Management method for improving power generation capacity of photovoltaic power station

Abstract

The invention relates to the technical field of intelligent scheduling of photovoltaic power stations, in particular to a management method for improving the power generation capacity of a photovoltaic power station, and the method comprises the steps: analyzing a power regulation and control policy text, meteorological early warning data and electricity price historical fluctuation data, and generating a three-dimensional dynamic constraint condition set; a basic generated output prediction sequence is generated by using an LSTM time sequence model, and a three-dimensional constraint generation policy is superposed to adapt to the schedulable capacity; screening adjacent reference power stations based on a digital twinborn model, and correcting to obtain meteorological correction schedulable capacity; starting in-station compensation when the generated output of the target power station is insufficient; if compensation is insufficient, screening the auxiliary power station and generating an incremental power generation instruction set; storing the instruction and the result by using the alliance chain; and cross-power-station cooperative training is realized through federal learning. The power generation capacity of the photovoltaic power station and the power grid dispatching efficiency can be effectively improved, and the problems of power generation loss and low cross-station cooperation efficiency caused by meteorological fluctuation are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent scheduling of photovoltaic power stations, and in particular to a management method for improving the power generation capacity of a photovoltaic power station. Background Art

[0002] In the early days, PV power plant management relied primarily on manual inspections and simple logic control, enabling only basic start-up and shutdown operations and fault alarms. Managers performed daily plant operations and power generation planning based on fixed schedules or simple weather forecasts. Responses to factors such as power regulation policies, meteorological warnings, and electricity price fluctuations were neither timely nor accurate, making it difficult to form an effective set of dynamic constraints to guide plant operations. This limited power generation efficiency and the stability of power supply. Furthermore, early data analysis methods were unable to deeply mine and accurately predict historical PV power plant data, resulting in low accuracy in predicting the plant's dispatchable capacity, impacting the scientific and rational nature of grid dispatch.

[0003] With the rapid development of the photovoltaic industry, the proportion of PV power plants in the energy supply system has gradually increased, placing higher demands on PV plant management technologies. Power supply stability and power generation efficiency have become key priorities, requiring more advanced data processing and analysis technologies to optimize the operation and management of PV power plants. Some power generation forecasting methods based on simple models have begun to emerge in the market. However, these technologies have significant drawbacks. First, they fail to decouple the combined effects of meteorological factors (such as irradiance and temperature) on system efficiency, resulting in forecast errors generally exceeding 20%. Second, they lack cross-site coordinated scheduling mechanisms, making them unable to meet the grid stability requirements of large-scale clusters.

[0004] Although some current solutions attempt to incorporate digital twin technology or blockchain evidence storage, technical bottlenecks remain. Existing technologies primarily rely on a single physical model to decouple meteorological deviations, without integrating historical data from neighboring power plants for comparative analysis. This results in insufficient correction accuracy, leading to inaccurate corrections for meteorological factors and impacting the precise assessment and improvement of power generation capacity. Cross-station scheduling also fails to simultaneously quantify the dual constraints of ramp rate and installed capacity, which can easily lead to grid voltage fluctuations. This makes it difficult to ensure grid stability when multiple power plants collaborate on power generation. Existing technologies lack automated mechanisms for policy compliance verification and model optimization, and there is a risk of privacy leakage when sharing data across multiple power plants. This not only increases management costs but also limits model optimization and promotion.

[0005] Chinese invention CN116526553B discloses a management method and system for improving the power generation capacity of photovoltaic power stations. The method uses modeling and simulation technology and a loss algorithm model to analyze power losses and generate an operation and maintenance plan. However, the invention does not involve a strategy for compensating for meteorological deviations within the station, lacks cross-station collaborative scheduling logic for generating incremental power generation instructions, and only passively responds to losses through an operation and maintenance plan.

