast data, the scheduling plan is dynamically adjusted through rolling optimization. The rolling optimization time window is ;

[0037] Add a real-time correction term based on the baseline objective function:

[0038]

[0039] in, To adjust the cost factor; is the penalty item for scheduling plan adjustment;

[0040] The photovoltaic output prediction error distribution parameters are updated through Bayesian online learning.

[0041] Optionally, the real-time layer includes:

[0042] Based on second-by-second PV output measurement data and grid frequency regulation demand signals, event-triggered Bayesian online learning is used to correct PV output forecast error distribution parameters in real time and dynamically adjust the reserve capacity deployment strategy, including:

[0043] Determine the grid frequency deviation based on the grid frequency regulation requirements and use it to calculate the standby power requirements:

[0044]

[0045] in, It is the frequency deviation of the power grid; For standby power requirements; For frequency modulation needs;

[0046] The calculation formula for the spare capacity allocation constraint is:

[0047]

[0048] in, For the The standby discharge power of each building charging pile; It is the standby discharge power of the energy storage device.

[0049] Optionally, optimizing the charging and discharging priorities of building charging piles and energy storage devices based on time-of-use electricity prices and real-time load demands includes:

[0050] Receive time-of-use electricity price signals. During peak hours, building charging piles and energy storage equipment are given priority for discharge, while during off-peak hours, building charging piles and energy storage equipment are given priority for charging.

[0051] Monitor real-time load demand and forecasted load demand. If the real-time load demand is greater than the forecasted load demand, call on energy storage equipment to discharge and reduce electricity purchase demand. If the real-time load demand is less than the forecasted load demand, give priority to absorbing surplus photovoltaic power and reduce abandoned light.

[0052] A second aspect of the present invention provides a virtual power plant control system based on dynamic parameter identification and hierarchical collaboration. Based on the virtual power plant control method based on dynamic parameter identification and hierarchical collaboration described in the first aspect of the present invention, the system includes:

[0053] The data collection and integration module is used to collect multi-dimensional data of building distributed resources and integrate them through federated learning; the resource aggregation and modeling module is used to aggregate building distributed resources according to power generation characteristics and economic efficiency, generate external characteristic parameters of virtual power plants and build energy and operation reserve optimization models; the parameter update and dynamic adjustment module is used to update power generation parameters online through the RLS algorithm, and dynamically adjust the optimization model in combination with the hierarchical timing collaborative framework; the optimization control and status monitoring module is used to generate control instructions and send them to building distributed resource equipment to control their operating status and power output, while monitoring the operating status of the virtual power plant in real time and adjusting the optimization model according to feedback information.

[0054] Compared with the prior art, the beneficial effects of the present invention include at least:

[0055] (1) Through federated learning technology, a balance is achieved between privacy protection and efficient data integration when collecting multidimensional data of distributed building resources. The fusion of multidimensional data provides comprehensive input for subsequent optimization, solving the model bias problem caused by the single data dimension of traditional methods and improving the refinement and scenario adaptability of scheduling strategies.

[0056] (2) The dynamic resource aggregation mechanism based on power generation characteristics and economics can flexibly adapt to the temporal and spatial heterogeneity of distributed resources, and improve resource utilization by generating virtual power plant external characteristic parameters to match grid dispatch requirements;

[0057] (3) The recursive least squares method is introduced to identify and update the power generation parameters online, which can correct the model errors caused by equipment aging and environmental fluctuations in real time;

[0058] (4) Combined with the hierarchical optimization of the hierarchical timing coordination framework, it can not only cover medium- and long-term risk predictions, but also quickly respond to the second-level grid frequency regulation needs. This dynamic adjustment mechanism can effectively improve the adaptability of virtual power plants to uncertainties. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of a method provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0061] In order to more clearly introduce the outstanding essential features of the present invention and the significant progress it brings to the prior art, an application example of implementing the present invention is introduced below.

