A new energy power station intelligent operation and maintenance scheduling and resource optimization method and system

Abstract

The application discloses a new energy power station intelligent operation and maintenance scheduling and resource optimization method and system, belongs to the technical field of data processing and management, and comprises the following steps: acquiring real-time power generation data, equipment sensor data, weather forecast data, power grid scheduling instructions and power market electricity price information of a new energy power station, and generating a unified data matrix through space-time alignment processing to execute bidirectional feedback prediction and generate a corrected power generation plan; generating a collaborative scheduling decision through a multi-target dynamic weighting algorithm and generating a personnel dispatching scheme and a material allocation path through real-time path planning; generating a historical decision dataset based on an execution result and an actual execution deviation, and adjusting prediction parameters and scheduling parameters through an incremental learning model. The closed-loop optimization method combining data fusion, collaborative prediction, multi-target scheduling and adaptive learning can realize global collaborative optimization of power generation, operation and maintenance, energy storage and the market and the like, and improves the economic benefits and the intelligent level of the overall operation of the power station.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing and management, in particular to a new energy power station intelligent operation and maintenance scheduling and resource optimization method and system. BACKGROUND

[0002] As an important part of modern power systems, the core business of new energy power stations is to generate electricity using renewable energy such as wind and solar energy, and to ensure safe, stable and economic operation through intelligent operation and maintenance. With the expansion of power station scale and the deepening of power market reform, how to effectively manage the massive data inside and outside the power station, and on this basis, realize the collaborative optimization of power generation planning, equipment maintenance, energy storage scheduling and market transactions and other business links, has become the key to improving the business competitiveness of power stations.

[0003] In the prior art, the operation and management of new energy power stations usually relies on multiple independent software systems. For example, the data acquisition and monitoring control system is responsible for collecting real-time operation data, the power prediction system provides power generation prediction based on weather forecasts, and the work order management system is used to arrange operation and maintenance tasks. When making scheduling decisions, the operator often first formulates a power generation plan, and then arranges maintenance according to the plan and equipment status. The decision-making process of each link is relatively independent, and there is a lack of unified optimization target and information interaction. For the allocation of operation and maintenance resources, it is also often assigned by relying on human experience, and the scheduling efficiency and response speed are limited.

[0004] However, the data standards and time scales between different systems are not the same, forming a data island, making it difficult to form a unified data view to support global decision-making. Secondly, power prediction is mostly in an open-loop mode, without fully integrating the real-time health status of the equipment, resulting in a deviation between the prediction results and the actual power generation capacity, and the inability to self-correct according to historical deviations. More importantly, the decision-making links of power generation, operation and maintenance, and energy storage are disconnected, often leading to resource conflicts and benefit offset, such as unnecessary shutdown maintenance at high electricity prices, or the expansion of power generation losses due to the lack of timely operation and maintenance response, which limits the improvement of the overall economic benefit of the power station. SUMMARY

[0005] To solve the above problems, the present application provides a new energy power station intelligent operation and maintenance scheduling and resource optimization system, which adopts a closed-loop optimization method combining data fusion, collaborative prediction, multi-objective scheduling and adaptive learning, and can realize the global collaborative optimization of power generation, operation and maintenance, energy storage and market, etc. multiple elements, and improve the economic benefit and intelligent level of the overall operation of the power station.

[0006] The above-mentioned object can be achieved by the following scheme:

[0007] The application discloses a new energy power station intelligent operation and maintenance scheduling and resource optimization method, which comprises the following steps: acquiring real-time power generation data, equipment sensor data, weather forecast data, power grid scheduling instructions and power market electricity price information of a new energy power station, and generating a unified data matrix through space-time alignment processing; based on the unified data matrix, performing bidirectional feedback prediction, which comprises generating long-term prediction results and short-term prediction results, and correcting the long-term prediction results by using the short-term prediction results to generate a corrected power generation plan; acquiring equipment maintenance urgency and energy storage charge state, and combining the corrected power generation plan and the power market electricity price information to generate a collaborative scheduling decision comprising a power station output plan, an energy storage charging and discharging strategy and an operation and maintenance task allocation instruction through a multi-objective dynamic weighting algorithm; based on the operation and maintenance task allocation instruction, acquiring technician position information and spare parts inventory position information, and generating a personnel dispatching scheme and a material allocation path through real-time path planning; recording the execution result of the collaborative scheduling decision and calculating the actual execution deviation, generating a historical decision dataset based on the execution result and the actual execution deviation, and adjusting the prediction parameters and the scheduling parameters through an incremental learning model by using the historical decision dataset.

[0008] Optionally, the generating a unified data matrix comprises: acquiring real-time power generation data, equipment sensor data, weather forecast data, power grid scheduling instructions and power market electricity price information of a new energy power station to obtain a basic data set; extracting timestamp and geographic coordinate information from the basic data set; synchronizing the data in the basic data set to a unified time step based on the timestamp, and mapping the data in the basic data set to a unified space reference based on the geographic coordinate information to obtain a processed basic data set; structurally integrating the processed basic data set to generate a unified data matrix.

[0009] Optionally, the method further comprises: extracting temperature parameters and vibration parameters from the equipment sensor data; analyzing the temperature parameters and vibration parameters by using a preset health state threshold to generate an equipment health state; calculating a fault risk coefficient as the equipment maintenance urgency based on the equipment health state and the time variation rates of the temperature parameters and vibration parameters.

[0010] Optionally, the generating a corrected power generation plan comprises: extracting weather trends and historical power generation rules from the unified data matrix, and performing long-term prediction to generate long-term prediction results; extracting real-time equipment state data from the unified data matrix, and performing short-term prediction to generate short-term prediction results; calculating a prediction error between the long-term prediction results and the short-term prediction results, and performing back propagation calibration on the long-term prediction results by using the prediction error to generate a corrected power generation plan.

[0011] Optionally, the generating the short-term prediction result comprises: extracting real-time device state data from the unified data matrix; correcting the real-time device state data by using the device maintenance urgency; generating a short-term prediction result based on the corrected real-time device state data.

[0012] Optionally, the generating the coordinated scheduling decision comprising a power station output plan, a storage charging and discharging strategy, and an operation and maintenance task allocation instruction comprises: obtaining the device maintenance urgency and the storage charge state; constructing a multi-objective function with power generation efficiency, device life, power grid demand, and economy as optimization objectives; taking the corrected power generation plan, the device maintenance urgency, the storage charge state, and the power market price information as input variables of the multi-objective function; solving the multi-objective function to obtain an optimal balanced solution, and analyzing the optimal balanced solution into the power station output plan, the storage charging and discharging strategy, and the operation and maintenance task allocation instruction to obtain the coordinated scheduling decision.