[0006] Therefore, the present invention discloses a management method for improving the power generation capacity of a photovoltaic power station. Summary of the Invention

[0007] The purpose of this invention is to solve the problems in the prior art of photovoltaic power stations such as serious power generation losses caused by meteorological fluctuations, low cross-station coordination efficiency and difficulty in policy compliance verification, and to propose a management method for improving the power generation capacity of photovoltaic power stations.

[0008] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a management method for improving the power generation capacity of a photovoltaic power station, comprising the following steps: Step S1, parsing the power regulation policy text, meteorological warning data and historical electricity price fluctuation data to generate a three-dimensional dynamic constraint condition set; Step S2: Based on the mapping relationship between real-time meteorological parameters and power generation output data, a basic power generation output forecast sequence is generated through the LSTM time series model, and a three-dimensional dynamic constraint condition set is superimposed to generate policy-adapted dispatchable capacity; Step S3: Based on the digital twin model, select neighboring reference power stations, analyze and adjust the policy to adapt the dispatchable capacity through meteorological deviation decoupling, and output the meteorologically corrected dispatchable capacity; Step S4: When the real-time power generation output of the target power station is lower than 8% of the meteorologically corrected dispatchable capacity for 15 consecutive minutes, the station meteorological deviation compensation strategy is activated, and corresponding compensation control is performed according to the type of meteorological deviation; Step S5: If the target power station continues to have insufficient power output after compensation, auxiliary power stations are selected by calculating the available margin of adjacent reference power stations, and an incremental power generation instruction set is generated with the goal of maximizing grid stability and power generation efficiency; Step S6: The incremental power generation instruction set and execution results are stored in a reliable manner through the alliance chain, and policy compliance is automatically verified based on the smart contract to generate audit records; Step S7: Optimize model parameters based on the alliance chain audit records, and use the federated learning mechanism with encrypted differential privacy to achieve cross-power station collaborative training.

[0009] The beneficial effects brought about by the technical solution provided by the present invention include at least: By combining the LSTM time series model with a three-dimensional dynamic constraint condition set, the present invention can accurately predict power generation output and make real-time corrections, thereby reducing the prediction error rate.

[0010] Through the decoupling analysis of meteorological deviations between digital twin models and adjacent reference power stations, the present invention can quickly identify the dominant factors of temperature and irradiance deviations, and implement a directional compensation strategy to reduce power generation losses caused by meteorological fluctuations.

[0011] The present invention coordinates scheduling through available margin calculation and multi-objective optimization algorithm, and can dynamically allocate the incremental power of auxiliary power stations and improve cluster power generation efficiency while ensuring the stability of the power grid.

[0012] The present invention uses a federated learning mechanism with encrypted differential privacy to achieve collaborative optimization of model parameters and improve the prediction accuracy of the global model while protecting the data privacy of each power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0014] Figure 1 A flow chart of a management method for improving the power generation capacity of a photovoltaic power station provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0015] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a management method for improving the power generation capacity of a photovoltaic power station according to the present invention, including its specific implementation, structure, features, and effectiveness. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0016] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0017] The following examples are for illustrative purposes only and are not intended to limit the scope of the present invention.

[0018] A specific solution of a management method for improving the power generation capacity of a photovoltaic power station provided by the present invention is described in detail below with reference to the accompanying drawings.

[0019] See also Figure 1 , which shows a flow chart of a management method for improving the power generation capacity of a photovoltaic power station provided by an embodiment of the present invention, the method comprising the following steps: Step S1, parsing the power regulation policy text, meteorological warning data and historical electricity price fluctuation data to generate a three-dimensional dynamic constraint condition set; Wherein, step S1 further includes the following sub-steps: S1-1, parse the power regulation policy text through natural language processing, extract the peak and valley electricity price periods, carbon emission indicators and peak and frequency regulation requirements, and generate the policy signal feature vector; S1-2, processes meteorological warning data through spatiotemporal data fusion algorithms to generate meteorological impact maps covering all power stations; S1-3, construct an electricity price fluctuation prediction model to predict the electricity price fluctuation trend in the future T period based on the historical electricity price fluctuation data and policy signal feature vector; S1-4, integrates the policy signal feature vector, meteorological impact map and electricity price fluctuation trend to generate a three-dimensional dynamic constraint condition set including time dimension, space dimension and policy dimension.

