A clean energy station power generation strategy optimization method and system

Abstract

The present application relates to the technical field of clean energy power generation, and discloses a clean energy station power generation strategy optimization method and system. The method comprises collecting real-time operation data of multiple power generation devices in the clean energy station, including power generation power, environmental parameters and device state parameters. Based on these real-time operation data, a station operation feature matrix representing the dynamic correlation relationship between different power generation devices is constructed. According to a preset power generation strategy optimization target, an initial power generation strategy set containing multiple candidate power generation strategies is generated. The station operation feature matrix is input into a strategy evaluation model, and the adaptability of each candidate strategy in the initial set is evaluated to obtain a strategy evaluation result. Based on the evaluation result, the optimal power generation strategy is screened out and is issued to the clean energy station for execution. The present application realizes real-time and collaborative optimization of the station power generation strategy.

Description

Technical Field

[0001] This invention relates to the field of clean energy power generation technology, specifically to a method and system for optimizing power generation strategies at clean energy power plants. Background Technology

[0002] With the acceleration of the global energy transition, the construction scale of clean energy power plants, represented by wind and solar power, continues to expand. These power plants typically include various power generation equipment, such as different types of wind turbine generators, fixed-axis and tracking photovoltaic arrays, forming a complex energy production system. Clean energy power generation is inherently intermittent and fluctuating, and its output level is directly constrained by natural conditions such as wind speed, sunlight intensity, and ambient temperature. Within a power plant, there are also complex interactions between power generation equipment of different locations and types. For example, the wake of upstream wind turbines can reduce the incoming wind speed of downstream wind turbines, and photovoltaic arrays may experience differences in power generation efficiency due to shading. These factors mean that the overall power generation capacity of a power plant is not a simple sum of the output of individual devices, but rather exhibits strong spatiotemporal coupling characteristics.

[0003] Existing power plant-level operation control strategies mostly employ relatively simplified methods. Common practices include setting a uniform power curve model or allocating power generation plans based on historical average data. These methods fail to fully consider the dynamic interactions between equipment and real-time changes in the external environment. For example, when local wind speeds suddenly increase, traditional methods may not be able to quickly coordinate the operating states of adjacent wind turbines to minimize wake losses; when clouds drift over a photovoltaic power plant, fixed strategies struggle to optimize the coordinated operation of photovoltaic arrays in different areas to smooth fluctuations in total power output. This control approach, based on static models or hysteresis responses, often results in actual power plant operating efficiency being lower than design values, leading to frequent wind and solar curtailment.

[0004] At the technical level, existing research has attempted to apply optimization algorithms for power generation strategy formulation, but most studies focus on single-unit optimization or power plant models based on ideal assumptions. These methods have limitations in handling multi-device collaborative operation. On the one hand, they typically require accurate physical models to describe the interactions between devices, but the physical environment of actual power plants is complex, making accurate modeling difficult and costly. On the other hand, most optimization algorithms have high computational complexity, making it difficult to meet the real-time requirements of power plant operation for strategy generation. Furthermore, existing methods often take maximizing instantaneous power generation as the sole objective, rarely considering multiple constraints such as equipment wear and tear and grid dispatch requirements, thus limiting the practicality of the strategies.

[0005] There is an urgent need in the field of clean energy power generation for a new method that can adapt to the complex operating environment of power plants and quickly generate collaborative optimization strategies in order to fully tap the power generation potential of power plants and improve the utilization efficiency and economy of clean energy. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for optimizing power generation strategies at clean energy power plants, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a method for optimizing power generation strategies at clean energy power plants, the method comprising:

[0008] Collect real-time operating data of various power generation equipment in clean energy power plants, including power generation, environmental parameters and equipment status parameters;

[0009] Based on the real-time operation data, a power station operation feature matrix is ​​constructed, which is used to characterize the dynamic correlation between different power generation equipment.

[0010] Based on the preset power generation strategy optimization objective, an initial power generation strategy set is generated, which contains multiple candidate power generation strategies;

[0011] The power station operation feature matrix is ​​input into the strategy evaluation model to perform an adaptive evaluation on each candidate power generation strategy in the initial power generation strategy set, and the strategy evaluation result is obtained.

[0012] The optimal power generation strategy is selected based on the strategy evaluation results, and the optimal power generation strategy is then distributed to clean energy power plants for execution.

[0013] Preferably, the step of constructing the station operation feature matrix based on the real-time operation data includes:

[0014] Extract the power generation sequence and environmental parameter sequence from the real-time operating data, and calculate the power correlation coefficient between different power generation devices;

[0015] A device correlation matrix is ​​constructed based on the power correlation coefficient, and the device correlation matrix is ​​used to describe the cooperative operation relationship between power generation equipment;

[0016] The equipment health vector is generated by combining the equipment status parameters, and the equipment correlation matrix is ​​fused with the equipment health vector to form the station operation feature matrix.

[0017] Preferably, generating an initial power generation strategy set according to a preset power generation strategy optimization objective includes:

[0018] Obtain grid dispatch instructions and energy storage system status from clean energy power plants to determine power generation strategy optimization objectives;

[0019] Based on the power generation strategy optimization objective, an initial power generation strategy set is generated using a multi-objective optimization algorithm. Each candidate power generation strategy in the initial power generation strategy set includes a power generation allocation scheme and an energy storage scheduling scheme.

[0020] Preferably, the step of inputting the power station operation feature matrix into the strategy evaluation model to perform an adaptive evaluation of each candidate power generation strategy in the initial power generation strategy set includes:

[0021] The station operation feature matrix is ​​matched with each candidate power generation strategy, and the strategy matching degree is calculated;

[0022] Based on the strategy matching degree and preset constraints, the feasibility of each candidate power generation strategy is evaluated, and strategy evaluation results are generated.