[0013] Optionally, the generating the personnel dispatching scheme and the material allocation path comprises: analyzing the operation and maintenance task allocation instruction to determine required technical personnel skills and spare part types; obtaining technical personnel location information and spare part inventory location information; combining the technical personnel location information and the technical personnel skills to screen out candidate technical personnel, and combining the spare part inventory location information and the spare part types to screen out candidate spare part warehouses; combining the candidate technical personnel and the candidate spare part warehouses, and performing path optimization calculation to select a combination according to a total response time, to generate the personnel dispatching scheme and the material allocation path.

[0014] Optionally, the adjusting the prediction parameters and the scheduling parameters comprises: recording an execution result of the coordinated scheduling decision and calculating an actual execution deviation, generating a historical decision dataset based on the execution result and the actual execution deviation; inputting the historical decision dataset into the incremental learning model for training to identify a systematic deviation mode; generating a deviation adjustment amount according to the systematic deviation mode; and applying the deviation adjustment amount to update the prediction parameters and the scheduling parameters.

[0015] Optionally, the method further comprises: judging whether a sudden drop in future power generation occurs based on the corrected power generation plan; if yes, triggering an emergency response, calculating a predicted power gap based on the corrected power generation plan, and updating the storage charging and discharging strategy according to the power gap; synchronously adjusting the operation and maintenance task allocation instruction, and generating an adaptive scheduling instruction in combination with the updated storage charging and discharging strategy.

[0016] Based on the same inventive concept, the application also provides a new energy power station intelligent operation and maintenance scheduling and resource optimization system, the system comprises: a data fusion module for acquiring real-time power generation data, equipment sensor data, weather forecast data, power grid scheduling instructions and power market electricity price information of the new energy power station, and generating a unified data matrix through space-time alignment processing; a collaborative prediction module for executing bidirectional feedback prediction based on the unified data matrix, the bidirectional feedback prediction including generating long-term prediction results and short-term prediction results, and correcting the long-term prediction results using the short-term prediction results to generate a corrected power generation plan; an intelligent scheduling module for acquiring equipment maintenance urgency and energy storage charge state, combining the corrected power generation plan and the power market electricity price information, and generating a collaborative scheduling decision including power station output plan, energy storage charging and discharging strategy and operation and maintenance task allocation instruction through a multi-objective dynamic weighting algorithm; a resource optimization module for acquiring technician location information and spare parts inventory location information based on the operation and maintenance task allocation instruction, and generating a personnel dispatching scheme and material allocation path through real-time path planning; and an adaptive learning module for recording execution results of the collaborative scheduling decision and calculating actual execution deviation, generating a historical decision dataset based on the execution results and the actual execution deviation, and adjusting prediction parameters and scheduling parameters through an incremental learning model using the historical decision dataset.

[0017] Compared with the prior art, the application has the following advantages:

[0018] 1. The application constructs a data fusion and closed-loop learning mechanism, aligns and deeply integrates scattered power station data, meteorological information and market signals in space and time, and continuously corrects systematic deviation using an incremental learning model, significantly improving prediction accuracy and the reliability of decision-making basis, and providing a high-quality data basis for fine operation and management of power stations;

[0019] 2. The application proposes a global collaborative optimization scheduling method, which integrates multiple mutually restrictive management objectives such as power generation efficiency, equipment life, power grid demand and economy into a unified decision-making framework through a multi-objective dynamic weighting algorithm, realizes deep coupling and integrated scheduling of power generation, operation and maintenance, energy storage and other businesses, overcomes the local optimum and global suboptimal problems caused by traditional step-by-step decision-making, and comprehensively improves the comprehensive operation benefit of power station assets;

[0020] 3. The application realizes dynamic and accurate allocation of operation and maintenance resources, generates optimal personnel dispatching and material allocation schemes by analyzing operation and maintenance requirements in the collaborative scheduling instruction and combining real-time location information and path planning of personnel and materials, significantly shortens fault response and repair time, reduces operation and maintenance cost and power generation loss, and enhances the agility and economy of power station operation and management.

[0021] Other features and advantages of the present application will be set forth in the descriptions that follow, and in part will be apparent from the description, or can be learned by practice of the application. The purposes and other advantages of the present application will be realized and attained by the structures particularly pointed out in the written description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0023] Figure 1 is a flowchart of a new energy power station intelligent operation and maintenance scheduling and resource optimization method according to an embodiment of the present application.

[0024] Figure 2 is an effect diagram of continuously optimizing prediction accuracy according to an embodiment of the present application.

[0025] Figure 3 is a structural diagram of a new energy power station intelligent operation and maintenance scheduling and resource optimization system according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0027] With reference to Figure 1 An embodiment of the present application proposes a new energy power station intelligent operation and maintenance scheduling and resource optimization method, which adopts a closed-loop optimization method combining data fusion, collaborative prediction, multi-objective scheduling and adaptive learning, can realize global collaborative optimization of multiple factors such as power generation, operation and maintenance, energy storage and market, and improve the economic benefit and intelligent level of the overall operation of the power station.

[0028] The method of the embodiment specifically includes:

[0029] Real-time power generation data, equipment sensor data, weather forecast data, power grid scheduling instructions and power market price information of the new energy power station are acquired, and a unified data matrix is generated through spatio-temporal alignment processing;

[0030] Based on the unified data matrix, bidirectional feedback prediction is performed, which includes generating long-term prediction results and short-term prediction results, and correcting the long-term prediction results with the short-term prediction results to generate a corrected power generation plan;

[0031] The equipment maintenance urgency and the energy storage charge state are obtained, and the corrected power generation plan and the power market price information are combined to generate a collaborative scheduling decision including a power plant output plan, a storage charging and discharging strategy, and an operation and maintenance task allocation instruction through a multi-objective dynamic trade-off algorithm;

[0032] Based on the operation and maintenance task allocation instruction, technician location information and spare parts inventory location information are obtained, and a personnel dispatching scheme and a material allocation path are generated through real-time path planning;

[0033] The execution results of the collaborative scheduling decision are recorded and the actual execution deviation is calculated, the historical decision data set is generated based on the execution results and the actual execution deviation, and the prediction parameters and the scheduling parameters are adjusted through the incremental learning model using the historical decision data set.

