Cooperative scheduling method and system for virtual power plant

Abstract

The invention relates to the technical field of electric power intelligent management, and discloses a cooperative scheduling method and system for a virtual power plant, and the method comprises the following steps: S1, collecting the real-time data of each distributed power supply, each load and an energy storage system in the virtual power plant, and carrying out the ultra-short-term prediction, and obtaining a prediction parameter; s2, dynamically calculating the dynamic operation boundary of the energy storage system based on the real-time state of the energy storage system; s3, on the day before the current operation day, generating a pre-scheduling plan through collaborative decision making of a multi-target fuzzy satisfaction function; s4, in the current running day, taking the pre-scheduling plan as a reference, updating boundaries and prediction parameters in a rolling manner, and generating a real-time scheduling instruction through model prediction; and S5, monitoring the deviation between the actual output of each resource and the real-time scheduling instruction in real time, and when the deviation exceeds a threshold value, starting a collaborative deviation compensation mechanism to carry out power balance. According to the invention, fine cooperative scheduling of different types of distributed resources can be realized in a complex environment with high uncertainty.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent power management, in particular to a collaborative scheduling method and system for a virtual power plant. BACKGROUND

[0002] With the deepening of the "double carbon" strategy, the proportion of new energy represented by wind power and photovoltaic power in the power system continues to increase, and the power grid structure is undergoing a transition from traditional centralized to clean and low-carbon, safe and efficient new power systems. Under this background, as an intelligent management system that aggregates distributed resources such as distributed power sources, energy storage systems, controllable loads, electric vehicles, and participates in the coordinated optimization of power grid operation and power market through advanced information communication technology, the virtual power plant has become a key technical path to improve the ability of the power grid to absorb renewable energy and ensure the balance of power supply and demand.

[0003] Currently, virtual power plant technology has entered the large-scale demonstration stage from the concept verification stage in China. The current mainstream mode mainly focuses on the preliminary aggregation and instruction response of resources, and most of them build software platforms with basic functions such as resource monitoring, load regulation, and demand response in terms of technical architecture, and preliminarily realize the "observable and measurable" of distributed resources. In terms of scheduling strategy, existing systems usually use methods based on price signals or direct load control, and according to the adjustment instructions issued by the power grid, the aggregated resources are simply started or stopped or power is adjusted. However, these modes still rely more on administrative means or fixed price incentive mechanisms, the target of scheduling optimization is relatively single, mainly focusing on meeting the instantaneous power balance demand of the power grid, and most of them still follow the "top-down" instruction scheduling logic, regarding the heterogeneous resources within the aggregate as a whole that can be uniformly scheduled, lacking detailed consideration of the individual differences and autonomy of the resources. There is still no systematic solution to how to coordinate the differentiated interest demands of different subjects within the resources and how to deal with the severe fluctuations of both sides of the source and load under high proportion of new energy access.

[0004] Further, although the virtual power plant has developed rapidly, it still faces several key technical difficulties in actual operation, which restricts the maximization of its effectiveness: First, it is difficult to achieve multi-objective collaborative optimization under complex operating environment. Virtual power plants aggregate photovoltaic, energy storage, flexible load and other resources, whose response speed, regulation capacity, cost structure and interest demands are different. The scheduling needs to meet the goals of power grid peak shaving, economy, high proportion of renewable energy consumption, user comfort and equipment life extension, etc. These goals often restrict each other, for example, frequent charging and discharging of energy storage for economic benefits may damage its life. Simple single-objective optimization or traditional methods such as weighted summation cannot find a fair and efficient optimal solution among all objectives, especially cannot handle the dynamic trade-off between objectives.

[0005] Secondly, the output of new energy has high uncertainty, which is easy to cause scheduling deviation. The output of photovoltaic is intermittent and volatile due to weather influence, and the load demand also has randomness. The existing prediction technology is difficult to be completely accurate, resulting in the difference between the current plan based on point prediction and the actual situation. Especially in the scenario of high proportion of new energy, the uncertainty is amplified, and the system is difficult to balance the scheduling.

[0006] Thirdly, the fine consideration of the whole life cycle loss of the energy storage system is insufficient. The existing scheduling model usually simplifies the energy storage as an ideal energy body, ignores its charging and discharging efficiency attenuation, cycle life loss and other dynamic characteristics, and fails to include the constraint of battery health state on the adjustable capacity into the optimization boundary. This not only may cause the scheduling instruction to exceed the safe operation range of the energy storage device, aggravate its aging, but also cause distortion in economic evaluation due to the failure to quantify the attenuation cost, so as to realize the real balance between short-term income and long-term asset value. SUMMARY

[0007] The present application aims to provide a collaborative scheduling method and system for a virtual power plant, which can realize fine collaborative scheduling of different types of distributed resources in a complex environment with high uncertainty, and can improve the operation economy and efficiency of the virtual power plant while ensuring the reliability of the power grid.