[0062] The following describes an embodiment of the present invention in detail with reference to the accompanying drawings. The application example specifically includes:

[0063] In embodiment 1, the present invention provides a virtual power plant control method based on dynamic parameter identification and hierarchical collaboration, such as Figure 1 As shown, the following steps are included:

[0064] Step 1: Collect multi-dimensional data of distributed building resources and integrate the multi-dimensional data through federated learning;

[0065] Preferably, in step 1, the building distributed resources include rooftop photovoltaic equipment, EV charging piles and energy storage equipment;

[0066] Preferably, in step 1, the multi-dimensional data of distributed building resources includes:

[0067] The V2G status of EV charging stations, user travel patterns, rooftop PV output curve, energy storage SOC, and building cooling / heating load data.

[0068] Preferably, in step 1, integrating multidimensional data through federated learning includes:

[0069] Computing nodes are deployed locally in each building, and the EV user behavior model is trained using the collected EV charging pile and user travel data. After training, only the EV user behavior model parameters are uploaded to the VPP central node to avoid leakage of original data and achieve user privacy protection.

[0070] Step 2: Aggregate the distributed resources of buildings according to their power generation characteristics and power generation economics, generate the external characteristic parameters of the virtual power plant and build the energy and operation reserve optimization model of the virtual power plant;

[0071] Preferably, in step 2, aggregating the building distributed resources according to their power generation characteristics and power generation economics includes:

[0072] In step 2.1, each device is considered as a separate cluster, including rooftop photovoltaic devices, EV charging piles, and energy storage devices.

[0073] Step 2.2, calculating the degree of consistency of power generation characteristics between different devices based on the power generation characteristic curve of each device;

[0074] Further preferably, the step 2.2 includes:

[0075] The degree of compatibility of power generation characteristics between different devices is calculated using the following formula:

[0076]

[0077] Where, For equipment and The degree of fit between For equipment Adjustment period point in the power generation characteristic curve The regulatory potential of For equipment Adjustment period point in the power generation characteristic curve The regulatory potential of is the number of adjustment cycle points used for comparison; this formula can be used to simply compare the degree of fit between two power generation characteristic curves. The higher the degree of fit, the better. The smaller.

[0078] Step 2.3, clustering the devices based on the degree of compatibility of power generation characteristics between each two devices;

[0079] Further preferably, the step 2.3 includes:

[0080] According to the agglomerative hierarchical clustering method, the devices are clustered, and the objective function is as follows:

[0081]

[0082] in, yes The number of all possible combinations of any two numbers; is the number of resources in the group after aggregation; For load equipment The load regulation potential, The total planned adjustment amount.

[0083] It is worth noting that the technical effects achieved by agglomerative hierarchical clustering using the above objective function include:

[0084] 1) The virtual power plant formed after aggregation can meet the expected regulation volume of the power grid company;

[0085] 2) The characteristic parameters of the aggregated resource equipment group are similar, which facilitates centralized control;

[0086] 3) Number of load devices that can be aggregated into a virtual power plant As small as possible, that is, the number of devices called is small to reduce the impact on users.

[0087] Preferably, in step 2, the external characteristic parameters of the virtual power plant include:

[0088] Rated power, economic power lower limit, economic power upper limit, unit control error deadband, upward ramp rate, downward ramp rate, secondary frequency regulation (AGC) upward limit, secondary frequency regulation (AGC) downward limit, minimum time interval between unit shutdown and startup, and cost-capacity function.

[0089] Step 2.4: Build a virtual power plant energy and operating reserve optimization model, including:

[0090] Baseline loss function:

[0091]

[0092] in, is the grid interaction cost, which represents the cost of purchasing or selling electricity between the building and the grid:

[0093]

[0094] in, is the power interaction value between the building and the grid, where a positive value indicates power purchase and a negative value indicates power sale; It is a time-of-use electricity price. Billing, when selling electricity Settlement, usually ;

[0095] Running costs for EV charging stations:

[0096]

[0097] in, Charging stations for EVs At the moment The charge and discharge power, discharge is negative; 、 and The device characteristic parameters of the EV charging pile are obtained through experiments or historical data fitting.

[0098] The penalty cost for abandoning light is:

[0099]

[0100] in, The penalty coefficient for curtailed solar power is RMB / kW and is set according to market rules. The discarded optical power.