[0034] Specifically, through data fusion and bidirectional feedback prediction, the accuracy of power generation prediction is greatly improved, providing a reliable basis for power plants participating in the power market and accepting grid dispatching. The core collaborative scheduling mechanism realizes the transition from independent decision of each subsystem to global optimization decision, effectively balancing the relationship between short-term power generation income and long-term equipment health and grid stability, and avoiding the overall loss that may be caused by local optimal decision. Real-time optimization of operation and maintenance resources greatly shortens the fault response and processing time, and improves the availability of equipment. Finally, the adaptive learning closed loop ensures that the system can continuously adapt to changes in the external environment and internal state, so that the operation and maintenance scheduling strategy of the power plant always remains in a near-optimal state, thereby fundamentally enhancing the competitiveness and sustainable operation ability of new energy power plants.

[0035] Optionally, the generating a unified data matrix comprises:

[0036] Obtaining real-time power generation data, equipment sensor data, weather forecast data, grid dispatching instructions and power market price information of the new energy power plant to obtain a basic data set;

[0037] Extracting timestamp and geographic coordinate information from the basic data set;

[0038] Synchronizing the data in the basic data set to a unified time step based on the timestamp, and mapping the data in the basic data set to a unified spatial reference based on the geographic coordinate information to obtain a processed basic data set;

[0039] The processed basic data set is structured and integrated to generate a unified data matrix.

[0040] Specifically, the process of generating a unified data matrix aims to standardize and fuse new energy power station related data with diverse sources and formats, providing a high-quality data foundation for subsequent prediction and scheduling. The process first involves data collection, where the system obtains raw information constituting the basic data set from multiple information sources. These information includes but is not limited to real-time power generation data recording power station real-time output, device sensor data reflecting device operating status, weather forecast data predicting future weather conditions, scheduling instructions from the power grid, and electricity market price information reflecting market value. After completing data collection, the system deeply analyzes the basic data set to extract the core spatio-temporal attributes of each data. Specifically, the system identifies and extracts the timestamp information embedded in each data record and the geographic coordinate information that can determine its physical source. Then, based on these spatio-temporal attributes, key spatio-temporal alignment processing is performed. In the time dimension, the system takes a preset uniform time step, such as 15 minutes, as the reference to synchronize the data in the basic data set with different collection frequencies. For data with a collection frequency higher than the reference, such as second-level sensor data, the mean or characteristic value within the time window can be used for aggregation; for data with a frequency lower than the reference, such as hourly weather forecasts, interpolation algorithms can be used to generate data points at corresponding time steps, ensuring that all data are aligned in time series. In the spatial dimension, the system maps all data to a unified spatial reference based on geographic coordinate information, such as associating specific wind turbines or photovoltaic arrays' sensor data to their unique device IDs and on-site coordinates, ensuring the uniqueness and comparability of data in physical space. After spatio-temporal alignment, the processed basic data set is obtained. The last step is structured integration, where the system organizes all data items in the processed basic data set according to a unified format to generate a multi-dimensional unified data matrix. The matrix can be represented as follows:

[0041] ,

[0042] where, represents the final unified data matrix. represents the data points in the matrix. Index i represents the spatial reference identifier, such as the specific power generation device number, which is obtained by analyzing the geographic coordinate information and mapping it to the device topology of the power station. Index j represents the unified time step sequence number, which is obtained after synchronizing the original timestamp. Index k represents the feature type of the data, such as active power, device temperature, wind speed, and electricity price, which is determined by the nature of the original data source. Each element is the quantified value of the kth feature of device i at time j.

[0043] Optionally, the method further comprises:

[0044] extracting temperature parameters and vibration parameters from the device sensor data;

[0045] analyzing the temperature parameters and vibration parameters using preset health state thresholds to generate a device health state;

[0046] calculating a failure risk coefficient as a device maintenance urgency based on the device health state and time change rates of the temperature parameters and vibration parameters.

[0047] In particular, the method for calculating the device maintenance urgency in the present application aims to convert the original device sensor data into key indicators guiding operation and maintenance decisions. The process begins with the directional extraction of core parameters closely related to the mechanical and electrical state of the device from the unified data matrix, namely the temperature parameters and vibration parameters. These parameters are direct physical quantities reflecting the health status of rotating equipment such as generators or gearboxes. Then, the system compares and analyzes the real-time acquired temperature parameters and vibration parameters with preset health state thresholds to generate an evaluation of the device health state. The preset health state thresholds are multi-level thresholds determined according to device design specifications, industry standards, and historical data statistical analysis of long-term operation, such as normal operation range, pre-warning range, and dangerous range. By judging the interval in which the current parameter value is located, the system can qualitatively or quantitatively divide the device health state, such as outputting "healthy", "attention", or "warning" levels. Then, in order to realize the vectorization and dynamic risk prediction from static evaluation, the system not only considers the current device health state, but also introduces the time change rate of the parameters as a key consideration dimension. The system calculates the change rate of the temperature parameters and vibration parameters within consecutive time steps. Finally, by combining the level of the device health state and the change trend of the key parameters, the failure risk coefficient is calculated and used as the device maintenance urgency. The calculation process can be represented by the following formula:

[0048] ,

[0049] wherein, represents the final generated device maintenance urgency, which is a quantitative numerical value used to sort the priority of operation and maintenance tasks. is a discrete score obtained according to the device health state rating such as healthy, attention, and warning, which is directly assigned by the system based on the severity of the health state. and represent the time change rates of the temperature parameters and vibration parameters, respectively, which are obtained by calculating the difference of parameter values within adjacent time steps. The function is a nonlinear mapping function used to amplify the impact of rapidly changing parameter trends on risk, i.e. the faster the parameter changes, the larger the function output value. , , are preset weight coefficients, which respectively define the importance of the current state of the device, the temperature change trend and the vibration change trend in the comprehensive risk assessment, and the weight coefficients are trained based on expert experience and historical failure data.

[0050] Optionally, the generating the revised generation plan comprises:

[0051] extracting weather trends and historical generation rules from the unified data matrix, and performing long-term prediction to generate a long-term prediction result;

[0052] extracting real-time device state data from the unified data matrix, and performing short-term prediction to generate a short-term prediction result;

[0053] calculating a prediction error between the long-term prediction result and the short-term prediction result, and using the prediction error to back-propagation calibration of the long-term prediction result to generate a revised generation plan.