[0020] It should be noted that when using natural language processing technology to perform semantic understanding and feature extraction on the power regulation policy text, the BERT pre-training model is specifically used to encode the policy text into word vectors, and the named entity recognition technology is used to identify key information such as peak and valley electricity price periods, carbon emission indicators, and peak and frequency regulation requirements. Finally, a policy signal feature vector containing the core elements of the policy is generated. The policy signal feature vector is coupled with the subsequent electricity price prediction model through the attention mechanism to realize the dynamic guidance of policy signals on scheduling decisions.

[0021] The steps for processing meteorological warning data using a spatiotemporal data fusion algorithm to generate a meteorological impact map covering each power station include: The inverse distance weighted method is used to perform weighted interpolation on the irradiance, temperature, and wind speed data of the Solargis meteorological source to construct a regional meteorological field matrix. The time correlation of meteorological series is extracted by sliding window, and short-term fluctuation trend is captured by combining LSTM. Taking the geographical location of power stations as nodes and meteorological similarity as edge weights, a visual meteorological impact map is generated to intuitively reflect the meteorological correlation in the region.

[0022] By building an LSTM time series model to predict electricity price fluctuation trends, the specific implementation steps include: The historical electricity price data and the policy signal feature vector are concatenated and normalized to the [0,1] interval through Min-Max. The nonlinear fluctuation pattern of electricity price series is learned through the LSTM time series model, and the policy signal feature vector is weighted and integrated into the LSTM hidden layer through the attention mechanism.

[0023] The model time step is adjusted according to the grid dispatch cycle. Short-term forecasts (T≤4 hours) focus on high-frequency fluctuation characteristics, while long-term forecasts (T>24 hours) are guided by the trend of policy signals.

[0024] The steps for generating a three-dimensional constraint set include: Z-Score standardization is used for policy feature vectors (numerical type), meteorological impact maps (spatial matrix), and electricity price fluctuation trends (time series); Time dimension: Generate time-of-use power economy interval based on electricity price fluctuation trend; In the spatial dimension, the regional power derating factor is set based on the meteorological impact map; The policy dimension is generated based on the policy signal feature vector, including: Set power output range based on peak and valley electricity price periods; Allocate power plant-level carbon quotas based on carbon emission indicators; Determine the ramp rate threshold based on peak and frequency regulation requirements; The constraints of each dimension are integrated into a three-dimensional constraint matrix, and the final constraint is generated by weighted summation.

[0025] Step S2: Based on the mapping relationship between real-time meteorological parameters and power generation output data, a basic power generation output forecast sequence is generated through the LSTM time series model, and a three-dimensional dynamic constraint condition set is superimposed to generate policy-adapted dispatchable capacity; Wherein, in step S2, the following sub-steps are also included: S2-1, collects real-time power output, historical power output sequence, real-time meteorological parameters and equipment status data of each power station; S2-2 uses the LSTM time series model to learn the mapping relationship between real-time meteorological parameters and real-time power generation output, and outputs a basic power generation output forecast sequence; S2-3, superimpose the three-dimensional dynamic constraint condition set to generate policy-adaptive dispatchable capacity.

[0026] It should be noted that the historical power generation output sequence collects minute-level power generation output data of each power station for at least 3 years, covering all operating scenarios such as normal operation and fault repair.

[0027] Real-time meteorological parameters, including irradiance, component temperature, wind speed, humidity, etc., are obtained simultaneously. Data sources include on-site weather stations and third-party meteorological services.

[0028] Equipment status data, including inverter efficiency, PV module temperature coefficient, bracket inclination and other equipment parameters, are collected and transmitted in real time through the SCADA system.

[0029] The collected data are preprocessed, including outlier filtering, normalization, and time series alignment.

