Auxiliary decision-making method, device and equipment for virtual power plant operation, medium and product

Abstract

The invention discloses an auxiliary decision-making method and device for virtual power plant operation, equipment, a medium and a product, and relates to the technical field of data processing, and the method comprises the steps: collecting an original data set of a virtual power plant; wherein the original data set comprises equipment data, market price data, meteorological prediction data, historical data and rule data of the Internet of Things equipment in the virtual power plant; preprocessing the original data set to obtain a structured feature data set; constructing a virtual power plant knowledge graph based on the structured feature data set; based on a preset operation target and operation constraints, generating a plurality of candidate operation schemes and reference information of each candidate operation scheme by using the structured feature data set and the virtual power plant knowledge graph; and outputting a visual evaluation chart of each candidate operation scheme, and obtaining an optimal operation scheme determined by decision makers of the virtual power plant from the plurality of candidate operation schemes based on the visual evaluation charts, thereby improving the accuracy of the finally selected operation scheme.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, equipment, medium and product for assisting decision-making in the operation of a virtual power plant. Background Technology

[0002] With the continuous increase in the proportion of renewable energy generation and the deepening of power market reforms, virtual power plants, as a key technological means to integrate distributed energy resources and improve the flexibility and reliability of the power system, are becoming increasingly important. In daily operation, virtual power plants need to precisely coordinate various resources such as air conditioning, photovoltaics, energy storage, and controllable loads across numerous scenarios including day-ahead / day-intraday scheduling, demand response, deviation assessment, ancillary services, and compliance risk control. They must quickly select operating schemes that excel in economics, safety, compliance, and carbon efficiency to adapt to the complex and ever-changing power market environment and power system operation requirements.

[0003] However, current virtual power plants severely lack automation and intelligent support in actual operation decision-making. Decision-makers typically need to manually construct possible operation plans based on their own experience, and when faced with multiple candidate operation plans, they find it difficult to quantitatively analyze the different candidate operation plans. The final decision is highly susceptible to the influence of the decision-maker's subjective factors, resulting in an inaccurate selection of the final operation plan. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, equipment, medium, and product for assisting decision-making in the operation of a virtual power plant, which can improve the accuracy of the final selected operation plan.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a decision support method for operating a virtual power plant, including: Collect the raw dataset of the virtual power plant; wherein, the raw dataset includes device data, market price data, weather forecast data, historical data, and rule data of the IoT devices in the virtual power plant; the device data of the IoT devices includes at least device ID, collection timestamp, and device parameters; The original dataset is preprocessed to obtain a structured feature dataset; A virtual power plant knowledge graph is constructed based on the structured feature dataset. Based on preset operational goals and constraints, the structured feature dataset and the virtual power plant knowledge graph are used to generate multiple candidate operational schemes and reference information for each candidate operational scheme; wherein, the reference information includes scheme scores, key factor contributions, confidence intervals, and scheme risk warnings; Output visual evaluation charts for each candidate operation plan, and obtain the optimal operation plan determined by the virtual power plant's decision-makers from multiple candidate operation plans based on the visual evaluation charts.

[0006] Secondly, this application provides an auxiliary decision-making device for virtual power plant operation, comprising: The data acquisition unit is used to acquire the raw dataset of the virtual power plant; wherein, the raw dataset includes device data, market price data, weather forecast data, historical data, and rule data of the IoT devices in the virtual power plant; the device data of the IoT devices includes at least device ID, acquisition timestamp, and device parameters; The preprocessing unit is used to preprocess the original dataset to obtain a structured feature dataset; The construction unit is used to construct a virtual power plant knowledge graph based on the structured feature dataset; The generation unit is used to generate multiple candidate operation schemes and reference information for each candidate operation scheme based on preset operation goals and operation constraints, using the structured feature dataset and the virtual power plant knowledge graph; wherein, the reference information includes scheme score, key factor contribution, confidence interval and scheme risk warning; The output unit is used to output visual evaluation charts of each candidate operation plan and obtain the optimal operation plan determined by the decision-makers of the virtual power plant from multiple candidate operation plans based on the visual evaluation charts.

[0007] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the auxiliary decision-making method for operating a virtual power plant as described in any of the above.

[0008] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the auxiliary decision-making method for operating a virtual power plant as described above.

[0009] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the auxiliary decision-making method for operating a virtual power plant as described above.

[0010] In a sixth aspect, this application provides a chip including a processor and a communication interface coupled to the processor. The processor is used to run a program or instructions, and when the processor executes the program or instructions, it implements the steps of the auxiliary decision-making method for virtual power plant operation as described above.

[0011] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, device, equipment, medium, and product for auxiliary decision-making in virtual power plant operation. It collects raw datasets from multiple sources, including device data from IoT devices, market price data, weather forecast data, historical data, and rule data, and preprocesses them into a structured feature dataset. Based on this structured feature dataset, a virtual power plant knowledge graph is constructed, achieving deep correlation and knowledge fusion of multi-source heterogeneous data, which helps to uncover the potential value of the data. Furthermore, based on preset operational goals and constraints, multiple candidate operational plans are generated, along with reference information including plan scores, key factor contributions, confidence intervals, and plan risk warnings, allowing decision-makers to comprehensively understand the advantages and disadvantages of each plan from multiple dimensions. Visual evaluation charts are output, presenting complex data intuitively, reducing the difficulty of understanding for decision-makers, and enabling them to make scientific judgments based on comprehensive and accurate information. This effectively avoids decision-making errors caused by incomplete information or misunderstandings, thereby improving the accuracy of the final selected operational plan. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating an auxiliary decision-making method for virtual power plant operation provided in an embodiment of this application; Figure 2 A schematic diagram of the functional modules of an auxiliary decision-making device for virtual power plant operation provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] In one exemplary embodiment, such as Figure 1 As shown, a decision support method for virtual power plant operation is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, it includes the following steps 101 to 105. Wherein: Step 101: Collect the raw dataset of the virtual power plant.

[0017] In this embodiment of the application, the original dataset includes device data, market price data, weather forecast data, historical data, and rule data of the IoT devices in the virtual power plant; the device data of the IoT devices includes at least device ID, collection timestamp, and device parameters.

[0018] Internet of Things (IoT) devices include at least photovoltaic (PV) devices, energy storage devices, and load devices (such as air conditioners). The device data of the PV devices includes at least the power generation (PV_power), the device data of the energy storage devices includes at least the state of charge (SOC), and the device data of the load devices includes at least the current load. For example, Load_baseline: baseline non-adjustable load (uncontrollable); Load_flex / Load_shiftable: adjustable / shiftable load (can participate in demand response); Load_EV: electric vehicle charging load; Load_AC: air conditioning power consumption, etc. Device data from IoT devices can be uploaded via MQTT / Modbus protocol, with a collection frequency of 5 minutes / time. This application does not limit this aspect.

[0019] The market price data includes multiple price data sets, each of which includes at least the price collection time, the energy price (Price_t), and the ancillary services market price (AS_price_t).

[0020] Meteorological forecast data can include temperature, wind speed, irradiance, etc., and is uniformly converted into an internal data structure using JSON format.

[0021] Historical data and rule data can be entered by operations and maintenance personnel through the web backend, and batch import of Word / PDF is supported. The system automatically extracts key information using NLP.

[0022] Step 102: Preprocess the original dataset to obtain a structured feature dataset.

[0023] In this embodiment of the application, preprocessing of the original dataset may include constant cleaning, linear interpolation of missing values, Min-Max normalization, and statistical feature extraction.

[0024] As an optional implementation, step 102, which preprocesses the original dataset to obtain a structured feature dataset, may include: The original dataset is cleaned of outliers to obtain the first preprocessed dataset; Missing data are imputed in the first preprocessed dataset to obtain the second preprocessed dataset; The second preprocessed dataset is normalized to obtain a structured feature dataset.

[0025] This implementation method first involves outlier cleaning, which effectively removes interference noise from the original data, preventing outliers from misleading subsequent analysis and resulting in a cleaner first preprocessed dataset. Next, missing data imputation ensures data integrity and prevents data gaps from affecting the accuracy of knowledge graph construction and solution generation. Finally, normalization is performed to unify data units, making different feature data comparable and laying the foundation for generating a high-quality structured feature dataset, thereby improving the reliability and accuracy of virtual power plant operation support decisions.