[0054] Specifically, first, long-term prediction is performed, the system extracts key features affecting the generation capacity in the future longer period from the previously generated unified data matrix, mainly weather trends and historical generation rules. Weather trends include meteorological forecast data such as wind speed, wind direction and solar irradiance in the future hours to days, while historical generation rules are the actual generation performance of the power plant under similar weather and seasonal conditions. Based on these data, the system uses time series analysis or machine learning model to generate a preliminary long-term prediction result, which depicts the ideal generation capacity curve of the power plant in the future dispatching cycle. At the same time, the system performs parallel short-term prediction to capture the actual generation potential at the present and immediate future. The system extracts real-time device state data from the unified data matrix to generate a short-term prediction result, which represents the maximum possible output of the power plant in the current and very short time window. The key of the two-way feedback prediction is to use the short-term prediction result to dynamically calibrate the long-term prediction result. The system calculates the prediction error between the long-term prediction result and the short-term prediction result at the current time point. This error essentially reflects the generation capacity deviation caused by real-time device state changes and other factors that are not fully captured by the long-term model. Subsequently, instead of simply replacing, the system uses a back-propagation calibration mechanism to revise the entire long-term prediction result. This calibration can be understood as a bias adjustment, and its influence will decay as the prediction time elapses. The revision process can be represented by the following formula:

[0055] ,

[0056] wherein, This represents the revised power generation plan value at time t in the future. This represents the value of the original long-term prediction at time t. It is a short-term forecast result at the current moment, while It is the result of long-term prediction at the current moment. The difference between the two values ​​is the prediction error at the current moment. It is a time-varying weight decay coefficient, whose value is close to 1 at the current time and gradually decreases to 0 as t increases. This ensures that the calibration effect is most significant in the near term and smoothly transitions to the original long-term forecast in the long term, avoiding excessive impact of transient disturbances on the entire planning cycle. The final result is... The sequence is a modified power generation plan that integrates macro-meteorological trends and the real-time status of micro-equipment.

[0057] Optionally, the generation of short-term prediction results includes:

[0058] Extract real-time device status data from the unified data matrix;

[0059] The real-time equipment status data is corrected based on the urgency of equipment maintenance.

[0060] Short-term predictions are generated based on the corrected real-time device status data.

[0061] Specifically, the method for generating short-term prediction results in this invention deeply optimizes the assessment of real-time power generation capacity within a bidirectional feedback prediction framework. This process first extracts real-time equipment status data directly related to the current power generation capacity from a unified data matrix. This data includes, but is not limited to, frequently updated operating parameters such as inverter DC-side voltage and current, wind turbine blade pitch angle, and generator speed. After obtaining the raw real-time equipment status data, the system does not directly use it for prediction. Instead, it incorporates a pre-calculated equipment maintenance urgency level to dynamically correct these raw data. The core idea of ​​this step is that the health status of the equipment directly limits its theoretical maximum output; therefore, this limitation must be quantified and reflected in the prediction input. The system employs a derating method to convert the equipment maintenance urgency level into an adjustment coefficient for a key performance parameter, thereby generating corrected real-time equipment status data. For example, the correction process for a key performance parameter can be expressed as the following formula:

[0062] ,

[0063] in, These are corrected real-time device status parameters, such as corrected maximum available power or conversion efficiency. These are raw, real-time device status parameters extracted directly from a unified data matrix. is the previously calculated equipment maintenance urgency representing the risk of equipment failure. is a preset upper limit of the urgency, used to normalize the equipment maintenance urgency, and its value is set according to the type of equipment and risk management strategy. is an adjustable sensitivity coefficient, taking a value between 0 and 1, used to control the degree of influence of the equipment maintenance urgency on the derating of the performance parameters of the equipment, and the coefficient can be calibrated through historical data analysis and expert experience. The correction process ensures that when the equipment maintenance urgency is high, the corresponding performance parameters will be reasonably adjusted, thus reflecting the true and limited power generation capacity. Finally, the system inputs these corrected real-time equipment state data into the short-term prediction model, and finally generates a short-term prediction result that is closer to the actual operating limit.

[0064] Optionally, the generating of the collaborative scheduling decision including power station output plan, energy storage charging and discharging strategy and operation and maintenance task allocation instruction comprises:

[0065] obtaining the equipment maintenance urgency and the energy storage charge state;

[0066] constructing a multi-objective function with power generation efficiency, equipment life, grid demand and economy as optimization objectives;

[0067] taking the corrected power generation plan, the equipment maintenance urgency, the energy storage charge state and the electricity market price information as input variables of the multi-objective function;

[0068] solving the multi-objective function to obtain an optimal balance solution, and analyzing the optimal balance solution into a power station output plan, an energy storage charging and discharging strategy and an operation and maintenance task allocation instruction to obtain a collaborative scheduling decision.

[0069] Specifically, the generation of the collaborative scheduling decision including the power station output plan, the energy storage charging and discharging strategy and the operation and maintenance task allocation instruction in the present application is realized by a multi-objective dynamic trade-off algorithm, which aims to find the optimal balance point between power generation income, equipment health, grid stability and economic benefits. The process first obtains all the input information required for decision-making, including the previously generated corrected power generation plan, the dynamically calculated equipment maintenance urgency, the real-time monitored energy storage charge state, and the price information obtained from the electricity market. Subsequently, the system constructs a multi-objective function with power generation efficiency, equipment life, grid demand and economy as optimization objectives. This function is not a single mathematical expression, but an optimization problem framework that aims to simultaneously optimize or satisfy a set of objective functions, which can be represented as seeking the optimal decision variable set with the optimization objective vector :

[0070] ,

[0071] In this framework, for a multi-objective optimization function, is a set of decision variables, mainly containing power plant output plan for each scheduling time step t in the future , energy storage charging and discharging strategy , positive value for discharging, negative value for charging, and operation and maintenance task allocation instruction , i.e. a Boolean value or a priority index, determines whether to perform maintenance in this period. is a vector containing four objective functions. represents the generation efficiency objective, aiming to maximize the ratio of actual power generation to the theoretical maximum power generation based on the revised generation plan. represents the equipment life objective, aiming to minimize the equipment wear and tear caused by high load operation, which converts higher equipment maintenance urgency E into higher penalty for higher output . represents the grid demand objective, aiming to minimize the deviation between actual power plant output and grid scheduling instruction. represents the economic objective, aiming to maximize the electricity sales revenue minus operation and maintenance costs and energy storage loss costs, which is directly related to electricity market price information.