[0008] To achieve the above-mentioned purpose, the present application provides the following basic scheme.

[0009] Scheme one The collaborative scheduling method for a virtual power plant comprises the following steps: S1, real-time acquisition of the output data of each distributed power source in the virtual power plant, the power consumption data of each load, and the state data of the energy storage system, and based on this, super-short-term prediction is carried out to obtain prediction parameters for the next N hours, the prediction parameters at least including prediction values and corresponding uncertainty intervals; N is an integer between 1 and 6; S2, based on the real-time state of the energy storage system, dynamically calculating its dynamic operation boundary considering life attenuation, the dynamic operation boundary including the safe charging and discharging boundary based on the battery health state; S3, based on the prediction parameters, establishing and solving a first optimization model to maximize the overall operation satisfaction of the virtual power plant as the target, and generating a pre-scheduling plan for the next 24 hours on the day before the current operation day; wherein the overall operation satisfaction is the collaborative decision result of the fuzzy satisfaction function based on multiple targets; the multiple targets include grid demand matching degree, total operation cost, energy storage attenuation cost and load comfort degree; S4, within the current operation day, taking the pre-scheduling plan as a reference, and rolling the following steps at fixed time intervals: S41, according to the latest operation data, updating the dynamic operation boundary and prediction parameters of each resource; S42, under the updated boundary constraint, a second optimization model with shortened optimization time domain is established and solved, real-time scheduling instructions of the current period are generated, and are issued to each resource for execution; S5, real-time monitoring of the deviation of the actual output of each resource from the real-time scheduling instructions, when the deviation exceeds the threshold, starting the collaborative deviation compensation mechanism based on the real-time adjustable capacity of each resource to balance the power.

[0010] Scheme two A collaborative scheduling system for a virtual power plant has a computer program stored thereon, and the computer program, when executed by a processor, implements the steps of the collaborative scheduling method for the virtual power plant according to scheme one.

[0011] The working principle and advantages of the present application are: The collaborative scheduling method and system for a virtual power plant can achieve fine collaborative scheduling of different types of distributed resources in a complex environment with high uncertainty, and can improve the operation economy and efficiency of the virtual power plant while ensuring the reliability of the power grid. The focus is on: First, the present scheme breaks through the limitations of traditional scheduling and can achieve fine collaborative scheduling; and the uncertainty of new energy output and the whole life cycle loss of the energy storage system are considered and processed in detail.

[0012] Traditional scheduling methods often regard the virtual power plant as a black box or a single entity, and use top-down command control, ignoring the heterogeneity and autonomy of internal resources. In view of this, the present scheme specially designs the concept of "dynamic operation boundary", models the energy storage, load and other resources as intelligent agents with their own states and constraints, so as to accurately model the resource characteristics.

[0013] In particular, for the energy storage system, the present scheme does not simplify it as an ideal energy container, but establishes a safe charging and discharging boundary linked with the real-time health status, and quantifies the dynamic attenuation cost, so that the scheduling model can actively avoid harmful operation strategies to the battery life, not only improving the economy of scheduling decisions (explicitly including long-term asset depreciation costs into the optimization objective), but also helping to enhance the physical feasibility and operation safety of the system, preventing the risk of equipment failure caused by excessive calling. In addition, through the rolling optimization based on model predictive control and the dynamic boundary negotiation mechanism, the present scheme can realize real-time fine tuning under the current plan macro guidance, effectively smoothing the fluctuations caused by prediction errors of renewable energy and load, and has better robustness and adaptive ability.

[0014] Second, the present scheme can achieve multi-objective collaborative optimization in a complex operating environment.

[0015] First, this scheme specifically applies the theory of "fuzzy cooperative game" to the multi-objective optimization of virtual power plants, aiming to maximize overall "satisfaction" rather than a single economic indicator—a significant conceptual shift. Most existing technologies handle multi-objective optimization using weighted summation, but the weighting often relies on experience and can easily lead to poor performance of a particular objective, dragging down overall efficiency. This scheme seeks Pareto optimality through the geometric mean of the satisfaction function, requiring all sub-objectives to reach a certain level of satisfaction. This collaborative decision-making mechanism overcomes the inherent shortcomings of traditional methods, and its application in the field of virtual power plant scheduling is pioneering.