[0023] Preferably, the step of selecting the optimal power generation strategy based on the strategy evaluation results includes:

[0024] The initial power generation strategy set is sorted according to the strategy evaluation results, and the candidate power generation strategy with the highest matching degree is selected as the optimal power generation strategy.

[0025] The optimal power generation strategy is fine-tuned based on the energy storage system status to ensure coordinated execution of the power generation allocation scheme and the energy storage scheduling scheme.

[0026] Preferably, the fine-tuning of the optimal power generation strategy based on the energy storage system status includes:

[0027] Obtain the current charge / discharge state and remaining capacity of the energy storage system, and calculate the energy storage regulation capability;

[0028] The power allocation scheme in the optimal power generation strategy is adjusted according to the energy storage regulation capability to ensure that the energy storage system maintains stable operation during the execution of the optimal power generation strategy.

[0029] Preferably, the method further includes:

[0030] During the execution of the optimal power generation strategy, the operating status of the clean energy power station is monitored in real time, and updated real-time operating data is collected.

[0031] The optimal power generation strategy is dynamically adjusted based on the updated real-time operating data to ensure that the power generation strategy is synchronized and optimized with the station's operating status.

[0032] Preferably, the step of dynamically adjusting the optimal power generation strategy based on the updated real-time operating data includes:

[0033] The station operation feature matrix is ​​recalculated based on the updated real-time operation data, and the strategy evaluation model is updated accordingly.

[0034] The updated strategy evaluation model is used to re-evaluate the currently implemented optimal power generation strategy and generate an adjusted power generation strategy.

[0035] Preferably, the step of re-evaluating the currently executed optimal power generation strategy using the updated strategy evaluation model includes:

[0036] Compare the differences in power generation strategies before and after the adjustment, calculate the adjustment range, and if the adjustment range exceeds a preset threshold, regenerate the optimal power generation strategy and issue it for execution.

[0037] Preferably, the present invention also includes a clean energy power plant power generation strategy optimization system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described clean energy power plant power generation strategy optimization method.

[0038] Compared with the prior art, the beneficial effects of the present invention are:

[0039] This invention dynamically captures the interrelationships between various power generation devices by collecting and analyzing operational data in real time. Traditional static models cannot accurately reflect the impact of time-varying factors such as wake effects and shading losses on the overall power output of the power station, while the operational characteristic matrix constructed in this invention can quantify these dynamic correlations, providing a more realistic data foundation for strategy optimization. This real-time data-based analysis method enhances the ability to characterize the complex operating states of power stations.

[0040] This method employs a screening mechanism based on a strategy evaluation model, enabling efficient handling of multi-objective optimization problems. The operation of clean energy power plants requires consideration not only of maximizing power generation but also of various factors such as equipment lifespan and grid dispatch instructions. This invention generates a candidate strategy set by pre-setting optimization objectives and utilizes an evaluation model for multi-dimensional adaptive assessment, ensuring that the final selected optimal strategy balances multiple optimization objectives and meets the comprehensive requirements of economic, safe, and stable operation of the power plant.

[0041] The strategy generation and evaluation process exhibits good real-time performance. Compared to optimization methods that rely on complex physical models and lengthy simulation calculations, this invention reduces the dependence on precise physical models and simplifies the calculation process by constructing a feature matrix and evaluation model. This shortens the strategy optimization cycle, enabling faster response to sudden changes in environmental conditions such as wind speed and sunlight, meeting the real-time control requirements of power plants, and helping to reduce power generation losses caused by response delays.

[0042] The optimization method provided by this invention possesses strong versatility and scalability. Its core lies in extracting operational characteristics from data and evaluating strategy adaptability, rather than being tied to specific equipment models or power plant layouts. Therefore, this method can be applied to hybrid energy power plants containing different types of power generation equipment and can adapt to changes brought about by power plant expansion or equipment upgrades. This flexibility makes it widely applicable in clean energy power plants of different sizes and configurations.

[0043] The application of this method helps improve the overall operational efficiency of clean energy power plants. By optimizing the coordinated operation between equipment, greater power generation potential can be tapped under the same resource conditions, reducing efficiency losses caused by internal losses. Simultaneously, by generating smoother and more predictable power output strategies, the impact on the power grid can be mitigated, improving the grid-friendliness of clean energy. In the long run, this has a positive effect on improving the utilization rate of clean energy and enhancing its competitiveness in the energy system. Attached Figure Description

[0044] Figure 1 This is a schematic diagram illustrating the working principle of a clean energy power plant power generation strategy optimization method according to the present invention.

[0045] Figure 2 A flowchart for constructing a station operation feature matrix based on real-time operation data;

[0046] Figure 3 A flowchart for generating an initial set of power generation strategies based on preset targets;