[0072] In the solving process, the revised generation plan, equipment maintenance urgency, energy storage state of charge and electricity market price information are input into the multi-objective function as known parameters or constraints. For example, must be less than or equal to the upper limit given by the revised generation plan; the charging and discharging behavior of the energy storage is constrained by its current state of charge and maximum / minimum state of charge. The system solves this multi-objective function through a multi-objective dynamic weighting algorithm, such as an optimization algorithm with adaptive weights, to find an optimal balanced solution that best balances the four conflicting objectives. This solution is essentially a sequence of optimal values for the decision variables X throughout the scheduling period. Finally, the system analyzes this optimal balanced solution and converts the sequence into a specific power plant output plan, sequence into an energy storage charging and discharging strategy, sequence into operation and maintenance task allocation instructions, thus forming a unified, internally coordinated collaborative scheduling decision.

[0073] Optionally, the generation of personnel dispatching scheme and material allocation path includes:

[0074] Analyzing the operation and maintenance task allocation instructions to determine the required technical personnel skills and spare parts types;

[0075] Obtaining technical personnel location information and spare parts inventory location information;

[0076] The candidate technicians are screened in combination with the technician position information and the technician skills, and the candidate spare part warehouses are screened in combination with the spare part type and the spare part inventory position information;

[0077] The candidate technicians and the candidate spare part warehouses are combined, and path optimization calculation is performed, a combination is selected according to a total response time, a personnel dispatching scheme and a material allocation path are generated.

[0078] Specifically, by analyzing the operation and maintenance task allocation instruction, the specific requirements of each operation and maintenance task can be accurately identified, including the skill level or professional qualification of the technician required for task execution, and the type and quantity of spare parts required for maintenance or replacement. After the task requirements are clear, the system enters the resource matching stage. It first obtains the current position information of all available technicians and the skill information pre-entered in their archives in real time. At the same time, the system also queries the inventory position information of each spare part, including the real-time inventory data of the central warehouse and the distributed site spare part warehouse. Based on the skill requirements of the task, the system screens the candidate technicians with corresponding qualifications from all technicians. Similarly, according to the required spare part type, the system screens the candidate spare part warehouse that has the spare part and sufficient inventory.

[0079] Then, the system enters the path optimization calculation stage, aiming to find the resource combination with the shortest response time. The system combines each candidate technician and each candidate spare part warehouse in Cartesian product to form multiple potential dispatching schemes of “personnel-material-task point”. For each combination scheme, the system needs to calculate its total response time. The calculation of the total response time is a multi-path planning problem, which can be represented by the following formula:

[0080] ,

[0081] Among them, represents the total response time of the combination scheme. is the shortest travel time between locations calculated by a real-time path planning algorithm, which takes into account real-time traffic conditions, passable areas determined by geographic fence technology, etc. is the current position information of the candidate technician, is the position information of the candidate spare part warehouse, is the device position information of the operation and maintenance task to be executed. The first term of the formula calculates the time for the technician to go to the spare part warehouse, and the second term calculates the time for the technician to go to the task point after carrying the spare part. is the estimated time needed for spare parts preparation and collection, which is a constant based on historical data or preset standards. If the technician can go directly to the task point, for example, the task does not require spare parts or the spare parts are already on site, the path planning is simplified to the technician's location to the task point. The system will calculate this for all possible combinations and select the combination with the smallest value. Finally, according to the principle of the shortest total response time, the system determines the optimal combination scheme. This scheme is parsed into a specific personnel dispatching scheme, i.e., which technician is assigned to perform the task, and a detailed material allocation path, i.e., which technician should collect spare parts from which spare parts warehouse and follow the planned optimal route to the task site.

[0082] Optionally, the adjusting the prediction parameters and the scheduling parameters comprises:

[0083] Recording the execution results of the collaborative scheduling decisions and calculating the actual execution deviations, generating a historical decision dataset based on the execution results and the actual execution deviations;

[0084] Inputting the historical decision dataset into the incremental learning model for training to identify systematic deviation patterns;

[0085] Generating a deviation adjustment amount according to the systematic deviation patterns;

[0086] Applying the deviation adjustment amount to update the prediction parameters and the scheduling parameters.

[0087] Specifically, as shown in Figure 2 The process of adjusting the prediction parameters and the scheduling parameters in the present application constructs a closed-loop adaptive learning mechanism, enabling the system to continuously optimize itself from historical experience. The first step of this process is data accumulation and deviation quantification. The system will record the complete execution results of each collaborative scheduling decision in detail, such as the actual power curve of the power plant, the actual charging and discharging power of the energy storage, and the actual completion of the operation and maintenance tasks. Then, the system compares these execution results with the original collaborative scheduling decisions item by item, accurately calculates the actual execution deviations, such as the deviation of the power generation plan, the execution error of the energy storage strategy, etc. Based on the original decisions, execution results and calculated deviations, the system constructs and continuously expands a structured historical decision dataset.

[0088] Next, the system trains an incremental learning model using the historical decision dataset. The incremental learning model is characterized by its ability to iteratively update the original model as new data arrives, without the need to retrain on the entire historical dataset, which ensures the efficiency and real-time nature of the learning process. The goal of the training is to allow the model to autonomously learn and identify systematic bias patterns by analyzing the historical decision dataset. Systematic bias patterns refer to recurring, regular prediction or decision errors under specific conditions. For example, the model may find that, under a certain specific weather forecasting pattern, the generation prediction is consistently systematically high. Once the incremental learning model identifies a specific systematic bias pattern, it can generate a corresponding bias adjustment amount based on the pattern. This bias adjustment amount can be understood as the model's prediction of the bias that may occur under similar conditions in the future. The generation process can be represented by the following formula:

[0089] ,

[0090] wherein, represents the bias adjustment amount generated for a certain specific prediction parameter or scheduling parameter. is the trained incremental learning model. represents the context feature vector describing the future decision-making context, which includes the expected weather conditions, device status, and electricity price information, which are consistent with the context format recorded in the historical decision dataset. The model outputs a quantitative prediction of the systematic bias in this scenario by inputting the future context, i.e., the bias adjustment amount. The last step is to apply the bias adjustment amount to update the core parameters of the system. The system applies the calculated bias adjustment amount to the corresponding prediction parameters and scheduling parameters. For example, if the model predicts that there is a systematic negative bias in the generation prediction for the next period, the system will apply this negative bias adjustment amount to the relevant parameters of the prediction model to improve the accuracy of future predictions. Similarly, if it is found that a certain scheduling weight setting leads to continuous suboptimal results, the weight will be adjusted accordingly. Through this continuous "recording, analysis, adjustment, and application" cycle, the prediction and scheduling capabilities of the system are continuously iterated and refined.