[0101] Constraints:

[0102] (1) Power balance constraints

[0103]

[0104] The sum of grid power purchases, PV output, EV charging and discharging power, and energy storage device charging and discharging power must equal the building load demand;

[0105] (2) Energy storage system constraints

[0106] SOC dynamic equation:

[0107]

[0108] in, yes The state of charge of the energy storage device at any moment, indicating the proportion of the current remaining power of the energy storage battery to the total capacity; yes The state of charge of the energy stored at all times; is the charging efficiency; is the discharge efficiency; yes Charging power at the moment; yes Discharge power at the moment; is the rated capacity of energy storage; is the time interval;

[0109] SOC safety range:

[0110]

[0111] Charge and discharge power limit:

[0112]

[0113] (3) EV constraints

[0114] SOC dynamic equation:

[0115]

[0116] in, yes The state of charge of the electric vehicle at any moment, indicating the proportion of the current remaining power of the energy storage battery to the total capacity; yes The state of charge of the electric vehicle at all times; is the charging efficiency of the electric vehicle; is the discharge efficiency of the electric vehicle; yes The charging power of the electric vehicle at the moment; yes The discharge power of the electric vehicle at the moment; is the battery capacity of the electric vehicle; is the time interval;

[0117] SOC safety range:

[0118]

[0119] Charge and discharge power limit:

[0120]

[0121]

[0122] in, is the upper limit of the charging power of the EV charging pile, The upper limit of the discharge power of the EV charging station;

[0123] User travel needs:

[0124]

[0125] in, is the off-grid time of electric vehicles, The minimum off-grid SOC required by users.

[0126] (4) Photovoltaic output constraints

[0127]

[0128]

[0129] in, is the maximum theoretical photovoltaic output (kW), calculated from irradiance and temperature.

[0130] (5) Grid interaction power constraints

[0131]

[0132] in, is the maximum electricity sales power, The maximum power purchase.

[0133] Step 3: Update the power generation parameters of the virtual power plant energy and operation reserve optimization model constructed in step 2 online through RLS algorithm parameter identification;

[0134] Preferably, the step 3 includes:

[0135] Step 3.1: Based on the multi-dimensional data related to the photovoltaic equipment collected in step 1, the photovoltaic output model is dynamically corrected using the recursive least squares method (RLS). The photovoltaic output model is expressed as:

[0136]

[0137] in, is the photovoltaic power output, To predict photovoltaic power, is the overflow photovoltaic power.

[0138] Further preferably, dynamically correcting the photovoltaic output model using recursive least squares (RLS) method includes:

[0139] Initialize parameter estimates and the initial covariance matrix ;

[0140] At each time step , get new measurement data (actual PV output) and input data (such as solar irradiance, temperature and other variables that affect photovoltaic output), calculate the gain matrix according to the RLS formula:

[0141]

[0142] in, It is The gain matrix of time steps is used to adjust the amplitude of parameter updates; It is The covariance matrix of the time steps reflects the uncertainty of the parameter estimates; It is The input data vector of time steps contains various factors that affect photovoltaic output; is the forgetting factor, and its value range is , used to adjust the algorithm's forgetting speed of historical data;

[0143] Update the parameter estimates and the covariance matrix, where the updated parameter estimates are:

[0144]

[0145] in, It is The parameter estimates after the update of time steps; It is parameter estimates for time steps;

[0146] The updated covariance matrix is:

[0147]

[0148] in, It is The updated covariance matrix after time steps; is the identity matrix.

[0149] It should be noted that, considering factors such as dust obstruction and weather changes, there may be deviations between the actual photovoltaic output and the predicted value. The present invention dynamically corrects the photovoltaic output model through the recursive least squares method, so that the predicted value of the photovoltaic output model is closer to the actual value, thereby improving the accuracy of the model and providing a more reliable basis for further updating the power generation parameters such as the photovoltaic inverter efficiency in step 3.

[0150] In step 3.2, based on the data related to the EV charging pile and energy storage equipment collected in step 1, the energy storage SOC constraint and EV frequency modulation time constant are updated online using the RLS algorithm.