[0047] Figure 4 This is a diagram illustrating the optimization of power generation strategies and the analysis of energy storage regulation. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Please see Figure 1This invention provides a method and system for optimizing power generation strategies at clean energy power plants. The method includes collecting real-time operational data from various power generation devices at the clean energy power plant. This data covers key information such as power generation capacity, environmental parameters, and equipment status parameters. Power generation capacity data includes the real-time output power values ​​of each device. Environmental parameters include factors affecting power generation efficiency such as temperature, humidity, wind speed, and solar irradiance. Equipment status parameters involve equipment operating status, fault indicators, and health indicators. Based on the collected real-time operational data, a power plant operational feature matrix is ​​constructed. This matrix represents the dynamic correlation between different power generation devices through mathematical modeling, such as using covariance analysis or time series analysis to capture the synergistic or competitive effects between devices. An initial set of power generation strategies is generated according to a preset power generation strategy optimization objective. The optimization objective may include maximizing power generation efficiency, minimizing operating costs, or meeting grid dispatch requirements. The initial set contains multiple candidate power generation strategies, each defining a power generation allocation and energy storage dispatch scheme. The power plant's operational feature matrix is ​​input into a strategy evaluation model, which can be a rule-based system or a machine learning model. This model adaptively evaluates each candidate power generation strategy, considering the match between the strategy and the power plant's operating status, as well as constraints such as equipment capacity limitations. The evaluation generates strategy evaluation results, such as scores or rankings. Based on these results, the optimal power generation strategy is selected using a ranking or optimization algorithm. This strategy is then deployed to the clean energy power plant for execution, achieving real-time optimization of the power generation process.

[0050] Example 1: See Figure 2 The construction of the power plant operation characteristic matrix begins with the collection and preprocessing of real-time operating data from various power generation devices in the clean energy power plant. The collected data stream includes second-level or minute-level operating parameters of key equipment such as photovoltaic inverters, wind turbines, and energy storage converters. Power generation data is presented as a continuous sequence changing over time, reflecting the instantaneous output of the equipment. The environmental parameter sequence covers external variables affecting power generation efficiency, such as irradiance, ambient temperature, wind speed and direction, and humidity. Equipment status parameters are obtained from the monitoring system, including indicators reflecting the health status of the equipment, such as equipment start / stop signals, alarm codes, internal temperature, and insulation resistance values. The raw data often contains outliers caused by communication interference, missing values ​​due to sensor failures, and data asynchrony issues caused by different sampling frequencies. The preprocessing stage uses a sliding window filtering algorithm to smooth the instantaneous fluctuations in power data, uses linear interpolation or spline interpolation methods to fill in short-term data gaps, and performs timestamp alignment operations on data from different sources to form a regularized data set with a unified time reference.

[0051] Extracting the preprocessed power generation sequence and environmental parameter sequence is the foundation for constructing the feature matrix. The power sequence is grouped by equipment identifier to form the power change curve of each equipment during the observation period. The environmental parameter sequence is classified according to the measurement point and type. For example, multiple parameters collected from the same meteorological station are grouped together. Calculating the power correlation coefficient between different power generation equipment requires selecting an appropriate statistical measurement method. The Pearson correlation coefficient is often used to measure the degree of linear correlation. It is calculated based on the ratio of the covariance of the equipment power sequence to its respective standard deviation. The result ranges from -1 to 1. A value close to 1 indicates that the output change trends of the two devices are highly synchronized, a value close to -1 indicates that the output shows inverse change characteristics, and a value close to 0 suggests that the output behavior is independent. For scenarios with obvious nonlinear relationships, mutual information or rank-based Spearman correlation coefficients can be used for supplementary analysis to capture dependencies beyond linear models.

[0052] Based on the calculated power correlation coefficients, an equipment correlation matrix is ​​constructed. This matrix is ​​an N-order square matrix (N is the total number of generating equipment in the power station). The main diagonal elements of the matrix are usually set to 1, indicating that the equipment is perfectly correlated with itself. Off-diagonal elements are set to... Then, the power correlation coefficient values ​​between the i-th device and the j-th device are stored. The symmetry of the matrix stems from the symmetry of the correlation calculation, i.e. equal The mathematical structure of this matrix intuitively depicts the collaborative operation network among equipment groups within a power station. Clusters with high correlation coefficients often correspond to equipment clusters dominated by the same environmental factors, such as wind turbine groups in the same wind farm area or photovoltaic strings under the same illumination conditions. Generating an equipment health vector by combining equipment status parameters requires designing a comprehensive evaluation model. This model integrates discrete alarm signals, continuous operating parameters, and accumulated operating time into a scalar value representing the overall health level of the equipment. This can be achieved using an expert rule-based scoring system, where a longer fault-free operating time results in a higher score, while a significant decrease occurs when severe alarms are present. Alternatively, machine learning models such as gradient boosting decision trees can be applied, training the health prediction model using historical fault data as labels. The final generated health vector is an N-dimensional column vector, with each element corresponding to the health score of a piece of equipment. The score values ​​are typically normalized to the range of 0 to 1.

[0053] The fusion process of the device association matrix and the device health vector begins with an in-depth analysis of the characteristics of the two data structures. The device association matrix, as a symmetric square matrix, has its row and column dimensions determined by the number of devices, and each element value represents the degree of statistical association between devices. The health vector, on the other hand, is a one-dimensional array structure with a length consistent with the number of devices, and each element value represents the operating status score of the corresponding device. The fusion operation requires effectively integrating these two types of data with different dimensions and physical meanings while maintaining the interpretability of the mathematical relationships. In the preprocessing stage, the health vector is normalized to ensure that its numerical range is consistent with the correlation coefficient scale of the association matrix. The normalized health values ​​are restricted to the range of zero to one to facilitate subsequent weighted calculations. At the same time, the correspondence between device identifiers is checked to ensure that the matrix row number and the vector index point to the same physical device. The data alignment stage also verifies the consistency of timestamps to ensure that the data period used for the association matrix calculation completely overlaps with the runtime period on which the health assessment is based, avoiding information mismatch caused by time asynchrony.