[0091] Optionally, the method further comprises:

[0092] Based on the modified generation plan, determining whether a sudden drop in future generation capacity occurs;

[0093] If so, triggering an emergency response, calculating a predicted power gap based on the modified generation plan, and updating the energy storage charging and discharging strategy according to the power gap;

[0094] Synchronously adjusting the operation and maintenance task allocation instructions, and generating adaptive scheduling instructions in combination with the updated energy storage charging and discharging strategy.

[0095] Specifically, the emergency response method for handling the sudden drop in power generation in the present application is a set of active risk management mechanism based on accurate prediction. The triggering condition of the process is the real-time monitoring of the system on the future power generation. The system will continuously analyze the modified generation plan generated by the two-way feedback prediction mechanism and calculate the first-order difference between adjacent time steps, i.e. the change rate of power generation. When the system detects that the negative value of the change rate exceeds a preset sudden drop threshold, for example, the predicted power generation drops sharply in a short time, the emergency response process will be automatically triggered. Once the emergency response is triggered, the system will first accurately calculate the expected power gap based on the current modified generation plan. The calculation method of the power gap is to compare the modified generation plan sequence within a certain period of time after the sudden drop (for example, 1 hour in the future) with the generation plan or grid dispatching instruction in the stable state before the sudden drop, and the sum of the difference is the predicted power gap. The calculation can be represented by the following formula:

[0096] ,

[0097] Wherein, represents the total predicted power gap. is the target output at time t, which can be the stable output value before the sudden drop or the dispatching instruction value issued by the grid. is the modified generation plan value at time t after the emergency response is triggered. is the preset emergency response time window. After quantifying the power gap, the system will immediately update the original energy storage charging and discharging strategy according to the power gap. The system will re-plan the discharging plan of the energy storage system to discharge with the optimal power curve in the period when the power gap occurs, aiming to make up for the lack of power generation and smooth the fluctuations of the total power output of the power station to the greatest extent. This updating process will consider the constraints such as the current state of charge of the energy storage, the maximum discharge power and the remaining available power, to ensure the feasibility of the strategy. At the same time, the emergency response mechanism will also adjust the operation and maintenance task allocation instructions synchronously. The system will analyze the root cause of the sudden drop in power generation. If the decrease in the modified generation plan is due to the predictive failure or sharp deterioration of a specific device, this information is already included in the generation process of the modified generation plan, and the system will immediately raise the priority of the operation and maintenance of the related device and may generate new and urgent operation and maintenance task allocation instructions. This step ensures that operation and maintenance resources can be quickly directed to the root cause of the problem. Finally, the system integrates the updated energy storage charging and discharging strategy with the synchronously adjusted operation and maintenance task allocation instructions to generate a new set of adaptive dispatching instructions that adapt to the current emergency situation. This set of instructions will replace the corresponding part of the original collaborative dispatching decision and be issued to the execution unit of the power station, including the energy storage control system and the operation and maintenance management platform, so as to realize the rapid and coordinated response to the emergency.

[0098] Based on the same inventive concept, the present application also provides a new energy power station intelligent operation and maintenance scheduling and resource optimization system, as shown in the accompanying drawings, the system comprises: Figure 3

[0099] A data fusion module is configured to acquire real-time power generation data, equipment sensor data, weather forecast data, power grid scheduling instructions and power market electricity price information of the new energy power station, and generate a unified data matrix through spatio-temporal alignment processing;

[0100] A collaborative prediction module is configured to perform bidirectional feedback prediction based on the unified data matrix, the bidirectional feedback prediction including generating long-term prediction results and short-term prediction results, and correcting errors of the long-term prediction results by using the short-term prediction results to generate a corrected power generation plan;

[0101] An intelligent scheduling module is configured to acquire equipment maintenance urgency and energy storage charge state, and combine the corrected power generation plan and the power market electricity price information to generate a collaborative scheduling decision including power station output plan, energy storage charging and discharging strategy and operation and maintenance task allocation instruction through a multi-objective dynamic weighting algorithm;

[0102] A resource optimization module is configured to acquire technician location information and spare parts inventory location information based on the operation and maintenance task allocation instruction, and generate a personnel dispatching scheme and a material allocation path through real-time path planning;

[0103] An adaptive learning module is configured to record execution results of the collaborative scheduling decision and calculate actual execution deviations, generate a historical decision dataset based on the execution results and the actual execution deviations, and adjust prediction parameters and scheduling parameters through an incremental learning model using the historical decision dataset.

[0104] In order to verify the feasibility of the present application in implementation, the present application is applied to a certain new energy demonstration power station. The power station includes large-scale photovoltaic arrays, wind turbine generators and supporting energy storage systems, aiming to optimize its power generation plan, improve operation and maintenance efficiency and maximize market benefits. After the system of the present application is deployed, real-time data acquisition, scheduling decision generation and closed-loop optimization verification are carried out for a period of months.

[0105] ​In this embodiment, the data fusion module of the present application first processes the multi-source heterogeneous data of the power plant. The system obtains real-time power generation data of the photovoltaic array, device sensor data of the wind turbine generator unit such as W11, including temperature and vibration parameters, regional minute-level weather forecast data, dispatch instructions issued by the power grid, and real-time electricity price information of the provincial power market. Through spatio-temporal alignment processing, the system synchronizes these data of different frequencies and formats to a unified 15-minute time step and maps them to a unified spatial reference such as a specific device ID, generating a structured unified data matrix, which can be represented as , providing a high-quality data foundation for subsequent accurate prediction and scheduling.

[0106] At 10:15 on August 5, 2024, the system extracted the vibration parameter of the W11 wind turbine gearbox from the unified data matrix, which showed a sustained slight upward trend. Based on the preset health state threshold, the system adjusted its health state from "healthy" to "attention", and combined with the time rate of change of its vibration parameter, calculated its device maintenance urgency through the following formula: . The calculated device maintenance urgency value increased from 15 to 45, with a full score of 100, indicating a potential risk of failure.