[0016] Secondly, this scheme embeds the long-term loss problem of energy storage degradation in the field of electrochemistry into the mathematical model of short-term operation optimization of power system in the form of dynamic cost factors, and combines it with the model predictive control framework to form a closed-loop optimization system that considers lifetime loss, which can effectively coordinate instantaneous decisions and long-term consequences.

[0017] Finally, through a closed loop from compensation for collaborative deviations to benefit distribution based on contribution, the former ensures that the overall collaborative effect meets expectations by dynamically adjusting the output of each participating entity (such as energy storage, distributed power sources, and controllable loads) when actual output deviates from the plan; the latter distributes benefits based on the actual contribution of each entity (such as adjustment amount, response speed, accuracy, etc.) after the collaboration is completed, reflecting the principle of "more work, more pay, and better performance, better reward". The combination of the two can link technical scheduling with market incentive mechanisms, ensuring the stability of the collaborative alliance and the enthusiasm of the participating entities. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method flow of the first embodiment of the collaborative scheduling method and system for virtual power plants of the present invention. Detailed Implementation

[0019] The following detailed explanation illustrates the specific implementation methods: Example 1 The basic implementation examples are as follows: Figure 1 The diagram illustrates a cooperative scheduling method for virtual power plants, comprising the following steps: S1: Real-time acquisition of output data from each distributed power source within the virtual power plant, electricity consumption data from each load, and status data from the energy storage system. Based on this, ultra-short-term forecasts are performed to obtain forecast parameters for the next N hours. These forecast parameters include at least the predicted value and its corresponding uncertainty interval; N is an integer between 1 and 6. In this embodiment, N is 4, meaning the ultra-short-term forecast period is 4 hours into the future. In practical applications, the value of N can be adjusted according to actual forecasting needs.

[0020] In this embodiment, real-time data is collected by terminal devices such as smart meters, photovoltaic inverters, energy storage converters (PCSs), battery management systems (BMSs) installed on the user side, and the data collection frequency can be set as needed. For example, for rapidly changing quantities such as power, voltage, and current, the collection frequency is set to once every 15 seconds; for slowly changing quantities such as SOC and temperature, it can be set to once every 1-5 minutes. The collected raw data needs to be processed through data cleaning (such as removing outliers and filling missing values) and format standardization, and can be stored in a time series database (such as InfluxDB) or a relational database (such as MySQL) for subsequent calling.

[0021] The output data of the distributed power source includes real-time active / reactive power, voltage, current, and inverter state of each distributed photovoltaic power station. The power consumption data of each load includes the total load power of each industrial and commercial user and electric vehicle charging pile, the current state and adjustable range of interruptible load. The state data of the energy storage system includes real-time state of charge (SOC), battery voltage, current, temperature, internal resistance, state of health (SOH) estimation value, and cumulative cycle count.

[0022] Specifically, the ultra-short-term prediction is realized by using an Attention-BiLSTM prediction model based on an attention mechanism, and the model is optimized by using a quantum genetic algorithm to obtain the prediction parameters.

[0023] In this embodiment, the prediction model is trained for loads and distributed power sources (photovoltaic power) respectively. Taking photovoltaic power prediction as an example, the model input features include: 72-hour historical photovoltaic output sequence, historical weather data (irradiance, temperature, humidity), 24-hour numerical weather prediction (NWP) data, and time features (hour, day type). The past year's historical data is used for model training (the historical data is divided into training data set and validation data set in the ratio of 8:2). The quantum genetic algorithm (QGA) is used to optimize the search of the hyperparameters (such as the number of hidden layer nodes, learning rate, and dropout rate) of the BiLSTM network, and the optimal model parameters are obtained by minimizing the root mean square error (RMSE) of the prediction.

[0024] In the ultra-short-term prediction, the rolling prediction of the next 4 hours is automatically triggered every 15 minutes. The latest real-time data and NWP data are input into the trained LSTM model, and the power prediction value sequence of the next 16 time points (15-minute interval) is output .

[0025] The uncertainty interval in the prediction parameter is calculated by a Conformal Prediction method, and provides a robust optimization constraint for the first and second optimization models, specifically including the following sub-steps: S101, using the verification data set, calculating the prediction error of the prediction model on each sample . is the actual value of the i-th sample, is the prediction value of the i-th sample by the prediction model. reflects the deviation degree of the model prediction value and the actual value.