[0054] The weighted fusion method employs a row-wise scaling logic, multiplying each element value in the health vector by all elements in the corresponding row of the association matrix. This operation essentially adjusts the importance of all associations based on the device's health status. When a device has a high health score, its association with other devices is strengthened in the fused matrix; conversely, if the health score is low, the weight of the corresponding association is weakened. The matrix after this row-weighted processing retains the original association patterns between devices while incorporating the moderating effect of device reliability on association strength. The matrix augmentation fusion method uses a horizontal splicing technique, converting the health vector into a square matrix through diagonalization. This square matrix has zeros at all positions except for the main diagonal elements, which are health values. This diagonal health matrix is ​​then spliced ​​side-by-side with the original association matrix. The new matrix generated by this fusion method is structurally divided into two blocks: the left block fully retains the original device association information, while the right block centrally displays the independent health status of each device. Subsequent analysis algorithms can process the features of these two dimensions separately.

[0055] Tensor fusion methods construct a three-dimensional data structure to accommodate more complex relationship patterns. The first and second dimensions correspond to the row and column dimensions of the device index and the association matrix, respectively, while the third dimension is used to distinguish different types of features. In the specific implementation, the device association matrix is ​​used as the first feature slice of the tensor, and the matrix generated by the outer product operation of the health vector is used as the second feature slice. This representation method can simultaneously capture the direct associations between devices and the indirect interactions generated through health status. The choice of fusion method needs to be tailored to the characteristics of the subsequent strategy evaluation model. If the evaluation model is based on a graph neural network, the weighted fusion method is preferred because the single tensor structure output by this method is highly compatible with the node feature representation of the graph structure. If the evaluation model uses a multi-channel convolutional neural network, the tensor fusion method is suitable, as different feature slices can correspond to different input channels of the convolutional network. For evaluation models that need to analyze relational features and state features separately, the matrix augmentation fusion method can provide more flexible feature separation capabilities. An anomaly handling mechanism also needs to be set during the fusion process. When an abnormally low value is detected in the health vector, it may indicate that the device is in a faulty state. At this time, a special processing procedure can be triggered, such as forcibly setting the weights of all associations of the device to zero and adding a fault marker bit to the matrix. Extreme correlation coefficients in the correlation matrix also need to be validated for reasonableness to prevent erroneous correlations caused by sensor malfunctions or data anomalies from being included in the fusion process. The resulting site operation feature matrix after fusion needs quality verification to check whether the matrix's symmetry has been disrupted by the fusion operation, verify whether the distribution range of eigenvalues ​​meets expectations, and confirm that there are no non-numerical or outlier values ​​outside the reasonable range. Visualization tools can be used to present the fused matrix as a heatmap, intuitively displaying the correlation patterns existing in the equipment group and the degree to which health status affects these patterns.

[0056] Example 2: See Figure 3The determination of the power generation strategy optimization objective is based on a comprehensive analysis of external instructions and internal conditions. Grid dispatch instructions convey the macro-level needs of the power system. These instructions may include power output targets for specific time periods, frequency regulation accuracy requirements, or obligations to participate in grid peak shaving. Meanwhile, the assessment of the energy storage system status involves real-time monitoring of various energy storage units such as battery packs, flywheels, or compressed air energy storage. It is necessary to obtain key parameters such as their current state of charge, health level, maximum charge and discharge rate, and cycle life degradation coefficient. The power generation strategy optimization objective formed after integrating these internal and external information is usually a composite objective function that includes multiple dimensions such as economy, safety, and reliability. Based on the above optimization objectives, a multi-objective optimization algorithm is used to generate the initial set of power generation strategies. This algorithm needs to handle the trade-offs between different objectives, such as the potential inherent conflict between the objectives of minimizing power generation costs and minimizing carbon emissions. The non-dominated sorting genetic algorithm is one of the commonly used choices. Its workflow includes initializing the population, crossover and mutation to generate new individuals, and iterative screening based on non-dominated levels and crowding. Each individual in the population encodes a complete candidate power generation strategy. Its gene sequence defines the power output setpoint sequence of different clean energy power generation equipment (such as photovoltaic arrays, wind turbines, and biogas generator sets) and the charging and discharging power plan of the energy storage system at each time segment within the scheduling cycle.

[0057] The construction of each candidate power generation strategy begins with a fine division of the dispatch cycle. The time range covered by the strategy is evenly divided into multiple consecutive time intervals, each typically set to five or fifteen minutes, depending on the station's operational requirements and the frequency of grid dispatch instructions. Within each time interval, the strategy needs to explicitly specify the planned output power value of each power generation device within the station. These power values ​​constitute the core of the power generation allocation scheme. Determining the power values ​​requires considering not only the device's nameplate capacity and current actual generating capacity but also ultra-short-term wind and solar power forecast data. For example, the planned power of a photovoltaic array needs to refer to the irradiance forecast curve for the next few minutes, while the planned power of a wind turbine is set based on wind speed forecast results. The power generation allocation scheme must meet the total output requirements issued by the grid. Within any time interval, the sum of the planned power of all power generation devices should equal the total power value that the station needs to transmit to the grid during that period. This total power value originates from grid dispatch instructions or the station's own power generation plan. The scheme development process must adhere to equipment operation safety constraints, including power ramp-up rate limits for individual devices, minimum technical output limits, and equipment start-up and shutdown timing requirements. For complex equipment such as combined cycle generator sets, the stability requirements of their operating points must also be considered. The energy storage dispatch scheme and the power generation allocation scheme are developed simultaneously, requiring millisecond-level precision coordination. The energy storage dispatch scheme details the operating mode of energy storage in each time interval—whether it's charging, discharging, or standby—and determines the corresponding power level. A negative charging power value indicates that the energy storage system absorbs energy from the grid or power plant, while a positive discharging power value indicates that energy is released to the grid. The power level setting must consider the capacity limitations of the energy storage converter and the charge / discharge rate limitations of the battery itself. The scheme also needs to ensure that the state of charge of the energy storage returns to the preset target range at the end of the entire dispatch cycle, reserving sufficient regulation capacity for subsequent dispatch cycles.