[0107] Subsequently, the collaborative prediction module uses this urgency to correct the real-time state data of the W11 wind turbine when generating short-term prediction results. According to the formula , the system performs derating processing on its theoretical maximum available power parameter. This corrected state data is input into the short-term prediction model to generate a short-term prediction result that reflects its true power generation capacity. At the same time, the long-term prediction model gives a higher power generation prediction based on the weather forecast (stable wind speed). The two-way feedback prediction mechanism is activated, and the long-term prediction result is calibrated through backpropagation using the adjusted short-term prediction result through the formula , generating a more accurate revised generation plan. The plan shows that the total power output of the power plant will be reduced by about 5 MW in the next 3 hours compared to the original ideal state prediction.

[0108] In the intelligent scheduling link, the intelligent scheduling module obtains the revised generation plan, the high maintenance urgency of the W11 wind turbine, the 70% state of charge of the energy storage system, and the electricity price information of the peak market price segment at that time. The system constructs a multi-objective function , and make a dynamic trade-off. To balance the economy (high electricity price) and the equipment life (high urgency), the algorithm generates a collaborative scheduling decision: 1. The power station output plan is executed according to the revised generation plan in the next 1 hour; 2. The energy storage system discharges 5 MWh in 1 hour to make up for the gap caused by the potential output drop of W11 fan, to ensure that the grid scheduling instructions are met and the maximum electricity sales revenue is obtained; 3. Generate maintenance task allocation instructions for W11 fan, plan to repair during the low wind speed period in 2 hours.

[0109] Based on the maintenance task allocation instructions, the resource optimization module is started. Instruction analysis determines that a mechanic with a secondary qualification and a specific type of gear box lubricating oil are needed. The system obtains the location information of technicians Zhang and Li, who have a secondary qualification, located in the main control building and F area inspection, respectively, as well as the location information of central spare parts warehouse A and No. 3 distributed spare parts warehouse B, both of which have the required spare parts. Through path optimization calculation , the system calculates that the total response time of "Zhang from the main control building to the spare parts warehouse A to pick up the materials and then to W11 fan" is 45 minutes, while other combination schemes are more than 1 hour. Therefore, the system generates a personnel dispatching scheme, assigns Zhang to perform the task, and plans the optimal material allocation path.

[0110] In a sudden event, on September 2, 2024, due to a cloud layer moving quickly to block the photovoltaic area, the revised generation plan determines that the power generation will drop by 30 MW in the next 15 minutes. The system immediately triggers the emergency response, calculates the predicted power gap through the formula , and updates the energy storage charging and discharging strategy according to the gap, instructing the energy storage system to discharge at maximum power for emergency support, generating adaptive scheduling instructions, successfully avoiding sharp fluctuations in power output to the grid.

[0111] After a quarter of a year of operation, the adaptive learning module records the execution results of all decisions and the actual execution deviation, generating a historical decision data set. Through incremental learning model training, the system finds that in cloudy and sunny weather patterns, its prediction of the ramp rate of photovoltaic output is always systematically slow. The model generates a deviation adjustment amount through the formula based on this deviation pattern, and updates the relevant weight parameters in the prediction model with this adjustment amount. After adjustment, the prediction accuracy of the system in this specific weather pattern improves from 88% to 95%.

[0112] Table 1 Collaborative scheduling decision data table for new energy power station

[0113]

[0114] Table 2 Maintenance resource optimization scheduling data table

[0115]

[0116] Table 3 System adaptive learning adjustment data table

[0117]

[0118] The above Tables 1 to 3 record the actual application data of the present application in a new energy demonstration power station, and details the performance of the system in coordinated scheduling, resource optimization and adaptive learning. Through these data, the advancement and practical value of the present application can be clearly seen.

[0119] In Table 1, the system successfully coped with the potential failure risk of W11 fan on August 5, 2024 by revising the power generation plan and energy storage scheduling, maintaining stable power output while ensuring equipment safety. On September 2, in the face of a sharp drop in power generation, the system quickly compensated for the power gap using energy storage through an emergency response mechanism, with only a slight fluctuation in actual output, demonstrating its excellent risk response capability.

[0120] In Table 2, the system quantitatively selects the best resource allocation process for maintenance tasks. By accurately calculating the response time of different "personnel-material" combinations, the system selects the optimal solution, shortening the response time from more than 1 hour to 45 minutes, significantly improving maintenance efficiency.

[0121] In Table 3, the system's strong self-learning and self-evolution capabilities are demonstrated. Through analysis and learning of historical deviations, the system can automatically adjust its internal model parameters, continuously improving prediction accuracy and the economic efficiency of scheduling strategies. For example, the prediction accuracy for a specific weather pattern improved by 7 percentage points, which directly translates into more optimal power generation plans and higher market returns. These data results fully demonstrate the significant technical effects of the present application in improving the intelligent level, operational efficiency and economic benefits of new energy power stations.

[0122] It should be noted that the above formulas can be translated into unitless standard values or same-dimension superimposable parameters by means of dimensional consistency principles and mathematical standardization methods (such as normalization processing, dimensionless parameter conversion or unit system unification), thereby eliminating the interference of different dimensions on the operation logic, making the formula retain the original data distribution characteristics while having mathematical operation rationality and objective law adaptability. These are conventional technical means and will not be described again. The electrical connections between the above-mentioned units do not necessarily represent direct connections, and indirect connection methods can also be used as long as the purpose of the present application is achieved. The above-described embodiments are only exemplary embodiments of the present application, and cannot limit the scope of the present application.

[0123] intended to encompass any and all embodiments of the application with equivalents as would be ascertained by those skilled in the art to which the application pertains. Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.