[0030] The LSTM timing model architecture design includes: The input layer has a feature dimension of 12 dimensions (including 3 dimensions of historical power, 6 dimensions of meteorological parameters, and 3 dimensions of equipment status), and the time step is set to 48 (corresponding to 12 hours of historical data); Hidden layer, each layer contains 128 LSTM units, the activation function is tanh, and Dropout (ratio 0.2) is used between layers to prevent overfitting; The output layer, the fully connected layer, is mapped to the power prediction value of the future T period (T∈[24,168] ​​hours), and the output dimension is T×1.

[0031] The loss function uses a combination of mean square error and mean absolute error loss, and the formula is expressed as:

[0032] The Adam optimizer (learning rate 0.001, β1=0.9, β2=0.999) is used, the batch size is set to 64, and the training period is set to 200 rounds.

[0033] The logic for superimposing a set of 3D dynamic constraints includes: Policy dimension correction: adjust the power upper and lower limits according to the policy signal feature vector; Meteorological dimension correction: Based on the meteorological impact map, power restrictions are relaxed for power plants in high-irradiance areas, and derating constraints are set for shadowed areas. The electricity price dimension is modified, and combined with the electricity price forecast trend, a "storage charging priority" constraint is generated during periods of low electricity prices, and the power limit is increased during peak periods.

[0034] The formula for generating the final policy-adapted dispatchable capacity is expressed as:

[0035] in, Indicates the dynamic correction coefficient; Indicates policy adaptation to dispatchable capacity; Represents the basic power output forecast value output by the LSTM time series model; and They represent the minimum capacity and the maximum capacity of policy adaptation respectively; Represents a truncation function that limits the intermediate values ​​to interval.

[0036] Step S3: Based on the digital twin model, select neighboring reference power stations, analyze and adjust the policy to adapt the dispatchable capacity through meteorological deviation decoupling, and output the meteorologically corrected dispatchable capacity; Wherein, in step S3, the following sub-steps are also included: S3-1: Build a digital twin model of the target power station, synchronize PV module temperature, irradiance, and inverter efficiency, and predict the PV module temperature and irradiance baseline values ​​for the target power station in the next 15 minutes based on the meteorological correlation with neighboring power stations; S3-2: Based on the digital twin model, select neighboring power stations with a meteorological similarity of ≥90% and a distance of ≤5 km to the target power station to obtain neighboring reference power stations; S3-3, calculates the meteorological deviation influencing factors of PV module temperature and irradiance through the Euclidean distance algorithm, outputs the comprehensive loss correction coefficient, and corrects the policy-adapted dispatchable capacity based on the comprehensive loss correction coefficient, and outputs the meteorologically corrected dispatchable capacity.

[0037] It should be noted that the specific steps for building a digital twin model of the target power station are as follows: Acquire the photovoltaic array layout based on 3D laser scanning and construct the physical topology of the module-bracket-inverter; Establish an equivalent circuit model of the photovoltaic module, and use a space vector control model in the dq coordinate system for the inverter; Based on Fourier heat conduction equation and fluid mechanics model, the temperature field distribution of components is simulated.

[0038] The temperature of the photovoltaic module is collected by an infrared thermal imager (accuracy ±0.5°C) or a temperature sensor embedded on the surface of the module.

[0039] Irradiance is measured using a photosynthetically active radiation sensor (accuracy ±5W / m²) deployed at the center of the module array.

[0040] The inverter efficiency is calculated in real time by collecting the DC side input power and AC side output power through the SCADA system.

[0041] The synchronization error of the above collected data is controlled at ≤1.5%, ensuring the real-time consistency between the model and the physical power station.

[0042] The temperature benchmark value of photovoltaic modules refers to the theoretical temperature value generated by the digital twin model through simulation of the heat conduction equation based on the meteorological data of the target power station in the past three years and the historical correlation data of the neighboring reference power stations, reflecting the ideal temperature level that the modules should reach under the meteorological conditions.