[0058] Candidate strategies also need to include backup capacity allocation schemes to cope with emergencies. A portion of the generation capacity should be reserved in the power generation allocation to smooth out fluctuations in renewable energy or respond to temporary grid adjustment needs. Backup capacity can be allocated proportionally among generating units or designated to specific units or energy storage systems. The strategy document will clearly indicate the size and priority of backup capacity for each time interval to ensure rapid response when needed. Each candidate strategy is ultimately encoded as a structured data object, containing multiple data fields such as timestamp sequences, equipment power setpoint matrices, energy storage dispatch command vectors, and backup capacity configuration parameters. This data can be directly sent to the underlying controller for execution through the application programming interface of the site monitoring system. The strategy is stored in a standardized format to facilitate subsequent evaluation, comparison, and retrospective analysis. A complete strategy document also includes version identifiers, generation timestamps, and relevant boundary condition descriptions, forming a traceable and executable detailed action plan.

[0059] After generating a rich set of candidate strategies, the strategy evaluation phase begins. The core of this phase is to match the previously constructed power plant operation characteristic matrix with each candidate power generation strategy. The power plant operation characteristic matrix dynamically reflects the correlation and health status between equipment. Calculating the strategy matching degree is a crucial step, quantifying the degree of fit between a specific strategy and the actual operating characteristics of the current power plant. For example, if the characteristic matrix shows a strong correlation between certain wind turbines (units in the same wind farm), then a strategy requiring a drastic difference in the output power of these units may have a low matching degree, as this contradicts their natural cooperative operating characteristics. Matching degree calculation can be achieved using various similarity or distance metrics. Essentially, it assesses whether the operating mode required by the strategy is consistent with the historical or current operating mode revealed by the characteristic matrix. The calculation process needs to consider the multidimensional characteristics of the matrix, potentially involving projection or transformation of the feature vectors before similarity comparison. The result is a numerical score used to characterize the strategy's adaptability to the unique operating environment of the power plant.

[0060] After obtaining the initial matching score, a feasibility assessment of each candidate power generation strategy is required based on a series of pre-set hard and soft constraints. These constraints are thresholds to ensure the safe and reliable execution of the strategy. Hard constraints include, but are not limited to: the upper and lower limits of the absolute power output of each power generation device, the maximum charging and discharging power and capacity limits of the energy storage system, the voltage and frequency stability range of the grid connection point, and the emission standards stipulated by environmental protection regulations. Any candidate strategy that violates any hard constraint will be marked as infeasible or given a very low feasibility score. Soft constraints involve more considerations of operating efficiency, equipment lifespan, and economic benefits. For example, prioritizing the use of equipment in better health, avoiding the operation of the energy storage system under extreme charging conditions to extend its lifespan, or tending to allow more efficient units to bear more load. The evaluation process needs to comprehensively consider the strategy matching score and the satisfaction of all these constraints, ultimately generating a comprehensive strategy evaluation result for each candidate strategy. This result may be a multi-dimensional vector or a comprehensive score, which clearly indicates the advantages and disadvantages of each strategy, providing a direct and quantitative decision-making basis for the final selection of the optimal strategy. The entire generation and evaluation process constitutes a complete decision support chain, from target analysis to automatic strategy generation, and then to fine-grained matching and feasibility verification with the actual state of the power station. This implementation method ensures that the set of strategies considered not only meets the macro-dispatch needs of the power grid, but is also deeply rooted in the real-time operating conditions and dynamic interaction between equipment within the clean energy power station.

[0061] Example 3: The screening process begins with the evaluation results of each candidate strategy in the initial power generation strategy set generated in the previous stage. These results typically include multi-dimensional scoring data, such as the matching degree between the strategy and the station's operating characteristics, economic indicators, technical feasibility scores, and the estimated impact on equipment lifespan. Normalizing and standardizing these heterogeneous evaluation data is a prerequisite for effective comparison. A comprehensive weight allocation method based on entropy weighting or analytic hierarchy process is used to combine multiple evaluation indicators into a comparable comprehensive evaluation value, thereby establishing a unified ranking benchmark for all candidate strategies. After sorting the initial power generation strategy set in descending or ascending order according to the comprehensive evaluation value, the candidate power generation strategy with the highest ranking, i.e., the best comprehensive evaluation value, is selected as the provisional optimal power generation strategy. This selection logic is based on its best performance in terms of its fit with the current station's operating characteristic matrix, its ability to meet various hard and soft constraints, and its overall benefits. However, this provisional optimal strategy has not yet considered the dynamic constraints and subtle mismatches that may exist in the energy storage system at the moment of execution. Therefore, a fine-tuning step based on the real-time status of the energy storage system must be introduced.

[0062] The fine-tuning process relies heavily on precise perception of the current operating status of the energy storage system. The data that needs to be acquired in real time includes the instantaneous charge and discharge power of the energy storage unit, the current state of charge (SOC), state of health (SOH), internal temperature, maximum allowable charge and discharge current, and remaining available capacity. The remaining available capacity refers not only to the absolute value of the energy currently stored, but also to the effective range of energy that can actually be dispatched and used within the strategy execution cycle, taking into account operating efficiency and lifespan degradation. Based on this real-time data, the instantaneous adjustment capability of the energy storage system is calculated. This adjustment capability quantitatively describes the plasticity of the energy storage to absorb or release power in the future to smooth fluctuations and track plans.