[0026] S102, arranging the absolute values of all sample prediction errors in ascending order, so that subsequent appropriate error values are selected to determine the uncertainty radius according to the sorting results.

[0027] S103, for a new prediction point t, selecting the error absolute value ranked at the position as the uncertainty radius . Wherein, n is the number of samples in the verification data set, is the significance level (for example, set , indicating 90% confidence), represents rounding up, and the error absolute value selected through this position can ensure that the actual value of the new prediction point falls within the subsequent determined interval under a certain confidence.

[0028] S104, generating the uncertainty interval of the prediction point as .

[0029] is the prediction value of the new prediction point t by the prediction model.

[0030] Finally, the obtained prediction parameter contains two parts: the prediction value sequence and the corresponding uncertainty interval sequence. The upper and lower limits of the uncertainty interval can be subsequently used as part of the constraint conditions of the optimization problem (i.e., providing a robust optimization constraint for the first and second optimization models), forcing the scheduling scheme to remain feasible and safe under any possible adverse scenario, thereby enhancing the ability of the virtual power plant to cope with uncertainty.

[0031] S2, dynamically calculating the dynamic operation boundary considering the life attenuation of the energy storage system based on the real-time state of the energy storage system, the dynamic operation boundary including a safe charging and discharging boundary based on the battery health state.

[0032] The safe charging and discharging boundary is defined as: The degradation cost coefficient is calculated online by a degradation model, which is a function of the current state of charge (SOC) of the energy storage system, the cumulative equivalent cycle number, the average discharge depth, and the battery temperature.

[0033] Specifically, the safe charging and discharging boundary is not fixed, but is dynamically adjusted according to the real-time state of the energy storage. In this embodiment, the following rules can be used to set the boundary: the basic boundary is set to [20%, 90%], if the average battery temperature is continuously higher than 35℃, the upper and lower boundaries are both contracted by 5% to [25%, 85%], so as to reduce high-rate charging and discharging and slow down the degradation; if the state of health (SOH) of the battery is lower than 80%, the basic boundary is further contracted to [30%, 80%], so as to prolong the remaining service life.

[0034] When the actual SOC of the energy storage system exceeds the safe charging and discharging boundary, a higher degradation cost coefficient is assigned to it in the optimization model In this embodiment, the coefficient is calculated by the following empirical model: ; wherein I(t) is the current charging and discharging current, reflecting the intensity of charging and discharging; C is the rated capacity, used to normalize the current for unified analysis; CycleCount is the cumulative cycle number, and CycleLife is the rated cycle life; exp is the abbreviation of exponential function, representing the exponential operation with the natural constant e as the base. The coefficient 、 、 、 is obtained by fitting the data of the battery accelerated aging test. The degradation cost coefficient is used to quantify the equivalent economic loss caused by the unit charging and discharging operation at the current time.

[0035] S3, on the day before the current operation day, a first optimization model is established and solved based on the predicted parameters to maximize the overall operation satisfaction of the virtual power plant, and a pre-scheduling plan for the next 24 hours is generated; wherein the overall operation satisfaction is the result of the collaborative decision of the fuzzy satisfaction function based on multiple objectives; the multiple objectives include the grid demand matching degree, the total operation cost, the energy storage degradation cost, and the load comfort degree.

[0036] The form of the fuzzy satisfaction function is the weighted geometric mean of multiple sub-objective satisfaction functions, that is, wherein is the satisfaction degree of the i-th sub-objective, is the weight thereof, and x is the scheduling decision variable (such as the energy storage charging and discharging power, the load adjustment amount, etc.).

[0037] Specifically, in this embodiment, .

[0038] wherein, is the cost satisfaction, which is inversely proportional to the total operation cost, the lower the cost, the closer the satisfaction to 1.

[0039] is the grid demand matching satisfaction, which is inversely proportional to the deviation of actual response and dispatching instruction.

[0040] is the energy storage attenuation satisfaction, which is inversely proportional to the total attenuation cost.

[0041] The total attenuation cost is equal to ; is the attenuation cost coefficient, is the absolute value of battery charging and discharging power, reflecting the charging and discharging intensity.

[0042] is the load comfort satisfaction, which is inversely proportional to the total load reduction and the degree of deviation from its comfort range.

[0043] NSGA-II multi-objective genetic algorithm is adopted for solution, and the algorithm outputs a set of Pareto optimal solutions. A compromise scheme is selected from the set of Pareto optimal solutions by the dispatcher according to the current strategy as the current plan, i.e. the pre-dispatching plan.