[0151] It is important to note that the PV output model RLS update trigger conditions include prediction error triggering and environmental mutation triggering. The prediction error threshold can be set to 5%. For example, the environmental mutation trigger includes an irradiance mutation rate exceeding a threshold. The energy storage SOC constraint RLS update trigger conditions include SOC trajectory deviation triggering and charge / discharge rate anomaly triggering. For example, the SOC trajectory deviation trigger includes an absolute error of more than 3% for the actual SOC to deviate from the predicted trajectory. The EV frequency modulation time constant RLS update trigger conditions include response delay triggering and power fluctuation triggering. For example, the response delay trigger includes an absolute error of more than 0.2s between the actual frequency modulation response time and the model value. The RLS algorithm prioritizes online updates of the generation parameters of the virtual power plant energy and operating reserve optimization model in the order of PV updates > energy storage updates > EV updates.

[0152] Step 4: Considering the uncertainty of virtual power plant operation, a hierarchical time-series collaborative framework is constructed to dynamically adjust the virtual power plant energy and operation reserve optimization model;

[0153] Preferably, the layered time series coordination framework includes a day-ahead layer, an intra-day layer, and a real-time layer;

[0154] Further preferably, in the day-ahead layer, PV-EV joint scenarios are generated based on relevant historical data and forecast models, extreme high-risk scenarios are screened for injection into the optimization model, and extreme risks are quantified through conditional value-at-risk (CVaR) constraints to formulate a benchmark dispatch plan;

[0155] Specifically, the historical data includes PV output, EV behavior, and historical data of the power grid and market. More specifically, the power grid and market data includes historical records of time-of-use electricity prices, load demand data, and extreme event records. The generation of PV-EV joint scenarios based on historical data and prediction models and the screening of extreme high-risk scenarios include:

[0156] The improved K-means algorithm is used to perform cluster analysis on the historical photovoltaic output data and EV behavior data collected in step 1, and the photovoltaic output and EV available capacity are used as feature vectors to generate Combined PV-EV scenario:

[0157]

[0158] in, for Photovoltaic output at all times, for The available EV capacity at the time;

[0159] Calculate the net load deviation of each scenario and filter out high-risk scenarios using the following formula:

[0160]

[0161] in, It is The scene in The net load deviation at the moment is used to measure the imbalance between power supply and demand in this scenario; for Load demand at the moment; select the one with the largest deviation As an extremely high-risk scenario, 10% of the scenarios can be injected into the optimization model. value.

[0162] Specifically, the process of quantifying extreme risks through conditional value at risk (CVaR) constraints and formulating a benchmark scheduling plan includes:

[0163] The conditional value at risk (CVaR) constraint is introduced into the objective function of the virtual power plant energy and operating reserve optimization model to minimize the conditional value at risk. The CvaR constraint formula is:

[0164]

[0165] in, is the quantile threshold, representing the value at risk (CvaR); is the confidence level, ranging from 0 to 1, and illustratively, the present invention selects a confidence level of 95%; is the expectation operator; is the baseline loss function;

[0166] Add the CvaR constraint to the baseline loss function to generate a baseline scheduling plan:

[0167]

[0168] Further preferably, in the intra-day layer, the scheduling plan is dynamically adjusted through rolling optimization based on more recent forecast data;

[0169] Specifically, the forecast data is updated every 15 minutes, and the scheduling plan is dynamically modified to cope with real-time fluctuations. The formula for the rolling optimization time window is:

[0170]

[0171] in, The rolling window length is set to 4 hours by way of example; is the current time point;

[0172] Based on the day-ahead objective function, a real-time correction term is added:

[0173]

[0174] in, To adjust the cost coefficient, the unit is yuan / kW; It is the penalty item for scheduling plan adjustment.

[0175] Specifically, the intra-day layer also includes updating the prediction error distribution parameters through Bayesian online learning based on the photovoltaic output prediction error data of the past few hours. The photovoltaic output prediction error calculation formula is:

[0176]

[0177] in, is the prediction error, for The photovoltaic output at the moment; the prediction error follows the Gaussian distribution ;

[0178] The updating of the prediction error distribution parameters by Bayesian learning includes:

[0179]

[0180]

[0181] in, and are the mean and variance of the historical error distribution (prior), and are the mean and variance (likelihood) of the real-time error in the current time window, and is the updated posterior distribution parameter.