[0043] The irradiance benchmark value refers to the theoretical irradiance value predicted by the digital twin model based on the historical irradiance time series characteristics, geographical latitude and atmospheric transparency parameters during the same period. It is generated synchronously with the temperature benchmark value and serves as the baseline for evaluating real-time irradiance deviations.

[0044] The screening process for nearby reference power plants includes: Obtain the coordinates of the target power station through the GIS system, search for all power stations within a 5km radius, and form a list of candidate power stations; Extract the minute-level meteorological data (irradiance and temperature as core parameters) of the candidate power station and the target power station over the past three hours and calculate the comprehensive similarity; The candidate power stations with a meteorological similarity of ≥90% are retained as adjacent reference power stations; The meteorological deviation impact factor is calculated by extracting the real-time temperature and irradiance of the target power station and the adjacent reference power stations to construct a two-dimensional feature vector. The meteorological difference between the target power station and the adjacent reference power stations is quantified by the Euclidean distance, and the comprehensive loss correction coefficient is calculated as follows:

[0045] in, represents the comprehensive loss correction coefficient; k represents the proportional coefficient, which is fitted by historical data; represents the normalized Euclidean distance.

[0046] Based on the impact of the measured PV module temperature and irradiance deviation on power generation efficiency, a comprehensive loss correction coefficient is generated and the policy-adapted dispatchable capacity is updated to obtain the meteorologically corrected dispatchable capacity. The formula is expressed as:

[0047] in, represents the meteorologically corrected dispatchable capacity; Indicates policy adaptation to dispatchable capacity; Represents the comprehensive loss correction factor.

[0048] The execution logic of policy adaptation dispatchable capacity correction includes: If there is only one adjacent reference station, use Modify the policy to adapt to dispatchable capacity: If there are N neighboring reference power stations, a weighted average is used (the weight is the similarity between each neighboring reference power station and the target power station).

[0049] Step S4: When the real-time power generation output of the target power station is lower than 8% of the meteorologically corrected dispatchable capacity for 15 consecutive minutes, the station meteorological deviation compensation strategy is activated, and corresponding compensation control is performed according to the type of meteorological deviation; Wherein, in step S4, the following sub-steps are also included: S4-1: When the real-time power generation output of the target power station is lower than 8% of the meteorologically corrected dispatchable capacity for 15 consecutive minutes, the station compensation is triggered; S4-2, implement compensation strategies based on the type of meteorological deviation, including: When temperature deviation dominates, adjust the inverter reactive power; When irradiance deviation is dominant, adjust the tilt angle of the PV panels.

[0050] It should be noted that the temperature sensitivity coefficient is calculated separately and irradiance sensitivity coefficients , quantify the influence of temperature B and irradiance G on output, the formula is expressed as:

[0051]

[0052] in, Indicates the deviation between real-time output and meteorologically corrected dispatchable capacity; Indicates the deviation of temperature from the reference temperature of the PV module; Indicates the deviation of irradiance from the irradiance reference value.

[0053] When temperature deviation dominates, the increase in component temperature will cause the open-circuit voltage to drop. If the grid voltage is too low, the inverter may limit its output power due to low voltage ride-through. By adjusting the inverter's reactive power, the voltage loss can be compensated, the system voltage can be maintained stable, and the power generation output can be indirectly improved.

[0054] When irradiance deviation is dominant, the irradiance is insufficient. By adjusting the inclination angle of the components, the photovoltaic panels can be made more perpendicular to the direction of sunlight incidence, the amount of radiation received can be increased, and the power generation capacity can be improved.