[0026] In this embodiment of the application, outlier cleaning is performed on the original dataset to remove outliers that exceed the safe range. For example, PV_power∈[0, SOC∈[0%,100%], exceeding [0, PV_power exceeding [0%, 100%] and SOC exceeding [0%, 100%] are both abnormal values ​​and need to be deleted.

[0027] In this embodiment, missing data can be filled using linear interpolation. When the data at time point t... When missing, linear interpolation is performed using the values ​​at time points t−1 and t+1 to estimate. The value of can be calculated using the following formula:

[0028] in, This represents the data value after filling in the gaps at time t. This represents the valid data at time t-1. This represents the valid data at time t+1.

[0029] For example, if data is lost in the 3rd minute, and the data for the 2nd minute is known to be 10 and the data for the 4th minute is known to be 14, then the data to be filled in the 3rd minute would be: 10 + (14-10) / 2 = 12.

[0030] In this embodiment of the application, the original dataset can also be normalized to map the original data X to the interval [0,1] so that when the subsequent model (such as LSTM) is trained, the feature values ​​are at the same scale, which facilitates the convergence of the gradient descent process.

[0031] In addition, feature extraction can be performed on the original dataset to enhance the separability of scene recognition and optimization solutions, for example: 1. Construct a statistical measure of daily average photovoltaic power generation, calculate the average photovoltaic power generation at all times within a day, and reflect the overall power generation level of the day:

[0032] in, This represents the average photovoltaic power generation; N represents the number of data points collected in a day (e.g., 288, collected every 5 minutes). This represents the photovoltaic power generation value recorded in the i-th instance.

[0033] For example: If samples are collected 4 times a day, at 10, 12, 14, and 8 respectively, then... =(10+12+14+8) / 4=11.

[0034] 2. The SOC of the energy storage system must operate within a safe range and must not exceed the upper or lower limits: SOCmin ≤ SOC(t) ≤ SOCmax Wherein, SOC(t): the state of charge (percentage of charge) of the energy storage system in time t; SOCmin: the minimum allowable state of charge of the battery (e.g., 20%, to avoid over-discharge); SOCmax: the maximum allowable state of charge of the battery (e.g., 90%, to avoid over-charging).

[0035] Step 103: Construct a virtual power plant knowledge graph based on the structured feature dataset.

[0036] In this embodiment, target entities such as "device / event / parameter" can be extracted using NLP models (e.g., BERT+CRF), and target relationships such as "device-parameter-value" can be extracted using regular expression templates. Ontology design: The "device" class (attributes: type, capacity, status, etc.) is defined using Protégé (an open-source platform for ontology editing and knowledge acquisition). A conceptual system of devices, topologies, markets, weather, events, rules, cases, constraints, and indicators is defined based on OWL / RDF (WebOntology Language / Resource Description Framework). Triples (subject-predicate-object) are generated and stored in Neo4j, supporting Cypher and graph algorithms. Real-time data is aligned using ID / attribute similarity, and nodes and edges are updated incrementally. Triple generation: e.g., (energy storage, SOC, 80%), (peak load, trigger, scheduling rule A). The graph database (Neo4j) is used for storage, and real-time data is aligned using ID or attribute similarity, dynamically updating nodes.

[0037] As an optional implementation, step 103, which involves constructing a virtual power plant knowledge graph based on the structured feature dataset, may include: Extract target entities from the structured feature dataset; Extract the target relationships between target entities from the structured feature dataset; Based on the target entity and the target relationship, multiple triples are generated; The multiple triples are stored in a graph database, and a virtual power plant knowledge graph is constructed based on the multiple triples.

[0038] This implementation method, by extracting target entities and relationships from structured feature datasets, can accurately locate key information and clarify the inherent connections between data. Generating multiple triples and storing them in a graph database allows for efficient organization and management of complex data relationships, achieving structured data storage and rapid retrieval. Based on this, the constructed virtual power plant knowledge graph can intuitively present the relationships between various elements of the virtual power plant, providing comprehensive and systematic knowledge support for the subsequent generation of candidate operation plans. This helps improve the scientific rigor and accuracy of decision-making and enhances the intelligent level of virtual power plant operation.

[0039] As an optional implementation, after step 103, the following steps may also be performed: Based on the market price data, determine the total cost of electricity purchase within the preset time period; Based on the device data of the IoT device, determine the system stability index within the preset time period; The total cost of electricity purchase and the system stability indicators are defined as operational targets. Based on the device data of the IoT device, the maximum power generation of the photovoltaic device, the state of charge range of the energy storage device, and the load balance of the load device are determined. The maximum power generation, the state of charge range, and the load balance are defined as operational constraints.

[0040] This implementation method, which determines the total cost of electricity purchase based on market price data and sets system stability indicators as operational targets based on IoT device data, comprehensively considers both economic efficiency and stability, making virtual power plant operation decisions more aligned with actual needs. Simultaneously, by defining the maximum power output of photovoltaic equipment, the state-of-charge range of energy storage equipment, and the load balance of load equipment as operational constraints based on IoT device data, the operational boundaries can be precisely defined. This clear quantification of targets and constraints provides clear guidance for the subsequent generation of candidate operation plans, helping to improve the scientific rigor and rationality of the plans and ensuring the efficient and stable operation of the virtual power plant.

[0041] In this embodiment of the application, the total cost of purchasing electricity within a preset time period can be an economic target (calculating the total electricity cost incurred from purchasing electricity from the grid within a time range, such as one day or one month):

[0042] Where TotalCost represents the total cost of electricity (in yuan); T represents the total number of time steps (e.g., data is collected once per hour, T = 24 represents one day); This represents the power (kW) purchased from the grid during time period t. This represents the electricity price (yuan / kWh) for time period t.

[0043] For example: If 10kW of electricity is purchased in the first hour at a price of 1 yuan, and 8kW of electricity is purchased in the second hour at a price of 1.2 yuan, then TotalCost = 10 × 1 + 8 × 1.2 = 19.6 yuan.

[0044] In this embodiment, the system stability index within a preset time period can be a safety target (a weighted sum of voltage and frequency deviations at different times of the day, reflecting system stability. The larger the deviation, the larger the value, indicating a more unstable system):

[0045] Where Stability represents the system stability index (the smaller the better); T represents the number of time periods; The reference value for voltage is indicated (e.g., 230V or 400V); V(t) represents the actual voltage at time t. The reference value represents the frequency (e.g., 50Hz or 60Hz); f(t) represents the actual frequency in time period t. , This represents a weighting coefficient used to adjust the relative importance of voltage and frequency deviations (e.g., ...). =1, =2); For example: If V = 218V at a certain time period, =220V, f=49.9Hz =50Hz, =1, =2, then the deviation for this period is 1×|220-218|+2×|50-49.9|=2+0.2=2.2.

[0046] In this embodiment of the application, the operational constraints can specifically be: (1) Photovoltaic power generation:

[0047] Meaning: Photovoltaic power generation cannot exceed the maximum power generation capacity of the photovoltaic equipment, nor can it be negative.

[0048] Explanation of each variable: This represents the photovoltaic power generation in time period t; This indicates the maximum power generation capacity of the photovoltaic equipment.

[0049] For example: If =100kW, then It can only operate between 0 and 100 kW.

[0050] (2) State of charge (SOC) range of energy storage devices: SOCmin ≤ SOC(t) ≤ SOCmax Meaning: The SOC (State of Charge) of an energy storage system must be within a safe range.

[0051] Explanation of each variable: SOC(t): State of charge (percentage of charge) of the energy storage system in time period t; SOCmin: The minimum allowable state of charge of the battery (e.g., 20%, to avoid over-discharge); SOCmax: The maximum allowable state of charge of the battery (e.g., 90%, to avoid overcharging); For example, if SOCmin=20% and SOCmax=100%, then SOC(t) can only be between 20% and 100%.

[0052] (3) Load balancing of load equipment:

[0053] Meaning: At any given time, the sum of the outputs of all power sources must equal the total load.