[0063] This regulatory capacity The following relationship can be used for quantitative evaluation: ,in: Indicates the period during which the strategy is executed. Internally, considering both charging and discharging efficiency With lifetime protection strategies in place, the actual energy capacity that the energy storage system can safely access is determined. The rated operating current specified by the energy storage unit manufacturer. It is the average operating current required by the energy storage system in the optimal power generation strategy before fine-tuning. It is a decay coefficient related to the chemical characteristics of the energy storage system, reflecting the degree to which deviations from the rated operating current affect system performance and lifespan. The exponential term is designed to penalize dispatch commands that require the energy storage system to deviate from its optimal operating range for extended periods, thus embedding considerations of long-term equipment reliability into the quantification of regulation capability.

[0064] Quantitative energy storage regulation capabilities have been achieved. Subsequently, based on this, targeted adjustments are made to the power allocation scheme in the provisional optimal power generation strategy. If calculations indicate that energy storage has sufficient regulation margin, fine-tuning may tend to make fuller use of energy storage to further optimize the operation of the entire power station. For example, the energy storage system may undertake more power fluctuation smoothing tasks, thereby allowing the generator units to operate at a more stable and efficient operating point, or to more accurately track the grid's dispatch instructions. Conversely, if the regulation capacity of energy storage is close to its safety boundary, the direction of fine-tuning shifts to reducing the burden on energy storage, allocating more power balancing tasks to the generator units themselves, or activating backup regulation resources. The specific operation of the adjustment may involve iterative optimization of the power setpoints of each generator unit in the strategy, ensuring that the power allocation scheme and the energy storage dispatch scheme can be coordinated and executed, forming an internally self-consistent and safely implementable final operating instruction. The fine-tuning process is a refined decision-making cycle. Its core objective is to ensure that the selected strategy, which is optimal at the macro level, can be precisely matched with the actual dynamic response capability of the energy storage system at the micro level. This prevents problems such as overload, over-discharge, or accelerated aging of the energy storage system caused by overly idealistic strategies. Ultimately, the optimal power generation strategy that is issued and implemented is not only excellent in theory, but also robust, feasible, and efficient in actual operation.

[0065] See Figure 4 The left subplot shows the comprehensive evaluation and ranking results of candidate power generation strategies. Each bar represents a comprehensive evaluation value for a strategy, calculated using the entropy weighting method by weighting four dimensions: matching degree, economic efficiency, technical feasibility, and equipment lifespan impact. The strategy marked with a red border is the current optimal choice, performing best in terms of fit with the site's operating characteristics, ability to meet constraints, and overall benefits. The strategy ranking provides a benchmark for subsequent fine-tuning. The right subplot compares the power allocation schemes of the energy storage system before and after fine-tuning. Dark bars represent the power allocation of the original optimal strategy, while light bars represent the adjustment results considering the real-time status of the energy storage system. The adjustment process is based on the calculation of the instantaneous regulation capability of the energy storage system, which comprehensively considers the remaining available capacity, charging and discharging efficiency, and the impact of the operating current deviating from the rated value on the system's lifespan. The text boxes in the charts list in detail the current key state parameters of the energy storage system, including regulation capability values, state of charge, and health status. These parameters directly determine the magnitude and direction of fine-tuning.

[0066] Example 4: The dynamic adjustment mechanism constitutes the feedback loop of the power generation strategy optimization closed-loop system. Its core lies in enabling the pre-defined static strategy to respond to unpredictable fluctuations that occur during the actual operation of the power station. Taking a hybrid power station containing wind turbines, photovoltaic arrays, and lithium-ion battery energy storage as an example, the power station is currently executing an optimal power generation strategy based on short-term wind and solar power forecasts. This strategy plans the power output targets of each power generation unit and the charging and discharging schedule of the energy storage within the next 15 minutes. After the strategy starts to be executed, the sensor network deployed on the wind turbine towers, photovoltaic inverters, energy storage converters, and key nodes of the power station works continuously, with a data acquisition frequency of up to the second level. The core operating status parameters monitored include: the real-time active power output, pitch angle change, and gearbox oil temperature of the wind turbines; the DC side current and voltage of the photovoltaic array and the AC output power of the inverter; the battery cluster voltage, current, state of charge, and internal temperature of the energy storage system; and the environmental monitoring station transmits data such as wind speed, wind direction, irradiance, and ambient temperature in real time. These massive amounts of real-time operational data are aggregated to the central controller via industrial Ethernet or private wireless networks within the site, forming a data set that is isomorphic to but updated with the initial data used when formulating the strategy.

[0067] Immediately after data acquisition, consistency verification and outlier removal are performed. For example, it compares whether sudden changes in wind turbine power at adjacent times match changes in wind speed. If the photovoltaic panel string current is found to be zero while the irradiance is good, the equipment fault diagnosis process may be triggered. Validated data is marked with a precise timestamp for updating the system's internal real-time database. Table 1 shows a comparison between the expected operating status at a specific moment (T0) and the updated operating data actually monitored shortly after (T1).