[0044] S4, within the current operating day, taking the pre-dispatching plan as a reference, the following steps are executed at a fixed time interval: S41, according to the latest operating data, updating the dynamic operating boundary and prediction parameters of each resource; S42, under the constraint of the updated boundary, establishing and solving a second optimization model with shortened optimization time domain to generate real-time dispatching instructions for the current period and issuing the instructions to each resource for execution.

[0045] The second optimization model adopts a model predictive control (MPC) framework, and the shortened optimization time domain is 2 to 6 hours. In this embodiment, the optimization time domain is 4 hours (16 points) in the future, i.e. at each optimization, the situation in the next 4 hours is analyzed and optimized based on the prediction data. The control time domain is 1 hour (4 points), i.e. the time range for actually executing the dispatching decision is the next 1 hour, and the focus is on optimizing the resource power setting in this time period.

[0046] Specifically, in this embodiment, every 15 minutes, the latest prediction parameters (such as load prediction, renewable energy output prediction, etc.) and the actual state of the resources (especially the state of charge (SOC) of the energy storage) are obtained. Then, taking the existing pre-dispatching plan as a reference trajectory, an optimization problem is solved again.

[0047] And the re-solved optimization problem is a simplified version of the first optimization model, which can be solved by a faster quadratic programming (QP) solver. The result is the accurate power setting of each resource (such as energy storage, distributed power, controllable load, etc.) in the next hour, which is used to guide the actual operation scheduling.

[0048] S5, real-time monitoring of the actual output of each resource and the deviation of the real-time scheduling instruction, when the deviation exceeds the threshold (such as 5%), start the collaborative deviation compensation mechanism based on the real-time adjustable capacity of each resource to balance the power.

[0049] The collaborative deviation compensation mechanism includes the following operations: According to the power shortage or surplus and the real-time adjustable margin of each resource, calculate the real-time contribution capacity of each resource; wherein the real-time adjustable margin is calculated based on the difference between the current actual output of the resource and the maximum output of the resource, and the closeness between the current state of the resource and the dynamic operation boundary.

[0050] Specifically, for each resource i, calculate its real-time adjustable margin . Wherein, is the maximum output of resource i, is the current actual output, is the rated output, which reflects how much adjustment space the resource has.

[0051] For energy storage, if its SOC (state of charge) has approached the dynamic operation boundary (i.e. the safe charging and discharging boundary), its margin will be reduced. Because when the energy storage approaches the boundary, continued charging and discharging may damage the life or exceed the safety range, so reduce its adjustable margin to avoid excessive calling.

[0052] Wherein, based on the normalized distance D of the SOC of the energy storage relative to the upper limit or lower limit of the safe charging and discharging boundary, when D is less than 0.1, it is determined that the boundary has been approached, and the adjustable margin is reduced according to the reduction coefficient . The reduction coefficient is calculated using a sigmoid function (Sigmoid function); .

[0053] According to the size of the real-time contribution capacity, the power value that needs to be compensated is proportionally distributed to one or more resources to jointly bear.

[0054] Specifically, let the total deviation be , and the compensation amount allocated to resource i is . By preferentially calling resources with large margins, the adjustment capacity of the resources can be more efficiently utilized, and the deviation can be quickly eliminated.

[0055] After a demand response or ancillary service is completed, the contribution of each resource in the collaborative bias compensation mechanism is calculated based on the actual compensation amount, response speed and response accuracy of the resource, and the Shapley value method is used for benefit distribution.

[0056] Specifically, after a response event is completed, the system calculates a contribution score for each resource j participating in the response ; is the actual compensation amount of the resource j, is the response speed of the resource j, is the response accuracy of the resource j; a, b and c are weights of the corresponding indexes, the actual compensation amount reflects how much the resource output is adjusted, the response delay reflects the timeliness of the resource response, and the response accuracy represents the accuracy of the resource in executing the instruction, and the three dimensions are comprehensively measured to measure the contribution of the resource.

[0057] The Shapley value method is used for benefit distribution. The Shapley value method is a method for fair distribution of total benefits in cooperative games, and the core is to calculate the average marginal contribution of each participant in all possible cooperative combinations. In actual implementation, an approximate algorithm can be used to calculate the average marginal contribution of each resource in all possible participation combinations. The benefit obtained by the resource j is , wherein, is the total benefit of the event. This method can fairly reflect the actual value of each resource in the collaborative alliance and ensure the enthusiasm of the participants.