[0054] Explanation of each variable: This represents the photovoltaic power generation in time period t; This represents the energy storage power during time period t; This represents the power purchased by the power grid during time period t; This represents the total load in time period t; For example: If =20kW, =10kW, =30kW, then =60kW.

[0055] Step 104: Based on the preset operational goals and constraints, use the structured feature dataset and the virtual power plant knowledge graph to generate multiple candidate operational schemes and reference information for each candidate operational scheme.

[0056] In this embodiment of the application, the reference information includes scheme score, key factor contribution, confidence interval, and scheme risk warning.

[0057] The scheme score refers to a quantitative score obtained after a multi-dimensional comprehensive evaluation of candidate operational schemes. It is used to assist decision-makers in quickly identifying and comparing the merits of different candidate operational schemes. The scheme score is calculated in the following way: Scheme score = Overall score = w1 × Benefit score + w2 × Stability score + w3 × Carbon efficiency score + w4 × Compliance score; Where w1 is the preset weight of the return score, w2 is the preset weight of the stability score, w3 is the preset weight of the carbon efficiency score, and w4 is the preset weight of the compliance score.

[0058] The purpose of the evaluation of the proposed solutions includes: 1) Quantitative evaluation: The economic efficiency, safety, carbon efficiency, compliance, and other dimensions of candidate operating solutions are quantified into a unified score, facilitating quick comparison by decision-makers; 2) Ranking and Filtering: Based on the scheme scores, multiple candidate operation schemes are ranked to identify the scheme with the best overall performance; 3) Decision support: Decision-makers can combine information such as scheme scores, key factor contributions, confidence intervals, and risk warnings to select the optimal operational scheme from multiple candidate operational schemes; 4) Visual presentation: The scheme score, as one of the core indicators of the visual evaluation chart, is presented intuitively in the form of radar charts, bar charts, etc., reducing the difficulty of understanding for decision-makers.

[0059] The score ranges from [0,1], with a higher score indicating a better overall performance. The score levels are: Excellent (0.9-1.0), Good (0.8-0.9), Average (0.7-0.8), Poor (0.6-0.7), and Unsatisfactory (0.0-0.6).

[0060] As an optional implementation, step 104, based on preset operational goals and constraints, uses the structured feature dataset and the virtual power plant knowledge graph to generate multiple candidate operational schemes and reference information for each candidate operational scheme, may include the following: Based on the preset operational objectives, rule reasoning is performed using the rule data and the meteorological forecast data to obtain the first candidate operational plan; Multiple historical schemes are obtained from the historical data; Determine the similarity between the stated operational objectives and various historical schemes; The historical solution with the highest similarity was selected as the second candidate operation solution. Based on the operational objectives and constraints, a third candidate operational plan is generated using the virtual power plant knowledge graph. The first candidate operation plan, the second candidate operation plan, and the third candidate operation plan are analyzed respectively to obtain reference information for each candidate operation plan.

[0061] This implementation method generates a first candidate operational plan based on rules and meteorological data through rule-based reasoning, providing a reasonable solution using established rules and meteorological trends. A second candidate plan, selected from historical data with the highest similarity to the operational objectives, leverages past successes. A third candidate plan is generated using a knowledge graph, uncovering deep data relationships. Finally, analysis of the three candidate plans yields reference information, generating candidate plans from multiple perspectives and dimensions and providing comprehensive reference. This effectively avoids the limitations of a single plan, enhances the diversity and comprehensiveness of the plans, and provides solid support for selecting the optimal operational plan.

[0062] In this embodiment of the application, the rule data may include scheduling rule data, market rule data, regulatory compliance rule data, and equipment operation rule data, specifically: 1. Scheduling rule data includes: (1) Power change rate limit rule: The maximum allowable value of power change per unit time (ramp / ramp rate) is specified, for example, the power change per hour shall not exceed 20% of the rated capacity; (2) Output planning rules: specify the output planning requirements for photovoltaic, energy storage and load in each time period; (3) Frequency adjustment rules: Specify the system frequency deviation range (e.g., ±0.2Hz) and the corresponding response strategy; (4) Voltage regulation rules: Specify the voltage deviation range (e.g., ±5%) and the corresponding adjustment measures.

[0063] 2. Market rule data includes: (1) Electricity pricing rules: Time-of-use pricing, peak-valley pricing, real-time pricing and other pricing mechanisms; (2) Trading rules: electricity trading time window, application rules, and settlement rules; (3) Demand response rules: conditions for participation in demand response, response capacity requirements, and response time requirements; (4) Ancillary service rules: rules for participation and compensation standards for ancillary services such as frequency regulation, voltage regulation, and standby.

[0064] 3. Regulatory compliance data includes: (1) Carbon emission rules: carbon emission limits, carbon efficiency standards (e.g., carbon efficiency <0.6 is considered compliant); (2) Safe operation rules: safe operating range of equipment and emergency response requirements; (3) Measurement reporting rules: data collection frequency, reporting format, and reporting time requirements.

[0065] 4. Equipment operation rule data includes: (1) SOC operation rules: safe operating range of SOC of energy storage system (e.g., 20%-90%). (2) Power limiting rules: Maximum / minimum output limits for each device; (3) Start-up and shutdown rules: restrictions on the number of times the equipment can be started and stopped, and minimum operating time requirements; (4) Lifetime protection rules: Lifetime protection strategy based on cycle depth, temperature, and C-rate.

[0066] In this embodiment of the application, the meteorological forecast data may include solar radiation data, temperature data, wind speed and wind direction data, and other meteorological elements, specifically: 1. Solar radiation data includes: (1) Total radiation intensity (GHI): Total radiation intensity on the horizontal plane (W / m²) 2 ), used to predict photovoltaic power generation; (2) Direct radiation intensity (DNI): Direct normal radiation intensity (W / m 2 ); (3) Diffuse radiation intensity (DHI): Diffuse radiation intensity (W / m 2 ); (4) Radiation prediction time series: the predicted value of radiation intensity for the next 24 hours or longer, with a time resolution of 15 minutes or 1 hour.

[0067] 2. Temperature data includes: (1) Ambient temperature: Ambient air temperature (°C) affects the efficiency of photovoltaic modules and the performance of energy storage systems; (2) Module temperature: The surface temperature (°C) of photovoltaic modules directly affects power generation efficiency; (3) Temperature forecast time series: temperature forecast values ​​for future time periods.

[0068] 3. Wind speed and direction data include: (1) Wind speed: Wind speed (m / s) affects component heat dissipation and system stability; (2) Wind direction: The wind direction angle (degrees) affects the temperature distribution of the components; (3) Wind speed / direction forecast time series: wind speed and direction forecast values ​​for future time periods.

[0069] 4. Other meteorological elements include: (1) Humidity: Relative humidity (%) affects the cleanliness of the component surface and heat dissipation; (2) Air pressure: Atmospheric pressure (hPa), which affects the system's operating environment; (3) Visibility: Visibility (km) indirectly reflects atmospheric transparency.

[0070] In this embodiment of the application, the meteorological forecast data can be sourced from the following four sources: 1. Weather forecast data released by the meteorological bureau; 2. Detailed forecast data provided by professional meteorological service providers; 3. Prediction results based on historical meteorological data and numerical weather prediction models; 4. Real-time monitoring data from local weather stations combined with forecasting models.

[0071] In this embodiment of the application, the method for generating the first candidate solution through rule-based reasoning is as follows: 1. Rule matching stage: 1.1 Scene Recognition: Based on the current operational objectives (such as minimizing the total cost of electricity purchase and maximizing system stability) and meteorological forecast data (such as solar radiation intensity and temperature in the next 24 hours), identify the current operational scene type (such as sunny high radiation scene, cloudy low radiation scene, extreme weather scene, etc.).

[0072] 1.2 Rule Activation: Match a set of rules applicable to the current scenario from the rule database, including: (1) Scheduling rules: Match the corresponding scheduling strategy rules according to the photovoltaic power generation forecast based on the weather forecast; (2) Market rules: Match the power purchase and sale strategy rules according to the current electricity price rules; (3) Equipment rules: Match the equipment operation rules according to the current status of the equipment.