[0055] Step S5: If the target power station continues to have insufficient power output after compensation, auxiliary power stations are selected by calculating the available margin of adjacent reference power stations, and an incremental power generation instruction set is generated with the goal of maximizing grid stability and power generation efficiency; Wherein, in step S5, the following sub-steps are also included: S5-1: If the target power station's power generation output remains insufficient after compensation, calculate the available margin of the adjacent reference power stations and select the power stations that meet M>0 among the adjacent reference power stations as auxiliary power stations; The available margin of the adjacent reference power station is calculated as follows:

[0056] Where M represents the available margin; represents the rated installed capacity of the adjacent reference power station; Indicates the real-time power output of the adjacent reference power station; Indicates the ramp rate limit of the adjacent reference power station; Indicates the scheduling period; S5-2: Under the conditions of voltage fluctuation ≤ 5%, photovoltaic system power generation efficiency ≥ 80%, and policy constraints, a multi-objective optimization algorithm is constructed with the goals of maximizing grid stability and power generation efficiency, and incremental power is allocated based on the multi-objective optimization algorithm; S5-3, generate a structured incremental power generation instruction set, including the auxiliary power plant ID, incremental power and scheduling period. The scheduling period is a specific execution time interval, and the time length is equal to .

[0057] It should be noted that the incremental power generation instruction set refers to a structured instruction set that includes the auxiliary power plant ID, incremental power, and scheduling period. It is encoded in JSON format to ensure cross-system compatibility. The instructions are sent to the auxiliary power plant through the MQTT protocol, using QoS level 2 to ensure transmission reliability.

[0058] Step S6: The incremental power generation instruction set and execution results are stored in a reliable manner through the alliance chain, and policy compliance is automatically verified based on the smart contract to generate audit records; S6-1 uses the SHA-256 algorithm to generate a hash summary of the incremental power generation instruction set and execution results, attaches a timestamp and digital signature, and stores the evidence through the alliance chain; S6-2, verify the execution results through smart contracts and generate audit records, including: When the self-generation and self-consumption rate is ≥70%, it is marked as meeting the settlement conditions for the highest electricity price level; When carbon emissions exceed policy indicators, an over-standard alarm is triggered and transactions are frozen.

[0059] It should be noted that the consortium chain stores evidence, including the incremental power generation instruction set and the execution results uploaded by the auxiliary power plant. A SHA-256 hash digest is generated for the incremental power generation instruction set and execution results. A trusted timestamp (accurate to the millisecond) and the digital signature of the instruction issuer are attached. This information is then uploaded to the consortium chain (nodes include power grid companies, PV power plants, and regulatory agencies, and authorization is required for participation). After consensus is reached among the nodes, the evidence is stored and the data is written to the blockchain, forming a chain structure to ensure that it cannot be tampered with.

[0060] Encode policy compliance rules (spontaneous consumption rate, carbon emission requirements) into smart contracts in advance and deploy them to the alliance chain; after the execution results are stored, the contract automatically triggers the verification audit record; the verification results generate structured records (including instruction ID, verification results, and timestamp) and are stored in the alliance chain.

[0061] Step S7: Optimize model parameters based on the alliance chain audit records and use a federated learning mechanism with encrypted differential privacy to achieve cross-plant collaborative training; Wherein, in step S7, the following sub-steps are also included: S7-1, based on the audit records stored in the alliance chain, calculate the actual execution deviation rate of the policy dimension constraints and update the policy library weight coefficient through the adaptive weighting algorithm; In S7-2, each power station uses the Paillier algorithm to encrypt the LSTM time series model and digital twin model gradients and adds Gaussian differential privacy noise to the encrypted gradients. In S7-3, the cloud aggregates the encrypted parameters of each power station through the FedProx algorithm, combines it with zero-knowledge proof to verify the credibility of the nodes, and generates a global optimization model; S7-4, distribute the optimized model parameters through the quantum key distribution channel and perform periodic or triggered model updates.

[0062] It should be noted that the adaptive weighting algorithm logic includes: The policy dimension with a higher deviation rate has a larger weight coefficient, so that the model will pay more attention to the constraints of this dimension in the future; For dimensions with a deviation rate ≤ 5%, the weights are maintained or fine-tuned to reduce invalid calculations; The updated weight coefficients are used to generate a three-dimensional dynamic constraint condition set.