[0058] The embodiment also provides a collaborative scheduling system for a virtual power plant, and a computer program is stored on the collaborative scheduling system. The computer program is executed by a processor to implement the steps of the collaborative scheduling method for the virtual power plant.

[0059] The collaborative scheduling method and system for the virtual power plant provided by the embodiment can realize fine collaborative scheduling of different types of distributed resources in a complex environment with high uncertainty, and can improve the operation economy and efficiency of the virtual power plant while ensuring the reliability of the power grid.

[0060] In addition, to verify the application effect of the scheme, a power dispatching case of an industrial park in Xijin district during a summer peak period is selected for simulation analysis. It is assumed that the participating resources include a photovoltaic resource (a roof photovoltaic resource in the park, with a total capacity of 2 MW), an energy storage resource (a 1 MW / 2 MWh lithium iron phosphate battery energy storage system), a load (an injection molding workshop (interruptible load, with a maximum power reduction of 500 kW) and a central air conditioning system (flexible load, with a power reduction of 300 kW) in the park.

[0061] The grid event is set as follows: the day before (14th), the virtual power plant platform receives a demand response instruction from the grid dispatch center: it is required to reduce a total of 1000 kW of load power during 12:00-14:00 on the 15th to alleviate the pressure on the regional power grid.

[0062] Step one: current collaborative optimization (afternoon of the 14th).

[0063] Based on the weather forecast (sunny, strong radiation), it is predicted that the photovoltaic output will reach a peak of 1.8 MW at noon on the 15th. Load prediction shows that the basic load is 3.5 MW at noon.

[0064] Query the state of energy storage, the current SOC is 70%, the state of health (SOH) is 92%, and the recent cycle count is stable. According to these data, the dynamic safety boundary of energy storage during the noon period is calculated as [25%, 85%], and a moderate decay cost factor is generated.

[0065] The current optimization is carried out with the goal of "maximizing overall satisfaction".

[0066] The optimization result (pre-scheduling plan) is: 12:00-14:00: photovoltaic: full output 1.8 MW, all uploaded to the grid.

[0067] Energy storage: from 12:00, discharge at a power of 500 kW for 2 hours, reduce SOC from 70% to 50%.

[0068] This decision is because its SOC is always far from the safety boundary, and the decay cost is controllable.

[0069] Load: injection molding workshop reduces 500 kW, central air conditioning reduces 200 kW.

[0070] Total reduction: 500 kW (energy storage) + 500 kW (injection molding) + 200 kW (air conditioning) = 1200 kW > 1000 kW, leaving a 200 kW margin to deal with uncertainties.

[0071] Step two: intra-day rolling optimization and real-time collaboration (noon on the 15th) 12:00: the plan starts. Everything is normal. 12:45: a sudden situation. The weather radar monitors a temporarily generated cloud cluster moving towards the park. The prediction model immediately updates the "prediction parameters" and predicts that the photovoltaic output will drop from 1.8 MW to 1.2 MW in the next 30 minutes, and the power balance will be broken (photovoltaic drops 600 kW, but the grid's 1000 kW reduction instruction still needs to be completed).

[0072] The current SOC of energy storage is 60%, and the output power is 500 kW. The injection molding workshop has reduced 500 kW as planned. The air conditioner has reduced 200 kW.

[0073] The calculation found that if the energy storage is allowed to increase output to make up for the gap in photovoltaic, the SOC will quickly decrease and touch the lower limit of the dynamic safety boundary (25%), at which time the attenuation cost factor will rise sharply.

[0074] The cooperative bias compensation mechanism is started. A negotiation request is sent to the park central air conditioning system: "Are you willing to increase the load reduction by 100kW on the existing basis for 25 minutes? The platform will pay additional incentive fees." The air conditioning system determines that the operation is feasible according to the indoor temperature comfort model, automatically confirms the request, and issues a new dispatching instruction - air conditioner: additional reduction of 100kW load (total reduction of 300kW). Energy storage: maintain the original planned 500kW discharge power unchanged to avoid entering the high attenuation risk area.

[0075] Total reduction: 500kW (energy storage) + 500kW (injection molding) + 300kW (air conditioner) = 1300kW. After the photovoltaic decreases by 600kW, the net on-grid power still meets the grid's 1000kW reduction requirement.

[0076] In summary, in the presence of sudden situations (such as weather deterioration), the scheme can ensure that the dispatching target is met through cooperative scheduling, and in the face of sudden situations, not by a single resource hard, but by mobilizing load resources through the "negotiation mechanism" to compensate for cooperation, achieving multi-agent game, while through dynamic boundaries and attenuation cost factors, successfully avoiding the operation of energy storage in unhealthy conditions, achieving the optimal total life cycle cost.