[0073] 2. Rule-based reasoning stage: Employing either forward chaining or backward chaining mechanisms: For example, the process using the forward reasoning mechanism can be specifically as follows: Input facts: Current system status (SOC, load, electricity price, etc.) + meteorological forecast data (radiation, temperature, etc. for the next 24 hours); Rule matching: Find rules that satisfy all preconditions; Rule execution: Execute the matching rules and generate new decision facts (such as "During the period from 10:00 to 14:00, photovoltaic power generation is sufficient, and photovoltaic power should be used first, while energy storage charging should be prioritized"). Iterative reasoning: Take newly generated facts as input, continue to match and execute rules until no new facts can be generated or the preset reasoning depth is reached; Solution generation: Summarize all reasoning results to form a complete operation plan.

[0074] 3. Solution optimization phase: Constraint check: Check whether the inference-generated solution meets all operational constraints (equipment capacity, SOC range, load balancing, etc.). Target evaluation: Calculate the economic target (total electricity purchase cost) and safety target (system stability index) of the scheme; Solution adjustment: For solutions that do not meet the constraints, adjustments are made according to the rule priority; for solutions that meet the constraints but have unsatisfactory target values, local optimization is performed.

[0075] 4. Solution Output Stage: Convert the reasoned and optimized solutions into a standard format, including decision variable values ​​(photovoltaic power, energy storage power, purchased power, etc.) for each time period and solution evaluation indicators.

[0076] In this embodiment of the application, the specific content of the first candidate solution may include: time series decision variables, solution evaluation indicators, solution execution strategy description, and solution constraint satisfaction status, specifically: 1. Time series decision variables include: Photovoltaic power generation sequence: P_PV(t), t=1,2,...,T, representing the planned power generation (kW) of photovoltaic equipment in each time period; Energy storage charge and discharge power sequence: P_ESS(t), t=1,2,...,T, positive values ​​represent charging, negative values ​​represent discharging (kW); Energy storage SOC sequence: SOC(t), t=1,2,...,T, represents the state of charge (%) of the energy storage system at each time period; Power purchased from the grid sequence: P_Grid(t), t=1,2,...,T, representing the power (kW) purchased from the grid in each time period; Load power sequence: P_Load(t), t=1,2,...,T, representing the predicted load (kW) for each time period.

[0077] 2. The evaluation indicators for the plan include: (1) Economic indicators: Total Cost of Electricity Purchase: The total cost of electricity purchase (in yuan) within the preset time period; Revenue from electricity sales (if any): Revenue from selling electricity to the grid (in RMB). Net cost: Electricity purchase cost - Electricity sales revenue (RMB); (2) Safety indicators: System stability index (Stability): a weighted sum of voltage and frequency deviations; Maximum voltage deviation: The maximum value of the voltage deviation at each time period; Maximum frequency deviation: The maximum value of the frequency deviation in each time period; (3) Equipment status indicators: SOC minimum, maximum, and average values; Energy storage charge / discharge cycles; Equipment utilization rate.

[0078] 3. The implementation strategy description includes: Time-segmentation strategy: Explain the operational strategies for different time periods (such as peak hours, normal hours, and off-peak hours); Charge and discharge strategy: Explain the timing of energy storage charging and discharging and the basis for power selection; Electricity purchase strategy: Explain the timing and basis for selecting power sources when purchasing electricity from the grid; Risk response strategy: Explain the strategies for dealing with abnormal situations such as weather forecast deviations and equipment failures.

[0079] 4. The constraints of the proposed solution are satisfied, including: Equipment capacity constraint compliance: Whether the photovoltaic power and energy storage power are within the allowable range for each time period; SOC constraint satisfaction status: Whether the SOC in each time period is within the safe range (20%-90%). Load balance constraint satisfaction: Whether the power balance equation is satisfied in each time period.

[0080] In this embodiment of the application, the method for determining the similarity between the operational objective and each historical solution can be to retrieve historical cases using cosine similarity, specifically:

[0081] Where A and B represent two feature vectors (such as current load, price, etc.); sim(A,B) represents the similarity score.

[0082] For example, if A=[1,2,3] and B=[2,4,6], then sim=1, indicating perfect similarity. The similarity between the current scene and historical cases is calculated; the closer to 1, the more similar they are.

[0083] In this embodiment of the application, the specific content included in the second candidate operation plan may be: time series decision variables, historical plan metadata, similarity information, plan evaluation indicators, plan applicability analysis, and plan adjustment suggestions, specifically: 1) Time series decision variables (with the same structure as the first candidate) include: Photovoltaic power generation sequence: P_PV(t), t=1,2,...,T (kW); Energy storage charge and discharge power sequence: P_ESS(t), t=1,2,...,T (kW); Energy storage SOC sequence: SOC(t), t=1,2,...,T (%); Power purchase sequence from the power grid: P_Grid(t), t=1,2,...,T (kW); Load power sequence: P_Load(t), t=1,2,...,T (kW).

[0084] 2) Historical scheme metadata includes: Historical Scheme ID: A unique identifier for a historical scheme; Historical plan execution date: The actual date and time that the historical plan was implemented; Historical load characteristics: average load, peak load, load curve shape characteristics, etc.; Historical price characteristics: average electricity price, peak-valley electricity price difference, price fluctuation characteristics, etc. Historical meteorological characteristics: average radiation intensity, temperature range, weather type, etc.

[0085] 3) Similarity information includes: Similarity score: The cosine similarity value between the current scene and the historical solutions (between 0 and 1, the closer to 1, the more similar). Description of similar features: Explain which features are similar (e.g., similar load curves, similar price patterns, similar weather conditions, etc.). Description of differences: Explain which features differ and how these differences may affect the effectiveness of the solution.

[0086] 4) Evaluation indicators for the program (based on historical implementation results) include: Historical economic indicators: actual electricity purchase cost and electricity sales revenue during the implementation of historical plans; Historical security metrics: Actual system stability metrics during the execution of historical solutions; Historical equipment status indicators: the equipment operating status when historical plans were executed.

[0087] 5) The applicability analysis of the proposed solution includes: Applicability assessment: Analyze the applicability of this historical solution in the current scenario; Aspects requiring adjustment: Identify decision variables that may need adjustment due to differences in scenarios; Risk warning: Based on abnormal events during historical execution, this section highlights the risks that may be encountered when using this solution.

[0088] 6) Suggestions for adjusting the plan include: Direct usage suggestion: If the similarity is very high (>0.9), it is recommended to use this historical solution directly; Recommended adjustments: If the similarity is moderate (0.7-0.9), it is recommended to adjust some time periods or some decision variables. Suggestions for reference: If the similarity is low (<0.7), it is recommended to use it only as a reference and focus on learning from its strategic ideas rather than specific numerical values.

[0089] In this embodiment of the application, the method for generating a third candidate scheme based on a knowledge graph is as follows: 1. Knowledge Graph Query: (1) Target node localization: Economic target node: Minimize total electricity purchase cost; Security objective: Maximize system stability; These nodes are typically associated with the "Operational Target" entity.

[0090] (2) Constraint node positioning: Equipment capacity constraint nodes: maximum photovoltaic power, maximum energy storage power, etc.; SOC constraint nodes: SOC upper and lower limits; Load balance constraint node: power balance equation; These nodes are typically associated with "operational constraint" entities.

[0091] (3) Related entity query: Search for historical cases, optimization strategies, and equipment configurations related to the goal of "minimizing total electricity purchase cost"; Query the operating rules, security policies, historical cases, etc. related to "SOC constraints"; Query the topological relationships and influence relationships between devices.

[0092] 2. Knowledge Reasoning: (1) Path reasoning: Starting from the operational goal node, traverse along the relationship edges (such as "achievement", "impact", "association" etc.) to the relevant strategy nodes, case nodes, and rule nodes; Starting from the operational constraint node, it traverses to the relevant equipment node, parameter node, and historical case node; Find the optimal path connecting the objective and the constraints; the nodes and edges on this path constitute the knowledge base for generating the solution. (2) Pattern matching: Match subgraph patterns in the knowledge graph that are similar to the current scene; For example: match the scenario pattern of "high radiation + low electricity price + high load" and find historical success cases and optimization strategies under this pattern; (3) Knowledge integration: The knowledge extracted from different paths and patterns will be integrated; For example, it integrates knowledge from multiple aspects such as "economic optimization strategies", "safety assurance strategies", and "equipment operation experience".