[0063] The policy library weight coefficient is a parameter that measures the impact of policy constraints (such as carbon emission indicators, peak and valley electricity price periods, and peak and frequency regulation requirements) on scheduling decisions. Its value range is [0, 1], and its sum is 1. The initial value is set based on the priority of the policy document and is subsequently dynamically updated using an adaptive weighting algorithm.

[0064] After local training at each power station, the gradient parameters of the LSTM model (power generation forecast) and the digital twin model (meteorological correction) are extracted; the gradients are encrypted using the Paillier homomorphic encryption algorithm. Calculations can be performed directly in the encrypted state without the need for decryption, ensuring that the gradient data is not leaked.

[0065] Gaussian differential privacy noise is added to the encrypted gradient with a privacy budget of ε=1.0 to ensure data privacy is not leaked. The noise formula is expressed as:

[0066] in, represents the gradient after adding noise; g represents the original gradient; represents Gaussian noise with a mean of 0; represents the variance, σ controls the noise size, the larger σ is, the stronger the noise is; I represents the unit matrix.

[0067] The cloud receives the encrypted gradients of each power station and uses the FedProx algorithm (which adds a proximal term on the basis of the traditional federated average) to solve the problem of large differences in data distribution among power stations and reduce aggregation bias.

[0068] Zero-knowledge proof verifies node credibility. Each power station needs to prove that gradient encryption and noise addition comply with the rules without leaking the original data. After cloud verification, the node gradient will be included in the aggregation.

[0069] Quantum key distribution is used to generate encryption keys for encrypting global model parameters; the keys are updated in real time (once per hour) to ensure the absolute security of the parameter transmission process.

[0070] The default periodic update is once a week (matching the grid dispatch cycle); for triggered updates, an update is triggered immediately when the policy execution deviation rate is ≥10% for three consecutive times.

[0071] In this way, a management method for improving the power generation capacity of a photovoltaic power station can be achieved.

[0072] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A management method for improving the power generation capacity of a photovoltaic power station, characterized in that: The method includes: Step S1, parsing the power regulation policy text, meteorological warning data and historical electricity price fluctuation data to generate a three-dimensional dynamic constraint condition set; Step S2: Based on the mapping relationship between real-time meteorological parameters and power generation output data, a basic power generation output forecast sequence is generated through the LSTM time series model, and a three-dimensional dynamic constraint condition set is superimposed to generate policy-adapted dispatchable capacity; Step S3: Based on the digital twin model, select neighboring reference power stations, analyze and adjust the policy to adapt the dispatchable capacity through meteorological deviation decoupling, and output the meteorologically corrected dispatchable capacity; Step S4: When the real-time power generation output of the target power station is lower than 8% of the meteorologically corrected dispatchable capacity for 15 consecutive minutes, the station meteorological deviation compensation strategy is activated, and corresponding compensation control is performed according to the type of meteorological deviation; Step S5: If the target power station continues to have insufficient power output after compensation, auxiliary power stations are selected by calculating the available margin of adjacent reference power stations, and an incremental power generation instruction set is generated with the goal of maximizing grid stability and power generation efficiency; Step S6: The incremental power generation instruction set and execution results are stored in a reliable manner through the alliance chain, and policy compliance is automatically verified based on the smart contract to generate audit records; Step S7: Optimize model parameters based on the alliance chain audit records, and use the federated learning mechanism with encrypted differential privacy to achieve cross-power station collaborative training.

2. A management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, step S1 further includes the following sub-steps: S1-1, parse the power regulation policy text through natural language processing, extract the peak and valley electricity price periods, carbon emission indicators and peak and frequency regulation requirements, and generate the policy signal feature vector; S1-2, processes meteorological warning data through spatiotemporal data fusion algorithms to generate meteorological impact maps covering all power stations; S1-3, construct an electricity price fluctuation prediction model to predict the electricity price fluctuation trend in the future T period based on the historical electricity price fluctuation data and policy signal feature vector; S1-4, integrating the policy signal characteristic vector, meteorological impact map and electricity price fluctuation trend to generate a three-dimensional dynamic constraint condition set including time dimension, space dimension and policy dimension.