[0077] Embodiment Two The cooperative scheduling method for a virtual power plant further includes S6, an adaptive learning step, based on embodiment one: Periodically collect historical operation data, including prediction bias, energy storage attenuation model bias, and use historical operation data to update and train the prediction model used to generate the prediction parameters and the attenuation model.

[0078] Specifically, the collection step of the prediction bias data includes: Collect the ultra-short-term load prediction value, photovoltaic output prediction value and their corresponding actual measured values every 15 minutes in the past week, as well as the corresponding weather data (actual irradiance, temperature, etc.) at the prediction time. From this, the absolute prediction error at each time point: |predicted value - actual value| can be calculated as the prediction bias data.

[0079] The collection step of the energy storage attenuation model bias data includes Collect detailed records of each charge-discharge cycle of the energy storage system in the past week, including: initial SOC, final SOC, average charge-discharge power, average temperature. Perform a complete capacity calibration on the energy storage system every week (which can be done through full charge and discharge tests or online evaluation by professional equipment) to obtain the actual measured capacity degradation value. According to the expected capacity degradation value calculated by the degradation model (for example, based on the cumulative ampere-hour throughput multiplied by a degradation coefficient), the prediction deviation of the degradation model can be calculated: | expected degradation value - measured degradation value |, as the energy storage degradation model deviation data.

[0080] Clean the collected raw historical operation data, and remove invalid data caused by communication interruption and equipment failure. Then align the data by time and construct them into a sample format suitable for model training. For example, for the prediction model, each sample contains "input features (historical data, weather forecast)" and "label (actual value)".

[0081] In the update training, the update of the prediction model adopts an incremental learning or regular full retraining strategy. The incremental learning strategy includes: calculating the average prediction error (such as MAE) of the past week every week. If the error exceeds the preset threshold (for example, 15% higher than the baseline error) for two consecutive weeks, an incremental learning is triggered immediately. Incremental learning only uses new data in the last few weeks to fine-tune the existing model parameters to adapt to the changes in data distribution in the short term, with less computational overhead.

[0082] The regular full retraining strategy includes: automatically triggering a full retraining of the prediction model at the first weekend of each month. Use all available historical data in the past year (or two years) as the training set to retrain the model to capture long-term seasonal and pattern changes, with larger computational overhead.

[0083] The update of the degradation model is performed every quarter, including the following steps: aggregate all degradation model prediction deviation data in the current quarter, and use the Bayesian update method to correct the key parameters (such as the coefficient 、 、 ) in the degradation model. Specifically, the existing parameters are considered as the prior distribution, and the measured deviation data in the current quarter are considered as the observed evidence. The posterior distribution of the parameters is calculated by the Bayesian formula. The mean of the posterior distribution is the new parameter after correction.

[0084] Optionally, the update of the degradation model can also be performed by fitting, taking "cumulative equivalent cycle number" or "cumulative charge-discharge energy" as the independent variable, and "measured capacity degradation" as the dependent variable, to re-fit the function form of the original degradation model.

[0085] Compared with the first embodiment, the embodiment provides a method and system for the coordinated scheduling of a virtual power plant, which can periodically update and train the prediction model and the attenuation model using historical operation data, can obtain more accurate prediction model and attenuation model, can make more reasonable arrangements in scheduling decisions, equipment maintenance and other aspects, can reduce system failures and power outage events caused by inaccurate prediction or incorrect estimation of equipment state, and thus can improve the reliability of the entire power system.

[0086] The above is only an embodiment of the present application, and the common knowledge of specific structures and characteristics in the scheme is not described in detail. The ordinary skilled person in the art knows all the ordinary technical knowledge in the field of the present application before the application date or the priority date, can know all the prior art in the field, and has the ability to apply conventional experimental means before that date. The ordinary skilled person in the art can improve and implement the present scheme based on their own ability under the guidance of the present application. Some typical known structures or known methods should not be an obstacle for the ordinary skilled person in the art to implement the present application. It should be noted that for those skilled in the art, without departing from the structure of the present application, a number of modifications and improvements can be made, which should be considered as the protection scope of the present application. These will not affect the effect and practicality of the present application.