[0093] 3. Solution Generation: (1) Strategy extraction: Extract decision-making patterns from historical success stories; Extract optimization strategies from the rule base; Extract best practices from equipment operation experience.

[0094] (2) Parametric reasoning: Using the "device-parameter-value" triple, infer the equipment's operating parameters; Use the "scenario-strategy-effect" triple to select the optimal strategy parameters; Utilize graph algorithms (such as PageRank, community detection, etc.) to identify key influencing factors and adjust parameters accordingly.

[0095] (3) Constraints are satisfied: Use the constraints in the knowledge graph to check whether the solution violates the constraints; If a constraint is violated, search the knowledge graph for constraint adjustment strategies or alternatives.

[0096] (4) Target optimization: By leveraging the "goal-strategy-effect" relationship, we can select a better combination of strategies. By utilizing historical case study data, adjust the solution parameters to improve the target value.

[0097] In this embodiment of the application, the specific content of the third candidate scheme may include: time series decision variables, knowledge reasoning path, scheme evaluation index, and scheme interpretability information, specifically: 1. Time series decision variables: Photovoltaic power generation sequence: P_PV(t), t=1,2,...,T (kW); Energy storage charge and discharge power sequence: P_ESS(t), t=1,2,...,T (kW); Energy storage SOC sequence: SOC(t), t=1,2,...,T (%); Power purchase sequence from the power grid: P_Grid(t), t=1,2,...,T (kW); Load power sequence: P_Load(t), t=1,2,...,T (kW).

[0098] 2. Knowledge Reasoning Path: Knowledge graph nodes used: List the main knowledge graph nodes used to generate this solution (such as device nodes, policy nodes, case nodes, etc.). Knowledge graph relationships used: List the main relationships used to generate this solution (such as "realization", "impact", "association" etc.); Reasoning path: Describes the reasoning path from the target node to the solution node, including the sequence of nodes and edges traversed; Knowledge integration method: Explain how to integrate multiple knowledge sources (rules, cases, experience, etc.) to generate a solution.

[0099] 3. Project evaluation indicators: Predicting economic indicators: Based on historical performance data in the knowledge graph, predict the economic indicators of the plan; Predicting security metrics: Based on historical performance data in the knowledge graph, predict the security metrics of the solution; Knowledge confidence: The knowledge confidence of a solution is evaluated based on the reliability and relevance of knowledge in the knowledge graph.

[0100] 4. Explanation information of the solution: Decision rationale: Based on the causal relationships in the knowledge graph, explain the rationale for each decision; Influencing Factor Analysis: Based on the influence relationships in the knowledge graph, analyze the key factors affecting the effectiveness of the solution; Risk Source Analysis: Based on the risk relationships in the knowledge graph, analyze the potential sources of risk that the solution may face.

[0101] In this embodiment of the application, the first candidate operation plan, the second candidate operation plan, and the third candidate operation plan are analyzed respectively, and the reference information for each candidate operation plan can be specifically as follows: 1. Scheme scoring generation: The specific scoring method for the proposed solutions is as follows: (1) Objective function normalization: Normalize objective functions with different dimensions to the interval [0,1]; ensure that the objective functions are comparable; (2) Weight allocation: Weights are allocated according to the importance of the business; dynamic weight adjustment is supported; (3) Weighted summation: Calculate the weighted comprehensive score; consider risk penalty items; (4) Standardization of scoring: Limit the scoring of the schemes to the range of [0,1]; ensure the consistency and comparability of the scoring of the schemes.

[0102] The role of alternative rating in decision support: Quantitative comparison: The multi-dimensional performance of candidate operation plans is quantified into a unified score, which makes it easier for decision-makers to quickly compare different plans; Ranking and filtering: Based on the scheme score, the candidate operation schemes are ranked and the scheme with the best overall performance is identified; Decision-making basis: Decision-makers can make scientific decisions by combining information such as scheme scores, key factor contributions, confidence intervals, and risk warnings; Visual presentation: The solution score, as the core indicator of the visual evaluation chart, is presented intuitively in the form of radar charts, bar charts, etc.

[0103] Calculation formula: Normalization of economic objectives: Economic efficiency score = 1 - (actual cost - minimum cost) / (maximum cost - minimum cost) Security target normalization: Safety score = 1 - (Actual deviation - Minimum deviation) / (Maximum deviation - Minimum deviation) Scheme scoring calculation: Solution score = w5 × economic score + w6 × safety score Where w5 and w6 are the weights of each objective.

[0104] In this embodiment of the application, the actual cost refers to the total electricity cost incurred by purchasing electricity from the power grid within a preset time period after the candidate operation plan is executed in the current scenario; the actual cost is calculated based on the power purchase sequence P_Grid(t) and the electricity price sequence Price(t) in the candidate operation plan.

[0105] P_Grid(t): Decision variable from candidate operating schemes, representing the power (kW) purchased from the grid in each time period; Price(t): Sourced from market price data, representing the electricity price (yuan / kWh) for each time period, which can be time-of-use pricing, real-time pricing, etc. Actual cost calculation method: Actual cost = Σ(t=1 to T) [P_Grid(t) × Price(t) × Δt]; in: T: The total number of time periods within the preset time period (e.g., 24 hours, with one time period per hour, then T=24). P_Grid(t): Power purchased from the grid in time period t (kW); Price(t): Electricity price for time period t (yuan / kWh); Δt: The duration of each time period (in hours). If there is one time period per hour, then Δt = 1. Calculation example: Assuming the preset time period is one day (24 hours), with one hour as one time period: Hour 1: P_Grid(1) = 10 kW, Price(1) = 1.0 yuan / kWh Hour 2: P_Grid(2) = 8 kW, Price(2) = 1.2 yuan / kWh ... (other time periods) The actual cost is calculated as follows: 10 × 1.0 × 1 + 8 × 1.2 × 1 + 5 × 0.8 × 1 + ... = 10 + 9.6 + 4 + ... = Total cost (yuan).

[0106] In this embodiment of the application, the minimum cost can be obtained in the following three ways: Method 1: Theoretical optimization calculation: Solve the following optimization problems using optimization algorithms (such as linear programming, mixed integer programming, etc.): (1) Objective function: minimize total electricity purchase cost; (2) Constraints: Equipment capacity constraints, SOC constraints, load balancing constraints, etc.; (3) The cost value corresponding to the theoretical optimal solution is obtained by solving the problem, which is the minimum cost.

[0107] Method 2: Historical Best Value (1) Find the lowest cost value actually achieved under similar scenarios (similar load, similar price, similar weather conditions) from historical operation data; (2) Take the historical best cost as the reference value for the minimum cost.

[0108] Minimum cost calculation method: (1) Establish the optimization model (minimize: TotalCost = Σ(t=1 to T) [P_Grid(t) ×Price(t) × Δt]); (2) Solve the optimization problem: Use an optimization solver (such as CPLEX, Gurobi, or an open-source solver) to solve the above model; (3) Obtain the minimum cost: The optimal objective function value obtained by optimization is the minimum cost.

[0109] In this embodiment of the application, the maximum cost can be determined by finding the highest cost value actually achieved in similar scenarios from historical operational data as the maximum cost.

[0110] In this embodiment of the application, the formula for calculating the actual deviation can be: Actual deviation = Σ(t=1 to T) [w_V × |V(t) - V_ref| + w_f × |f(t) - f_ref|]; Where T represents the total number of time periods within the preset time period; V(t) represents the actual voltage (V) in time period t; V_ref represents the voltage reference value (V); f(t) represents the actual frequency (Hz) in time period t; f_ref represents the frequency reference value (Hz); w_V represents the voltage deviation weighting coefficient; and w_f represents the frequency deviation weighting coefficient.

[0111] In this embodiment, the minimum deviation can be the lowest deviation value actually achieved under similar scenarios, found from historical operational data. The maximum deviation can be the highest deviation value actually achieved under similar scenarios, found from historical operational data.