[0068] Table 1: Comparison between Expected Station Operation Strategies and Actual Monitoring Data

[0069] Device / Parameters Expected value of the strategy at time T0 Actual monitored value at time T1 Deviation Explanation #1 fan power 1500kW 1450kW Actual wind speed was lower than predicted, resulting in insufficient power output. #2 Photovoltaic Array Power 800kW 950kW Cloud movement caused a sudden increase in irradiance. Energy storage system SOC 65% 63% Load fluctuations caused the discharge to slightly exceed expectations. Main line grid connection power 2300kW 2400kW Total output deviates from the scheduled value

[0070] Based on the updated real-time operating data shown in the table above, the dynamic adjustment process is activated. The adjustment algorithm assesses the nature and magnitude of these deviations, determining whether they are short-term random fluctuations or indicate a trend change. For example, if the wind turbine power is consistently lower than expected and the wind speed sensor confirms a decrease in wind speed, it is judged as a trend deviation. The goal of dynamic adjustment is to generate a new power generation strategy that matches the actual state of the power plant at time T1, ensuring that the power generation strategy and the power plant's operating state remain synchronized and optimized. The adjustment logic does not simply overwrite the expected value of the original strategy with the actual value, but performs a new round of rapid optimization calculations. This process borrows the rolling optimization idea of ​​model predictive control, but with a shorter calculation cycle and a simpler model. The algorithm uses the actual state at time T1 as the new initial conditions, such as the exact state of charge of the current energy storage and the actual output capacity of each power generation device, and incorporates the latest ultra-short-term wind and solar power prediction data. Under the premise that the overall goal of the original grid dispatch instructions remains unchanged (such as the total on-grid electricity within the cycle), it re-solves a simplified version of the optimization problem and quickly calculates a new set of device power setpoints covering the next adjustment cycle (such as the next 5 minutes).

[0071] The dynamic adjustment module needs sufficient intelligence to distinguish between disturbances of different natures. For second-level random fluctuations in wind turbine power caused by turbulence, the adjustment module may choose not to act, relying instead on the automatic frequency response of the energy storage system to smooth the flow and avoid the impact of frequent strategy switching on the equipment. However, for significant drops in photovoltaic power lasting several minutes due to large-area cloud cover, the adjustment module will initiate re-optimization calculations. It may instruct the energy storage system to increase discharge power to compensate for the photovoltaic deficit, or appropriately increase the available potential of wind power generation (if wind speed conditions permit). At the same time, it will recalculate the charging and discharging plan for energy storage in subsequent periods to ensure that the state of charge target at the end of the cycle can still be achieved. The dynamic adjustment process is continuously cyclical. Whenever a new batch of real-time data arrives, the system will re-execute the "monitor-compare-judge-adjust" process. This design makes the power generation strategy no longer a fixed sequence of commands, but a breathing, flexible, and dynamic plan that closely follows every pulse of the internal and external conditions of the power plant, thereby achieving continuous optimization of the power generation process in complex and ever-changing environments.

[0072] Example 5: The station's operational feature matrix and strategy evaluation model are updated based on updated real-time operational data. The updated model is then used to re-evaluate the current execution strategy, thereby determining whether to generate and issue a new strategy. For example, a large wind-solar hybrid power station is executing an optimal power generation strategy aimed at stable output. This strategy was formulated based on relatively stable weather conditions an hour ago, assuming wind speeds fluctuate around 8 meters per second and uniform irradiance. However, weather radar echoes show a dense cloud cluster rapidly moving towards the station's area, and wind speed sensors begin recording continuously increasing gusts. As a result, the real-time operational data undergoes a systematic change. The wind turbine SCADA system reports significant oscillations in power output, while the photovoltaic inverter output begins to decline sharply due to cloud cover. The rate of change of the energy storage system's state of charge deviates from expectations. These updated data are immediately sent to the data processing module, and the station operation characteristic matrix is ​​recalculated according to the method defined in Example 1. The matrix elements that originally represented the weak correlation between wind turbines and photovoltaics may be transformed into a moderate positive correlation because the output of both decreases synchronously. The vector describing the health status of equipment may also be updated because some equipment is operating under harsh conditions.

[0073] After recalculating the power plant operation characteristic matrix, the strategy evaluation model enters the update process. This model may contain cluster centers reflecting typical operating ranges of the equipment, regression coefficients characterizing the mapping relationship between environmental variables and power output, or weight parameters of a neural network model. Model updates are achieved through incremental learning algorithms, which absorb the latest patterns contained in the new operation characteristic matrix and gradually adjust their internal parameters without requiring complete retraining. For example, the model may learn that under the combination of strong gusts and rapidly changing irradiance, the optimal response strategy of the energy storage system should be more forward-looking. The updated strategy evaluation model's decision boundaries and evaluation criteria already reflect the current dynamic operating environment of the power plant. The updated strategy evaluation model is then used to re-evaluate the currently implemented optimal power generation strategy. This process is similar to a "stress test" of a predetermined plan. The evaluation input includes the updated power plant operation characteristic matrix and detailed parameters of the current strategy. The model outputs a comprehensive score on the adaptability of the strategy in the new environment. It may be found that a previously excellent strategy receives a significant drop in score under the influence of clouds and gusts because it did not reserve sufficient energy storage backup for sudden drops in photovoltaic power, or because the power fluctuations of the wind turbines were too strictly limited, leading to wind curtailment.

[0074] After reassessment, the system generates an adjusted power generation strategy. This strategy is a partial modification of the original strategy based on the new model and data, rather than a complete overhaul. The adjusted strategy may maintain consistency with the original strategy in overall objectives, but there may be differences in specific parameters. Comparing the differences between the power generation strategies before and after adjustment is a key step in decision-making. The system calculates a quantitative strategy adjustment magnitude, which may be a multi-dimensional distance metric, comprehensively considering the changes in the setpoints of each device in the power generation allocation scheme, the degree of deviation in the energy storage dispatch plan, and the expected changes in overall cost or efficiency indicators. The calculation of the strategy adjustment magnitude needs to avoid overreacting to normal fluctuations. If the calculated strategy adjustment magnitude does not exceed a preset threshold, it indicates that the change in the current operating state is still within the tolerance range of the original strategy, or that the adjusted strategy is not fundamentally different from the original strategy. In this case, the system may decide to continue implementing the original strategy or only apply fine-tuning, and record the evaluation results for model self-learning. This threshold is usually set comprehensively based on factors such as the regulation performance of the station equipment, grid assessment requirements, and economic benefits, such as setting an allowable total output change threshold or the maximum strategy change frequency per unit time.