3. A management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, in step S2, the following sub-steps are also included: S2-1, collects real-time power output, historical power output sequence, real-time meteorological parameters and equipment status data of each power station; S2-2, learning the mapping relationship between the real-time meteorological parameters and the real-time power generation output through the LSTM time series model, and outputting a basic power generation output prediction sequence; S2-3, superimpose the three-dimensional dynamic constraint condition set to generate policy-adaptive dispatchable capacity.

4. The management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, in step S3, the following sub-steps are also included: S3-1: Build a digital twin model of the target power station, synchronize PV module temperature, irradiance, and inverter efficiency, and predict the PV module temperature and irradiance baseline values ​​for the target power station in the next 15 minutes based on the meteorological correlation with neighboring power stations; S3-2, based on the digital twin model, screening neighboring power stations with a meteorological similarity of ≥90% and a distance of ≤5 km to the target power station to obtain neighboring reference power stations; S3-3, calculates the meteorological deviation influencing factors of PV module temperature and irradiance through the Euclidean distance algorithm, outputs the comprehensive loss correction coefficient, and corrects the policy-adapted dispatchable capacity based on the comprehensive loss correction coefficient, and outputs the meteorologically corrected dispatchable capacity.

5. The management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, in step S4, the following sub-steps are also included: S4-1: When the real-time power generation output of the target power station is lower than 8% of the meteorologically corrected dispatchable capacity for 15 consecutive minutes, the station compensation is triggered; S4-2, implement compensation strategies based on the type of meteorological deviation, including: When temperature deviation dominates, adjust the inverter reactive power; When irradiance deviation is dominant, adjust the tilt angle of the PV panels.

6. A management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, in step S5, the following sub-steps are also included: S5-1: If the target power station's power output remains insufficient after compensation, calculate the available margin of each power station and select nearby reference power stations that meet M>0 as auxiliary power stations; The available margin of the adjacent reference power station is calculated as follows: Where M represents the available margin; represents the rated installed capacity of the adjacent reference power station; Indicates the real-time power output of the adjacent reference power station; Indicates the ramp rate limit of the adjacent reference power station; Indicates the scheduling period; S5-2, under the conditions of voltage fluctuation ≤ 5%, photovoltaic system power generation efficiency ≥ 80%, and policy dimension constraints, construct a multi-objective optimization algorithm with the goals of maximizing grid stability and power generation efficiency, and allocate incremental power based on the multi-objective optimization algorithm; S5-3, generate a structured incremental power generation instruction set, including the auxiliary power station ID, incremental power and scheduling period, the scheduling period is a specific execution time interval, the time length is equal to .

7. The management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, in step S6, the following sub-steps are also included: S6-1 uses the SHA-256 algorithm to generate a hash summary of the incremental power generation instruction set and execution results, attaches a timestamp and digital signature, and stores the evidence through the alliance chain; S6-2, verify the execution results through smart contracts and generate audit records, including: When the self-generation and self-consumption rate is ≥70%, it is marked as meeting the settlement conditions for the highest electricity price level; When carbon emissions exceed policy indicators, an over-standard alarm is triggered and transactions are frozen.

8. The management method for improving the power generation capacity of a photovoltaic power station according to claim 1, characterized in that: Wherein, in step S7, the following sub-steps are also included: S7-1, based on the audit records stored in the alliance chain, calculate the actual execution deviation rate of the policy dimension constraints and update the policy library weight coefficient through the adaptive weighting algorithm; In S7-2, each power station uses the Paillier algorithm to encrypt the LSTM time series model and digital twin model gradients and adds Gaussian differential privacy noise to the encrypted gradients. In S7-3, the cloud aggregates the encrypted parameters of each power station through the FedProx algorithm, combines it with zero-knowledge proof to verify the credibility of the nodes, and generates a global optimization model; S7-4, distribute the optimized model parameters through the quantum key distribution channel and perform periodic or triggered model updates.

Images