Claims

1. A collaborative scheduling method for a virtual power plant, characterized in that, Includes the following steps: S1. Real-time acquisition of output data of each distributed power source, power consumption data of each load, and status data of the energy storage system within the virtual power plant, and based on this, ultra-short-term forecasting is performed to obtain forecast parameters for the next N hours. The forecast parameters include at least the forecast value and the corresponding uncertainty interval; N is an integer between 1 and 6. S2, Based on the real-time state of the energy storage system, dynamically calculate its dynamic operating boundary considering lifetime degradation, the dynamic operating boundary including the safe charge and discharge boundary based on the battery health state. S3, the day before the current operating day, based on the predicted parameters, with the goal of maximizing the overall operational satisfaction of the virtual power plant, establish and solve the first optimization model to generate a pre-schedule plan for the next 24 hours; wherein, the overall operational satisfaction is the result of collaborative decision-making based on a fuzzy satisfaction function of multiple objectives; the multiple objectives include grid demand matching degree, total operating cost, energy storage attenuation cost, and load comfort; S4. Within the current operating day, referencing the pre-scheduled plan, execute the following steps on a rolling basis at fixed time intervals: S41, based on the latest operational data, update the dynamic operational boundaries and prediction parameters of each resource; S42, under the updated boundary constraints, establish and solve the second optimization model that shortens the optimization time domain, generate the real-time scheduling instructions for the current period, and issue them to each resource for execution; S5, monitor the deviation between the actual output of each resource and the real-time scheduling command in real time. When the deviation exceeds the threshold, initiate a collaborative deviation compensation mechanism based on the real-time adjustable capability of each resource to balance the power.

2. The collaborative scheduling method for virtual power plant according to claim 1, wherein, In S2, the safe charging and discharging boundary is defined as: , which is a function of the current state of charge (SOC), cumulative equivalent cycle number, average discharge depth and battery temperature of the energy storage system, and is calculated online by a degradation model; when the actual SOC of the energy storage system exceeds the safe charging and discharging boundary, a higher degradation cost coefficient is assigned to it in the optimization model.

3. The method for collaborative scheduling of virtual power plant according to claim 1, wherein, In S3, the form of the fuzzy satisfaction function is a weighted geometric mean of multiple sub-target satisfaction functions, i.e. wherein, is the satisfaction of the i-th sub-target, is the weight thereof, and x is the scheduling decision variable.

4. The collaborative scheduling method for virtual power plant according to claim 1, wherein, In S4, the second optimization model adopts the Model Predictive Control (MPC) framework, and the shortened optimization time domain is 2 to 6 hours.

5. The method for collaborative scheduling of virtual power plant according to claim 1, wherein, In S5, the cooperative deviation compensation mechanism includes the following operations: The real-time contribution capacity of each resource is calculated based on the power deficit or surplus and the real-time adjustable margin of each resource; wherein, the real-time adjustable margin is calculated based on the difference between the current actual output of the resource and its maximum output, and the degree of proximity of the current state of the resource to the dynamic operating boundary. Based on the magnitude of the real-time contribution capability, the power value that needs to be compensated is proportionally allocated to one or more resources to share the burden.

6. The method for collaborative scheduling of virtual power plant according to claim 5, wherein, After completing a demand response or ancillary service, the contribution of each resource in the collaborative deviation compensation mechanism is calculated based on the actual compensation amount, response speed, and response accuracy, and the Shapley value method is used for benefit allocation. 7.The method for collaborative scheduling of virtual power plant according to claim 1, wherein, The ultra-short-term prediction is achieved using a bidirectional long short-term memory network prediction model based on an attention mechanism, and the model is optimized for hyperparameters using a quantum genetic algorithm to obtain the prediction parameters. 8.The method for collaborative scheduling of virtual power plant according to claim 1, wherein, The uncertainty interval in the prediction parameters is calculated using the Conformal Prediction method, and provides robust optimization constraints for the first and second optimization models.

9. The method for collaborative scheduling of virtual power plant according to claim 2, wherein, It also includes S6, the adaptive learning step: Historical operating data, including prediction bias and energy storage attenuation model bias, are collected periodically, and the historical operating data is used to update and train the prediction model used to generate the prediction parameters and the attenuation model.

10. A collaborative scheduling system for a virtual power plant, characterized in that, A computer program product, comprising a computer readable medium having stored thereon computer program, the computer program comprising instructions executable by a processor to cause the processor to perform the steps of the method of any one of claims 1 to 9. A computer program product, comprising a computer readable medium having stored thereon computer program, the computer program comprising instructions executable by a processor to cause the processor to perform the steps of the method of any one of claims 1 to 9.

Images