[0112] Please also refer to Table 1, which is the rating scale for the proposed solutions: Table 1. Scheme Evaluation Level Classification Table

[0113] 2. Key Factor Contribution Generation: Sensitivity analysis methods: Parameter perturbation: Perturb each decision variable by ±5% and calculate the rate of change of the objective function.

[0114] Contribution calculation: Calculate the degree of influence of each factor on the objective function and identify the key factors with the greatest influence.

[0115] Contribution ranking: Sort by contribution level and output Top-K key factors.

[0116] Calculation formula: Key factor contribution: Contribution = Change in objective function / Change in parameters.

[0117] Photovoltaic power contribution: PV contribution = |Objective function(PV+5%) - Objective function(PV-5%)| / (2×5%).

[0118] Energy storage SOC contribution: SOC contribution = |Objective function(SOC+5%) - Objective function(SOC-5%)| / (2×5%).

[0119] 3. Confidence interval generation: Monte Carlo simulation method: Parameter perturbation: Randomly sample the input parameters to account for parameter uncertainty.

[0120] Multiple optimizations: Run the optimization algorithm multiple times using the perturbed parameters and collect the optimization results.

[0121] Statistical analysis: Calculate the mean and standard deviation of the results, and determine the confidence interval.

[0122] Calculation formula: Disturbance parameter = original parameter × (1 + random noise).

[0123] Confidence interval: Confidence interval = [mean - t × standard deviation, mean + t × standard deviation]; Wherein: the t-value is determined based on the confidence level and degrees of freedom; at a 95% confidence level, the t-value is approximately 1.96.

[0124] 4. Generation of solution risk warnings: 4.1) Risk Identification and Quantification: Risk factor identification: Identify potential risk factors and create a list of risk factors.

[0125] Risk quantification: Calculate the probability and impact of risks to generate risk scores.

[0126] Risk level classification: Based on the risk score, risk levels are classified and risk warning information is generated.

[0127] Calculation formula: SOC risk = 1 - min(lower distance limit, upper distance limit) / (upper limit - lower limit).

[0128] Voltage deviation risk: Voltage risk = Maximum voltage deviation / Maximum allowable deviation.

[0129] Economic risks: Economic risk = (actual cost - budgeted cost) / budgeted cost.

[0130] Please also refer to Table 2, which is the risk level classification table: Table 2 Risk Level Classification Table

[0131] 4.2) Risk Score of the Solution Calculation formula: Solution risk score = w7 × SOC risk + w8 × voltage risk + w9 × economic risk Where w7, w8, and w9 are the risk weights; the sum of the weights is 1.

[0132] Step 105: Output the visual evaluation charts of each candidate operation plan, and obtain the optimal operation plan determined by the decision-makers of the virtual power plant from multiple candidate operation plans based on the visual evaluation charts.

[0133] In this embodiment of the application, the visual evaluation chart may include a multi-dimensional scoring radar chart, a comprehensive scoring comparison bar chart, and a time series comparison chart, etc., specifically: 1. Multidimensional Scoring Radar Chart: Chart content: Displays the scores of each candidate operation plan in four dimensions: revenue, stability, carbon efficiency, and compliance; Chart format: Radar chart, with each option represented by a line and four dimensions as four axes; Purpose: To intuitively compare the performance of each solution across different dimensions and identify the advantages and disadvantages of each solution.

[0134] 2. Overall score comparison bar chart: Chart content: Displays the overall score of each candidate operational plan; Chart format: bar chart, with the horizontal axis representing the scheme name and the vertical axis representing the overall score; Function: Quickly identify the solution with the highest overall score.

[0135] 3. Time series comparison chart: Chart content: Shows the curves of key decision variables for each candidate operational plan changing over time; Photovoltaic power generation time series; Energy storage charging and discharging power time series; Energy storage SOC time series; Time series of power purchases by the power grid; Chart format: Line chart, with multiple options represented by lines of different colors; Purpose: To compare the differences in the operational strategies of various schemes and identify the execution characteristics of the schemes.

[0136] In this embodiment of the application, the method for determining the optimal operation plan from multiple candidate operation plans can be as follows: 1. Preliminary screening stage: Compliance screening: First, based on the multi-dimensional scoring radar chart, all candidate operation plans with a compliance score <0.7 or with serious violations are excluded, and candidate operation plans with a compliance score ≥0.7 are identified as the first screening plan; candidate operation plans with a compliance score ≥0.7 but <0.9 are marked as requiring attention. Risk screening: Based on the risk score of the plan, candidate operation plans with extremely high risk (risk score ≥ 0.8) are excluded, and candidate operation plans with risk score < 0.8 are determined as the second screening plan; for high-risk plans (risk score 0.6-0.8), they are marked as requiring careful consideration. Preliminary screening based on comprehensive scores: Review the comparative bar chart of comprehensive scores, identify candidate operation plans with higher comprehensive scores (e.g., scores > 0.8) and determine them as the third screening plan; exclude candidate operation plans with excessively low comprehensive scores (e.g., scores < 0.6). The first, second, and third screening schemes were selected as target candidate operation schemes.

[0137] 2. In-depth evaluation phase: Uncertainty analysis: Based on the obtained confidence intervals, compare the confidence interval widths of the key indicators of each target candidate operation plan (plans with narrow confidence intervals have less uncertainty and are more reliable; plans with wide confidence intervals have greater uncertainty and require caution); prioritize the first plan with a narrow confidence interval and high reliability; Key factor analysis: Based on the contribution of key factors, analyze the key driving factors of each candidate operational plan; identify the key factors that have the greatest impact on the effectiveness of the plan; evaluate the controllability and stability of these key factors; and select the second plan with more controllable and stable key factors. Comprehensive evaluation and decision-making: Based on the analysis results of the first and second schemes, a comprehensive evaluation of each target candidate operation scheme is conducted; a scoring card method can be used to score and weight each indicator of each scheme; finally, the target candidate operation scheme with the best comprehensive evaluation is selected as the optimal operation scheme.

[0138] In this embodiment, revenue is obtained by calculating the net revenue of the virtual power plant within a preset time period (e.g., one day, one week, one month). Net revenue includes electricity sales revenue, demand response revenue, ancillary service revenue, etc., minus the total cost of electricity purchase, equipment operating costs, maintenance costs, etc.

[0139] Wherein, electricity sales revenue = Σ(photovoltaic power generation (t) × electricity sales price (t) + energy storage discharge power (t) × electricity sales price (t)), where t is the time period index; Demand response revenue = Σ(demand response power (t) × demand response price (t)); Ancillary service revenue = Σ(Ancillary service power (t) × Ancillary service price (t)); Equipment operating cost = fixed operating cost + variable operating cost (related to power); Maintenance cost = Fixed cost of equipment maintenance.

[0140] In this embodiment, the stability deviation is calculated by real-time collection of voltage V(t) and frequency f(t) data of the system at various time periods through IoT devices (such as smart meters, PMUs, etc.) and comparison with reference values.

[0141] In this embodiment, carbon efficiency is assessed by calculating the carbon emission intensity of a virtual power plant over a preset time period. The carbon efficiency value reflects the carbon emissions per unit of electricity generated or per unit of revenue.

[0142] Carbon efficiency can be calculated as follows: Carbon efficiency = 1 - (actual carbon emissions - minimum carbon emissions) / (maximum carbon emissions - minimum carbon emissions); in: Actual carbon emissions = Σ(Photovoltaic power generation (t) × Photovoltaic carbon emission coefficient + Energy storage charging and discharging power (t) × Energy storage carbon emission coefficient + Power purchased by the grid (t) × Grid carbon emission coefficient); Photovoltaic carbon emission factor: The carbon emission factor of photovoltaic power generation (usually close to 0, because photovoltaic power generation produces virtually no carbon emissions). Energy storage carbon emission coefficient: The carbon emission coefficient of the energy storage system during the charging and discharging process (usually small, mainly considering the indirect carbon emissions corresponding to the loss of charging and discharging efficiency). Carbon emission factor of power grid: Carbon emission factor of electricity purchased from the power grid (determined according to the energy structure of the power grid, for example 0.5-0.8 kg CO2 / kWh); Minimum carbon emissions: The theoretically achievable minimum carbon emissions (usually assuming all electricity is generated by photovoltaics, the minimum carbon emissions are close to 0). Maximum carbon emissions: The theoretically achievable maximum carbon emissions (usually assuming all electricity is purchased from the grid, maximum carbon emissions = total load × grid carbon emission coefficient).