Claims

1. A new energy power station intelligent operation and maintenance scheduling and resource optimization method, characterized in that, The method comprises: Obtaining real-time power generation data, equipment sensor data, weather forecast data, grid dispatching instructions and power market price information of a new energy power station, and generating a unified data matrix through space-time alignment processing; wherein it also includes: extracting temperature parameters and vibration parameters from the equipment sensor data; using a preset health state threshold, analyzing the temperature parameters and vibration parameters to generate a device health state; based on the device health state and the time variation rate of the temperature parameters and vibration parameters, calculating a fault risk coefficient as the device maintenance urgency; Based on the unified data matrix, a two-way feedback prediction is performed, which includes generating long-term prediction results and short-term prediction results, and using the short-term prediction results to correct the long-term prediction results to generate a corrected power generation plan; which includes: extracting weather trends and historical power generation rules from the unified data matrix and performing long-term prediction to generate long-term prediction results; extracting real-time equipment state data from the unified data matrix; using the device maintenance urgency to correct the real-time equipment state data; based on the corrected real-time equipment state data, performing short-term prediction to generate short-term prediction results; calculating the prediction error between the long-term prediction results and the short-term prediction results, and using the prediction error to correct the long-term prediction results through back propagation, to generate a corrected power generation plan; Obtaining the device maintenance urgency and the energy storage charge state, and combining the corrected power generation plan and the power market price information, generating a collaborative scheduling decision including power station output plan, energy storage charging and discharging strategy and operation and maintenance task allocation instruction through multi-objective dynamic weighting algorithm; Based on the operation and maintenance task allocation instruction, obtain the technician location information and the spare parts inventory location information, and generate a personnel dispatching scheme and a material allocation path through real-time path planning; Record the execution result of the collaborative scheduling decision and calculate the actual execution deviation, generate a historical decision data set based on the execution result and the actual execution deviation, and adjust the prediction parameters and the scheduling parameters through the incremental learning model using the historical decision data set. 2.The new energy power station intelligent operation and maintenance scheduling and resource optimization method of claim 1, characterized in that, The generation of the unified data matrix comprises: Obtaining real-time power generation data, equipment sensor data, weather forecast data, grid dispatching instructions and power market price information of a new energy power station, to obtain a basic data set; Extracting timestamp and geographic coordinate information from the basic data set; Based on the timestamp, synchronizing the data in the basic data set to a unified time step, and based on the geographic coordinate information, mapping the data in the basic data set to a unified spatial reference to obtain a processed basic data set; Structurally integrating the processed basic data set to generate a unified data matrix.

3. The intelligent operation and maintenance scheduling and resource optimization method for a new energy power station according to claim 1, characterized in that, The generation of the collaborative scheduling decision including the power station output plan, the energy storage charging and discharging strategy and the operation and maintenance task allocation instruction comprises: Obtaining the device maintenance urgency and the energy storage charge state; Constructing a multi-objective function with power generation efficiency, equipment life, grid demand and economy as optimization objectives; The modified power generation plan, the equipment maintenance urgency, the energy storage charge state and the power market price information are taken as input variables of the multi-objective function; The multi-objective function is solved to obtain an optimal balance solution, and the optimal balance solution is analyzed into a power station output plan, an energy storage charging and discharging strategy and an operation and maintenance task allocation instruction, to obtain a coordinated scheduling decision. 4.The new energy power station intelligent operation and maintenance scheduling and resource optimization method of claim 3, characterized in that, The generation of the personnel dispatching scheme and the material allocation path includes: The operation and maintenance task allocation instruction is analyzed to determine the required technical personnel skills and spare part types; Technical personnel position information and spare part inventory position information are obtained; The candidate technical personnel are screened in combination with the technical personnel position information and the technical personnel skills, and the candidate spare part libraries are screened in combination with the spare part inventory position information and the spare part types; The candidate technical personnel and the candidate spare part libraries are combined, and path optimization calculation is performed, and a combination is selected according to the total response time, to generate the personnel dispatching scheme and the material allocation path.

5. The intelligent operation and maintenance scheduling and resource optimization method for a new energy power station according to claim 3, characterized in that, The adjustment of the prediction parameters and the scheduling parameters includes: The execution result of the coordinated scheduling decision is recorded, and the actual execution deviation is calculated, a historical decision dataset is generated based on the execution result and the actual execution deviation; The historical decision dataset is input into an incremental learning model for training, to identify a systematic deviation mode; A deviation adjustment amount is generated according to the systematic deviation mode; The deviation adjustment amount is applied to update the prediction parameters and the scheduling parameters.

6. The intelligent operation and maintenance scheduling and resource optimization method for a new energy power station according to claim 3, characterized in that, The method further includes: Based on the modified power generation plan, it is determined whether a sudden drop in future power generation occurs; If so, an emergency response is triggered, a predicted power gap is calculated based on the modified power generation plan, and the energy storage charging and discharging strategy is updated according to the power gap; The operation and maintenance task allocation instruction is adjusted synchronously, and an adaptive scheduling instruction is generated in combination with the updated energy storage charging and discharging strategy.

7. A new energy power station intelligent operation and maintenance scheduling and resource optimization system applied to the new energy power station intelligent operation and maintenance scheduling and resource optimization method of any one of claims 1-6, characterized in that, The system includes: A data fusion module is configured to obtain real-time power generation data of a new energy power station, device sensor data, weather forecast data, power grid scheduling instructions and power market price information, and generate a unified data matrix through spatiotemporal alignment processing; wherein the temperature parameters and the vibration parameters are extracted from the device sensor data; the temperature parameters and the vibration parameters are analyzed by using a preset health state threshold to generate a device health state; and a fault risk coefficient is calculated as an equipment maintenance urgency based on the device health state and a time variation rate of the temperature parameters and the vibration parameters. The cooperative prediction module is configured to perform bidirectional feedback prediction based on the unified data matrix, the bidirectional feedback prediction including generating long-term prediction results and short-term prediction results, and correcting errors of the long-term prediction results by using the short-term prediction results to generate a corrected power generation plan; wherein the cooperative prediction module is configured to extract weather trends and historical power generation rules from the unified data matrix, and perform long-term prediction to generate the long-term prediction results; extract real-time equipment state data from the unified data matrix; correct the real-time equipment state data by using the equipment maintenance urgency; perform short-term prediction based on the corrected real-time equipment state data to generate the short-term prediction results; calculate prediction errors between the long-term prediction results and the short-term prediction results, and correct the long-term prediction results by using the prediction errors to generate the corrected power generation plan; The intelligent scheduling module is configured to obtain the equipment maintenance urgency and the energy storage charge state, and combine the corrected power generation plan and the power market price information to generate a cooperative scheduling decision including a power plant output plan, an energy storage charging and discharging strategy, and an operation and maintenance task allocation instruction by using a multi-objective dynamic weighting algorithm; The resource optimization module is configured to obtain technician location information and spare parts inventory location information based on the operation and maintenance task allocation instruction, and generate a personnel dispatching scheme and a material allocation path by using real-time path planning; The adaptive learning module is configured to record an execution result of the cooperative scheduling decision and calculate an actual execution deviation, generate a historical decision dataset based on the execution result and the actual execution deviation, and adjust prediction parameters and scheduling parameters by using the historical decision dataset through an incremental learning model.

Images