[0182] Further preferably, in the real-time layer, the photovoltaic forecast error distribution is corrected in real time through high-frequency Bayesian learning, and the reserve capacity call strategy is dynamically adjusted in combination with the grid frequency regulation demand and the real-time error distribution. The charging and discharging priority of the charging piles and energy storage equipment is optimized according to the time-of-use electricity price and the real-time load demand.

[0183] Specifically, real-time correction of photovoltaic prediction error distribution through high-frequency Bayesian learning includes:

[0184] Bayesian learning is performed using instantaneous error data from an extremely short time window of less than 1 minute to obtain real-time error distribution and perform real-time corrections on photovoltaic forecasts.

[0185] It's worth noting that both the real-time and intraday layers use Bayesian online learning techniques to correct the distribution of PV output forecast errors. The differences lie in their objectives, data time windows, update frequency, and parameter adjustment logic. The goal of Bayesian learning in the intraday layer is to periodically correct the forecast model and optimize subsequent rolling scheduling plans. Its data time window uses accumulated error data from the past few hours (e.g., four hours) to reflect forecast deviation trends over longer time periods. Its update frequency is low, synchronized with the rolling optimization cycle. Its parameter adjustment logic involves adjusting the prior distribution based on the posterior distribution parameters from the previous rolling optimization, and the likelihood function based on the statistics of the error data within the current time window. The goal of Bayesian learning in the real-time layer is to respond instantaneously to real-time fluctuations. Its data time window uses instantaneous error data from an extremely short time window. Its update frequency is high (every second to every minute), synchronized with the real-time control cycle. Its parameter adjustment logic involves adjusting the prior distribution based on the posterior distribution parameters from the previous real-time update, and the likelihood function based on the statistics of the latest instantaneous error.

[0186] In a further preferred embodiment, the parameters updated in the real-time layer can be used as a priori input for the optimization of the next cycle of the intra-day layer, forming a data closed loop.

[0187] Specifically, based on the grid frequency regulation requirements and real-time error distribution, the dynamic adjustment of the reserve capacity call strategy includes:

[0188] Determine the grid frequency deviation based on the grid frequency regulation requirements and use it to calculate the standby power requirements:

[0189]

[0190] in, is the grid frequency deviation, in Hz; is the reserve power demand, in MW; Frequency regulation demand, determined according to grid rules, unit: MW / Hz;

[0191] The calculation formula for the spare capacity allocation constraint is:

[0192]

[0193] in, For the The standby discharge power of each building charging pile, in kW; It is the standby discharge power of the energy storage device.

[0194] Specifically, the optimization of charging and discharging priorities of building charging piles and energy storage equipment based on time-of-use electricity prices and real-time load demand includes:

[0195] (1) Electricity price triggering rules:

[0196] peak hours:

[0197] Building charging pile priority: discharge > charging;

[0198] Energy storage equipment priority: discharge > charge;

[0199] Low period:

[0200] Building charging pile priority: charging > discharging;

[0201] Energy storage equipment priority: charging > discharging.

[0202] (2) Load demand response rules:

[0203] If the real-time load :

[0204] Call on energy storage equipment to discharge and reduce the demand for electricity purchase;

[0205] If the real-time load :

[0206] Give priority to absorbing surplus photovoltaic power and reduce wasted light.

[0207] It is worth noting that in response to the uncertainty problem existing in the operation of virtual power plants in the existing technology, the present invention proposes a hierarchical timing coordination framework. Through the hierarchical optimization strategy, the virtual power plant achieves the optimal balance of economy, reliability and flexibility in an uncertain environment, providing core control capabilities for the new power system.

[0208] Specifically, the day-ahead layer in the layered time series collaborative framework is used for long-term risk perception and benchmark plan formulation, the intraday layer is used for short-term rolling correction and prediction accuracy improvement, and the real-time layer is used for instantaneous response and dynamic resource allocation; the error parameters of the real-time layer are fed back to the intraday layer, and the intraday layer correction model is transmitted to the day-ahead layer, forming a global optimization closed loop and realizing data closed-loop feedback, from day-ahead risk perception to real-time second-level response, covering the entire "prediction-optimization-control" process of the virtual power plant.