[0143] In this embodiment, compliance data originates from the results of compliance checks on the operational plan. The operational plan is checked item by item using a compliance rule base, and the pass rate of each compliance check is statistically analyzed to generate compliance data. Compliance data includes the number of regulatory approvals, scheduling approvals, market approvals, the total number of compliance items, and the total score for violation penalties.

[0144] In this embodiment of the application, the specific scores for each candidate operating scheme in the four dimensions of revenue, stability, carbon efficiency, and compliance can be as follows: (1) Profitability score: Profit score = (actual profit - minimum profit) / (maximum profit - minimum profit), ensuring the result is in the range [0,1].

[0145] (2) Stability score: Stability score = 1 - (actual deviation - minimum deviation) / (maximum deviation - minimum deviation), ensuring the result is in the range [0,1].

[0146] (3) Carbon efficiency score: Directly use the input carbon efficiency value (0-1 range).

[0147] (4) Compliance score: The compliance score calculation uses a normalization method to unify the pass rate of compliance items and the penalty items for violations into the [0,1] interval: Compliance score = Compliance pass rate × (1 - Normalized total score of violation penalties); in: Compliance pass rate = (Number of regulations passed + Number of scheduling passed + Number of market passed) / Total number of compliance items; The pass rate for compliance items ranges from [0,1]. Normalized total score for violation penalties = min(total score for violation penalties / maximum possible penalty score, 1.0); The normalized total score for violation penalties ranges from [0,1]. Maximum possible penalty score = total number of compliance items × maximum penalty score for a single item; The maximum penalty score for a single item is usually 1.0.

[0148] In this embodiment of the application, the total number of compliance items refers to the total number of all compliance check items defined in the compliance rule base, including regulatory compliance check items, scheduling procedure check items, market rule check items, etc.

[0149] The number of regulations passed refers to the number of items that pass the regulatory compliance check, which is obtained by performing regulatory compliance checks on the operation plan.

[0150] The number of items that pass the scheduling procedure check refers to the number of items that pass the check in the scheduling procedure check, which is obtained by performing the scheduling procedure check on the operation plan.

[0151] Market pass count refers to the number of items that pass the market rule check, which is obtained by performing market rule checks on the operation plan.

[0152] The compliance pass rate refers to the proportion of compliance items that pass the inspection out of the total number of compliance items, reflecting the overall compliance level of the operation plan.

[0153] The total score for violation penalties refers to the sum of the preset penalty scores for all violations, reflecting the severity of the operational plan's violation of compliance requirements.

[0154] In this embodiment of the application, the comprehensive score of each candidate operation plan can be calculated as follows: Overall score = w10 × profitability score + w11 × stability score + w12 × carbon efficiency score + w13 × compliance score - total penalty score; Where: w10+w11+w12+w13=1, and the total score of the penalty items is ≤1.0.

[0155] In this embodiment of the application, the comprehensive scoring level is divided as follows: Excellent: 0.9-1.0, Good: 0.8-0.9, Average: 0.7-0.8, Poor: 0.6-0.7, Unsatisfactory: 0.0-0.6.

[0156] As an optional implementation, after step 105, the following steps may also be performed: The optimal operation plan was simulated and verified using a power simulation platform to obtain the actual load; The target deviation value is obtained by comparing the predicted load based on the optimal operation plan with the actual load. If the target deviation value is greater than the preset deviation threshold, the optimal operation plan is analyzed to determine the cause of the deviation; Based on the target deviation value and the cause of the deviation, the virtual power plant knowledge graph is written back to obtain an updated virtual power plant knowledge graph.

[0157] This implementation method uses a power simulation platform to simulate and verify the optimal operation plan, obtaining actual load and comparing it with the predicted load to determine the target deviation value. This allows for accurate evaluation of the plan's actual effectiveness. If the deviation value exceeds a threshold, in-depth analysis can determine the cause of the deviation and identify potential problems with the plan. Based on the deviation value and its cause, knowledge is written back to the virtual power plant knowledge graph, allowing actual operational experience to be fed back into the knowledge system and enabling dynamic updates to the knowledge graph. In this way, the accuracy and practicality of the knowledge graph can be continuously improved, providing a more reliable basis for subsequent operational decisions and enhancing the scientific rigor and adaptability of virtual power plant operation decisions.

[0158] For example, a scenario: A virtual power plant receives a demand response event from a power company and needs to formulate a response strategy within 10 minutes, taking into account both the response effect and the lifespan of the equipment.

[0159] Implementation steps: 1) The system quickly identifies DR events and initiates the "rapid decision" mode.

[0160] 2) Quickly retrieve historical cases based on features such as similar load, price, and time, and output "template + parameterized suggestions".

[0161] 3) Generate multiple response strategies based on the current SOC status and equipment availability, and evaluate the response effectiveness, equipment lifespan impact, and compliance risks of each strategy.

[0162] 4) Based on the risk assessment results, the operation and maintenance personnel select the optimal strategy and distribute the response strategy to the energy storage system and controllable load.

[0163] 5) Track response effects in real time, record deviations and experiences, and write back to the knowledge graph.

[0164] Results: Response time was reduced by 60%, equipment lifespan was reduced by 25%, and fast and reliable decision-making was achieved.

[0165] As can be seen, the embodiments of this application, through precise scheduling, reduce peak grid load, improve economic efficiency and system security, increase user benefits, and make the system operation more intelligent. Specifically: 1. Unified semantic foundation: It unifies the expression of scattered multi-source data and expert knowledge in the knowledge graph, realizes cross-source reasoning and knowledge association, breaks down information silos, and improves data utilization.

[0166] 2. Multiple candidates and interpretability: Upgrading from "giving an answer" to "giving options + reasons + confidence level", significantly improving decision-making quality and efficiency, and supporting multi-objective trade-offs and risk assessment.

[0167] 3. Human-machine collaborative closed loop: Through mechanisms such as What-if analysis, solution comparison, and debriefing, human-machine collaborative decision-making is achieved. The system continues to evolve with use, improving the level of intelligent decision-making.

[0168] 4. Multi-objective integration: Achieve synergistic optimization among objectives such as economy, safety, compliance, and carbon efficiency, support scenario-based weight configuration and strategy templates, and improve the comprehensiveness of decision-making.

[0169] 5. Auditable and compliant: Establish a full-chain traceability mechanism to meet internal audit and external regulatory requirements, support the division of responsibilities and compliance verification, and enhance system credibility.

[0170] 6. Continuous self-learning: Through execution feedback and experience accumulation, the knowledge base is continuously updated and optimized to improve decision-making accuracy and adapt to the dynamically changing operating environment.

[0171] 7. Strong applicability: It adopts standardized interfaces and modular design, making it easy to integrate into existing virtual power plant systems. It supports distributed deployment and horizontal expansion, and has broad application prospects.

[0172] Implementing steps 101 to 105 above enables decision-makers to make scientific judgments based on comprehensive and accurate information, effectively avoiding decision-making errors caused by incomplete information or misunderstandings, thereby improving the accuracy of the final selected operating plan. Furthermore, this application can improve the reliability and accuracy of virtual power plant operation auxiliary decision-making. In addition, this application can enhance the intelligence level of virtual power plant operation. Furthermore, this application can improve the scientific rigor and rationality of the plan, ensuring the efficient and stable operation of the virtual power plant. Furthermore, this application can effectively avoid the limitations of a single plan, enhancing the diversity and comprehensiveness of plans, providing solid support for selecting the optimal operating plan. In addition, this application can continuously improve the accuracy and practicality of the knowledge graph, providing a more reliable basis for subsequent operational decisions.

[0173] Based on the same inventive concept, this application also provides a virtual power plant operation auxiliary decision-making device for implementing the above-mentioned auxiliary decision-making method for virtual power plant operation. The solution provided by this device is similar to the solution described in the above-described method. Therefore, the specific limitations of one or more virtual power plant operation auxiliary decision-making device embodiments provided below can be found in the limitations of the virtual power plant operation auxiliary decision-making method described above, and will not be repeated here.