[0075] If the strategy adjustment exceeds a preset threshold, a strategy regeneration process is triggered. This usually indicates a qualitative change in the operating environment, rendering the original strategy's foundation obsolete. The system will then revert to the initial strategy generation module, using the current instantaneous operating state as a new starting point. Combining this with the latest ultra-short-term forecasts, a multi-objective optimization algorithm will be run to regenerate a completely new optimal power generation strategy. This new strategy is logically a completely new solution, better suited to the actual situation currently faced by the power plant. After verification, the new strategy is immediately sent to the power plant monitoring system for execution, thus completing a full closed-loop control cycle from monitoring and evaluation to decision-making and execution.

[0076] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing power generation strategies at clean energy power plants, characterized in that, include: Collect real-time operating data of various power generation equipment in clean energy power plants, including power generation, environmental parameters and equipment status parameters; Based on the real-time operation data, a power station operation feature matrix is ​​constructed, which is used to characterize the dynamic correlation between different power generation equipment. Based on the preset power generation strategy optimization objective, an initial power generation strategy set is generated, which contains multiple candidate power generation strategies; The power station operation feature matrix is ​​input into the strategy evaluation model to perform an adaptive evaluation on each candidate power generation strategy in the initial power generation strategy set, and the strategy evaluation result is obtained. The optimal power generation strategy is selected based on the strategy evaluation results, and the optimal power generation strategy is then distributed to clean energy power plants for execution. The construction of the station operation feature matrix based on the real-time operation data includes: Extract the power generation sequence and environmental parameter sequence from the real-time operating data, and calculate the power correlation coefficient between different power generation devices; A device correlation matrix is ​​constructed based on the power correlation coefficient, and the device correlation matrix is ​​used to describe the cooperative operation relationship between power generation equipment; The equipment health vector is generated by combining the equipment status parameters, and the equipment correlation matrix is ​​fused with the equipment health vector to form a site operation feature matrix; The step of inputting the power station operation feature matrix into the strategy evaluation model and performing an adaptive evaluation on each candidate power generation strategy in the initial power generation strategy set includes: The station operation feature matrix is ​​matched with each candidate power generation strategy, and the strategy matching degree is calculated; Based on the strategy matching degree and preset constraints, the feasibility of each candidate power generation strategy is evaluated, and strategy evaluation results are generated.

2. The method for optimizing power generation strategy at clean energy power plants according to claim 1, characterized in that, The step of generating an initial set of power generation strategies based on a preset power generation strategy optimization objective includes: Obtain grid dispatch instructions and energy storage system status from clean energy power plants to determine power generation strategy optimization objectives; Based on the power generation strategy optimization objective, an initial power generation strategy set is generated using a multi-objective optimization algorithm. Each candidate power generation strategy in the initial power generation strategy set includes a power generation allocation scheme and an energy storage scheduling scheme.

3. The method for optimizing power generation strategy at clean energy power plants according to claim 2, characterized in that, The process of selecting the optimal power generation strategy based on the strategy evaluation results includes: The initial power generation strategy set is sorted according to the strategy evaluation results, and the candidate power generation strategy with the highest matching degree is selected as the optimal power generation strategy. The optimal power generation strategy is fine-tuned based on the energy storage system status to ensure coordinated execution of the power generation allocation scheme and the energy storage scheduling scheme.

4. The method for optimizing power generation strategy at clean energy power plants according to claim 3, characterized in that, The fine-tuning of the optimal power generation strategy based on the energy storage system status includes: Obtain the current charge / discharge state and remaining capacity of the energy storage system, and calculate the energy storage regulation capability; The power allocation scheme in the optimal power generation strategy is adjusted according to the energy storage regulation capability to ensure that the energy storage system maintains stable operation during the execution of the optimal power generation strategy.

5. The method for optimizing power generation strategy at clean energy power plants according to claim 4, characterized in that, The method further includes: During the execution of the optimal power generation strategy, the operating status of the clean energy power station is monitored in real time, and updated real-time operating data is collected. The optimal power generation strategy is dynamically adjusted based on the updated real-time operating data to ensure that the power generation strategy is synchronized and optimized with the station's operating status.

6. The method for optimizing power generation strategy at clean energy power plants according to claim 5, characterized in that, The step of dynamically adjusting the optimal power generation strategy based on the updated real-time operating data includes: The station operation feature matrix is ​​recalculated based on the updated real-time operation data, and the strategy evaluation model is updated accordingly. The updated strategy evaluation model is used to re-evaluate the currently implemented optimal power generation strategy and generate an adjusted power generation strategy.

7. The method for optimizing power generation strategy at clean energy power plants according to claim 6, characterized in that, The re-evaluation of the currently implemented optimal power generation strategy using the updated strategy evaluation model includes: Compare the differences in power generation strategies before and after the adjustment, calculate the adjustment range, and if the adjustment range exceeds a preset threshold, regenerate the optimal power generation strategy and issue it for execution.

8. A clean energy power plant power generation strategy optimization system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the clean energy power generation strategy optimization method according to any one of claims 1 to 7.

Images