[0209] Step 5: Based on the virtual power plant energy and operating reserve optimization model dynamically adjusted according to the hierarchical timing coordination framework, control instructions are generated and sent to the distributed resource equipment in each building to control the operating status and power output of the equipment, and the operating status of the virtual power plant is monitored in real time. The virtual power plant energy and operating reserve optimization model is adjusted according to the feedback information of the actual operating data.

[0210] In embodiment 2, the present invention provides a virtual power plant control system based on dynamic parameter identification and hierarchical collaboration. Based on the virtual power plant control method based on dynamic parameter identification and hierarchical collaboration provided in embodiment 1, the system includes:

[0211] The data collection and integration module is used to collect multi-dimensional data of building distributed resources and integrate them through federated learning; the resource aggregation and modeling module is used to aggregate building distributed resources according to power generation characteristics and economic efficiency, generate external characteristic parameters of virtual power plants and build energy and operation reserve optimization models; the parameter update and dynamic adjustment module is used to update power generation parameters online through the RLS algorithm, and dynamically adjust the optimization model in combination with the hierarchical timing collaborative framework; the optimization control and status monitoring module is used to generate control instructions and send them to building distributed resource equipment to control their operating status and power output, while monitoring the operating status of the virtual power plant in real time and adjusting the optimization model according to feedback information.

[0212] The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0213] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A virtual power plant control method based on dynamic parameter identification and hierarchical collaboration, characterized in that: The steps include: Collect multi-dimensional data on distributed building resources, including rooftop photovoltaic equipment, EV charging stations, and energy storage equipment, and integrate their multi-dimensional data through federated learning; Aggregate building distributed resources based on their power generation characteristics to generate virtual power plant external characteristic parameters and build a virtual power plant energy and operation reserve optimization model. Use the RLS algorithm to update the power generation parameters of the virtual power plant energy and operation reserve optimization model online. Considering the uncertainty of virtual power plant operation, build a hierarchical time series coordination framework, and dynamically adjust the virtual power plant energy and operation reserve optimization model based on the real-time power generation parameters updated by the RLS algorithm. The hierarchical time series coordination framework includes a day-ahead layer, an intraday layer, and a real-time layer; At the day-ahead level, PV-EV joint scenarios are generated based on PV output, EV behavior, historical grid and market data, and PV output forecast models. High-risk scenarios that meet preset risk thresholds are screened out, and their economic losses are quantified using conditional value-at-risk constraints and injected into the optimization model to develop a benchmark dispatch plan. In the intra-day layer, based on the periodically collected PV output forecast data, the dispatch plan is dynamically adjusted through rolling optimization, and the PV output forecast error distribution parameters are updated through Bayesian online learning. In the real-time layer, based on second-by-second PV output measurement data and grid frequency regulation demand signals, event-triggered Bayesian online learning is used to correct the PV output forecast error distribution parameters in real time, dynamically adjust the reserve capacity deployment strategy, and optimize the charging and discharging priorities of charging piles and energy storage devices based on time-of-use electricity prices and real-time load demand. Based on the dynamically adjusted virtual power plant energy and operating reserve optimization model within the hierarchical timing coordination framework, control instructions are generated and sent to the distributed resource equipment in each building to control the operating status and power output of the equipment. The operating status of the virtual power plant is monitored in real time, and the virtual power plant energy and operating reserve optimization model is adjusted based on feedback from actual operating data.

2. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 1 is characterized by: Aggregating building distributed resources according to their power generation characteristics, generating virtual power plant external characteristic parameters, and constructing a virtual power plant energy and operation reserve optimization model includes: Each device is considered as a separate cluster, including rooftop photovoltaic devices, EV charging piles and energy storage devices; Calculate the degree of consistency between the power generation characteristics of different devices based on the power generation characteristic curve of each device; Aggregate devices based on the degree of compatibility between the power generation characteristics of each pair of devices to generate external characteristic parameters of the virtual power plant; Based on the equipment aggregation results and the external characteristic parameters of the virtual power plant, a virtual power plant energy and operation reserve optimization model is constructed, including a baseline loss function, power balance constraints, energy storage system constraints, EV constraints, photovoltaic output constraints, and grid interaction power constraints. The baseline loss function is: in, is the grid interaction cost, For the The operating cost of an EV charging station, The penalty cost for abandoning light.

3. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 2 is characterized by: The external characteristic parameters of the virtual power plant include rated power, economic power lower limit, economic power upper limit, unit control error dead zone, upward climbing rate, downward climbing rate, secondary frequency regulation upward limit, secondary frequency regulation downward limit, minimum time interval between unit shutdown and startup, and cost-capacity function.

4. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 3 is characterized by: The online updating of the power generation parameters of the virtual power plant energy and operation reserve optimization model by the RLS algorithm includes: Based on multi-dimensional data related to photovoltaic equipment, the RLS algorithm is used to dynamically correct the photovoltaic output prediction model and update the photovoltaic inverter efficiency online; Based on the multi-dimensional data related to EV charging piles and energy storage devices, the RLS algorithm is used to update the energy storage SOC constraint and EV frequency modulation time constant online.

5. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 4 is characterized by: The day-ahead layer includes: Based on the historical data of PV output, EV behavior, grid and market, and the PV output prediction model, a PV-EV joint scenario is generated. High-risk scenarios that meet the preset risk threshold are selected and injected into the optimization model. Cluster analysis is performed on the historical PV output data and EV behavior data. PV output and EV available capacity are used as feature vectors to generate Combined PV-EV scenarios; Calculate the net load deviation of each scenario and filter out high-risk scenarios that meet the preset risk threshold; The conditional risk value constraint is introduced into the objective function of the virtual power plant energy and operating reserve optimization model to minimize the conditional risk value. The conditional risk value constraint formula is: Add the conditional value-at-risk constraint to the benchmark loss function to generate a benchmark scheduling plan: in, is the quantile threshold; is the confidence level; is the expectation operator; is the baseline loss function.

6. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 5 is characterized by: The intraday layer includes: Based on the periodically collected photovoltaic output forecast data, the scheduling plan is dynamically adjusted through rolling optimization. The rolling optimization time window is ; Add a real-time correction term based on the baseline objective function: in, To adjust the cost factor; is the penalty item for scheduling plan adjustment; The photovoltaic output prediction error distribution parameters are updated through Bayesian online learning.

7. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 6 is characterized by: The real-time layer includes: Based on second-by-second PV output measurement data and grid frequency regulation demand signals, event-triggered Bayesian online learning is used to correct PV output forecast error distribution parameters in real time and dynamically adjust the reserve capacity deployment strategy, including: Determine the grid frequency deviation based on the grid frequency regulation requirements and use it to calculate the standby power requirements: in, It is the frequency deviation of the power grid; For standby power requirements; For frequency modulation needs; The calculation formula for the spare capacity allocation constraint is: in, For the The standby discharge power of each building charging pile; is the standby discharge power of the energy storage device.

8. The virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to claim 7 is characterized by: Optimizing the charging and discharging priorities of building charging piles and energy storage equipment based on time-of-use electricity prices and real-time load demand includes: Receive time-of-use electricity price signals. During peak hours, building charging piles and energy storage equipment are given priority for discharge, while during off-peak hours, building charging piles and energy storage equipment are given priority for charging. Monitor real-time load demand and forecasted load demand. If the real-time load demand is greater than the forecasted load demand, call on energy storage equipment to discharge and reduce electricity purchase demand. If the real-time load demand is less than the forecasted load demand, give priority to absorbing surplus photovoltaic power and reduce abandoned light.

9. A virtual power plant control system based on dynamic parameter identification and hierarchical collaboration, based on a virtual power plant control method based on dynamic parameter identification and hierarchical collaboration according to any one of claims 1 to 8, characterized in that: The system includes: The data collection and integration module is used to collect multi-dimensional data of building distributed resources and integrate them through federated learning; the resource aggregation and modeling module is used to aggregate building distributed resources according to power generation characteristics, generate external characteristic parameters of virtual power plants and build energy and operation reserve optimization models; the parameter update and dynamic adjustment module is used to update power generation parameters online through the RLS algorithm, and dynamically adjust the optimization model in combination with the hierarchical timing collaborative framework; the optimization control and status monitoring module is used to generate control instructions and send them to building distributed resource equipment to control their operating status and power output, while monitoring the operating status of the virtual power plant in real time and adjusting the optimization model according to feedback information.

Images