[0174] In one exemplary embodiment, such as Figure 2 As shown, a decision support device for virtual power plant operation is provided, comprising: The data acquisition unit 201 is used to acquire the raw dataset of the virtual power plant. The raw dataset includes device data, market price data, weather forecast data, historical data, and rule data from IoT devices within the virtual power plant. The device data of the IoT devices includes at least device ID, acquisition timestamp, and device parameters. The IoT devices include at least photovoltaic devices, energy storage devices, and load devices. The device parameters in the photovoltaic device data include at least power generation capacity, the device parameters in the energy storage device data include at least state of charge, and the device parameters in the load device data include at least the current load. The market price data includes multiple price data groups, each of which includes at least the price acquisition time, energy price quote, and ancillary service market price.

[0175] Preprocessing unit 202 is used to preprocess the original dataset to obtain a structured feature dataset; Construction unit 203 is used to construct a virtual power plant knowledge graph based on the structured feature dataset; The generation unit 204 is used to generate multiple candidate operation schemes and reference information for each candidate operation scheme based on preset operation goals and operation constraints, using the structured feature dataset and the virtual power plant knowledge graph; wherein, the reference information includes scheme score, key factor contribution, confidence interval and scheme risk warning; Output unit 205 is used to output visual evaluation charts of each candidate operation scheme and obtain the optimal operation scheme determined by the decision-makers of the virtual power plant from multiple candidate operation schemes based on the visual evaluation charts.

[0176] Implementing the above-described methods enables decision-makers to make scientific judgments based on comprehensive and accurate information, effectively avoiding decision-making errors caused by incomplete information or misunderstandings, thereby improving the accuracy of the final selected operational plan. Furthermore, this application can improve the reliability and accuracy of virtual power plant operation auxiliary decision-making. Additionally, this application can enhance the intelligence level of virtual power plant operation. Furthermore, this application can improve the scientific rigor and rationality of the plan, ensuring the efficient and stable operation of the virtual power plant. Furthermore, this application can effectively avoid the limitations of a single plan, enhancing the diversity and comprehensiveness of plans, providing solid support for selecting the optimal operational plan. Moreover, this application can continuously improve the accuracy and practicality of the knowledge graph, providing a more reliable basis for subsequent operational decisions.

[0177] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 3 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores auxiliary decision-making data for virtual power plant operation. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an auxiliary decision-making method for virtual power plant operation.

[0178] Those skilled in the art will understand that Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0179] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0180] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0181] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0182] In one exemplary embodiment, a chip is provided, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the steps in the above method embodiments and achieve the same technical effect, and will not be described again here to avoid repetition.

[0183] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0184] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0185] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0186] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0187] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0188] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A decision support method for virtual power plant operation, characterized in that, The auxiliary decision-making methods for the operation of the virtual power plant include: Collect the raw dataset of the virtual power plant; wherein, the raw dataset includes device data, market price data, weather forecast data, historical data, and rule data of the IoT devices in the virtual power plant; the device data of the IoT devices includes at least device ID, collection timestamp, and device parameters; The original dataset is preprocessed to obtain a structured feature dataset; A virtual power plant knowledge graph is constructed based on the structured feature dataset. Based on preset operational goals and constraints, the structured feature dataset and the virtual power plant knowledge graph are used to generate multiple candidate operational schemes and reference information for each candidate operational scheme; wherein, the reference information includes scheme scores, key factor contributions, confidence intervals, and scheme risk warnings; Output visual evaluation charts for each candidate operation plan, and obtain the optimal operation plan determined by the virtual power plant's decision-makers from multiple candidate operation plans based on the visual evaluation charts.

2. The auxiliary decision-making method for virtual power plant operation according to claim 1, characterized in that, The IoT devices include at least photovoltaic devices, energy storage devices, and load devices. The device parameters in the photovoltaic device data include at least power generation; the device parameters in the energy storage device data include at least state of charge; and the device parameters in the load device data include at least the current load. The preprocessing of the original dataset to obtain a structured feature dataset specifically includes: The original dataset is cleaned of outliers to obtain the first preprocessed dataset; Missing data are imputed in the first preprocessed dataset to obtain the second preprocessed dataset; The second preprocessed dataset is normalized to obtain a structured feature dataset.

3. The auxiliary decision-making method for virtual power plant operation according to claim 1, characterized in that, The construction of the virtual power plant knowledge graph based on the structured feature dataset specifically includes: Extract target entities from the structured feature dataset; Extract the target relationships between target entities from the structured feature dataset; Based on the target entity and the target relationship, multiple triples are generated; The multiple triples are stored in a graph database, and a virtual power plant knowledge graph is constructed based on the multiple triples.

4. The auxiliary decision-making method for virtual power plant operation according to claim 2, characterized in that, The market price data includes multiple price data sets, each of which includes at least the price collection time, energy price quote, and ancillary service market price. After constructing the virtual power plant knowledge graph based on the structured feature dataset, the method further includes: Based on the market price data, determine the total cost of electricity purchase within the preset time period; Based on the device data of the IoT device, determine the system stability index within the preset time period; The total cost of electricity purchase and the system stability indicators are defined as operational targets. Based on the device data of the IoT device, the maximum power generation of the photovoltaic device, the state of charge range of the energy storage device, and the load balance of the load device are determined. The maximum power generation, the state of charge range, and the load balance are defined as operational constraints.

5. The auxiliary decision-making method for virtual power plant operation according to claim 4, characterized in that, Based on preset operational goals and constraints, the structured feature dataset and the virtual power plant knowledge graph are used to generate multiple candidate operational schemes and reference information for each candidate operational scheme, specifically including: Based on the preset operational objectives, rule reasoning is performed using the rule data and the meteorological forecast data to obtain the first candidate operational plan; Multiple historical schemes are obtained from the historical data; Determine the similarity between the stated operational objectives and various historical schemes; The historical solution with the highest similarity was selected as the second candidate operation solution. Based on the operational objectives and constraints, a third candidate operational plan is generated using the virtual power plant knowledge graph. The first candidate operation plan, the second candidate operation plan, and the third candidate operation plan are analyzed respectively to obtain reference information for each candidate operation plan.

6. The auxiliary decision-making method for virtual power plant operation according to claim 1, characterized in that, After the decision-makers of the virtual power plant determine the optimal operating scheme from multiple candidate operating schemes based on visual evaluation charts, the method further includes: The optimal operation plan was simulated and verified using a power simulation platform to obtain the actual load; The target deviation value is obtained by comparing the predicted load based on the optimal operation plan with the actual load. If the target deviation value is greater than the preset deviation threshold, the optimal operation plan is analyzed to determine the cause of the deviation; Based on the target deviation value and the cause of the deviation, the virtual power plant knowledge graph is written back to obtain an updated virtual power plant knowledge graph.

7. A decision support device for virtual power plant operation, characterized in that, The auxiliary decision-making device for the operation of the virtual power plant includes: The data acquisition unit is used to acquire the raw dataset of the virtual power plant; wherein, the raw dataset includes device data, market price data, weather forecast data, historical data, and rule data of the IoT devices in the virtual power plant; the device data of the IoT devices includes at least device ID, acquisition timestamp, and device parameters; The preprocessing unit is used to preprocess the original dataset to obtain a structured feature dataset; The construction unit is used to construct a virtual power plant knowledge graph based on the structured feature dataset; The generation unit is used to generate multiple candidate operation schemes and reference information for each candidate operation scheme based on preset operation goals and operation constraints, using the structured feature dataset and the virtual power plant knowledge graph; wherein, the reference information includes scheme score, key factor contribution, confidence interval and scheme risk warning; The output unit is used to output visual evaluation charts of each candidate operation plan and obtain the optimal operation plan determined by the decision-makers of the virtual power plant from multiple candidate operation plans based on the visual evaluation charts.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the auxiliary decision-making method for operating a virtual power plant according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the auxiliary decision-making method for operating a virtual power plant as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the auxiliary decision-making method for operating a virtual power plant as described in any one of claims 1-6.

Images