Hydraulic engineering construction operation parameter optimization and data management method and system

Abstract

The invention discloses a water conservancy project construction operation parameter optimization and data management method and system, and the method comprises the steps: firstly, carrying out the vector feature extraction, feature extraction and other operations of a water scheduling parameter, determining a plurality of task types, and predicting the abnormal event confidence of the task types, and extracting an influence factor weighting coefficient, and performing weighting processing to obtain a project overall abnormal event assessment confidence coefficient. And if abnormity exists, the subsystems and the key equipment are monitored in real time, and a preset treatment strategy is pushed for optimization adjustment, so that the operation management efficiency of the water conservancy project is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a method and system for optimizing operation parameters and data management in water conservancy project construction. Background Art

[0002] Accurate water scheduling and reliable data management are crucial in the operation and management of water conservancy projects. When dealing with complex working conditions and potential abnormal events, traditional methods have problems such as insufficient analysis of water scheduling parameters and difficulty in comprehensively evaluating abnormal risks under various task types. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for optimizing operation parameters and data management of water conservancy project construction.

[0004] In a first aspect, an embodiment of the present invention provides a method for optimizing operation parameters and managing data of a water conservancy project, including: Performing a vector feature extraction operation on the water scheduling parameters of the target water conservancy project to obtain a water scheduling vector representation corresponding to the water scheduling parameters; Performing a feature extraction operation on the water scheduling vector representation to obtain a target water scheduling feature corresponding to the water scheduling vector representation; Determine a plurality of task types corresponding to the water scheduling parameters, perform abnormal event prediction processing on the target water scheduling characteristics according to each task type, and obtain abnormal event prediction confidence corresponding to each task type; Extracting influencing factors from the water quantity dispatching vector representation to obtain influencing factor weighting coefficients corresponding to each task type; Based on the weighted coefficients of the impact factors corresponding to the respective task types, the prediction confidences of the abnormal events corresponding to the respective task types are weighted to obtain the weighted abnormal event prediction coefficients corresponding to the respective task types; the conversion value of the impact factor corresponding to a task type is used to weight the prediction confidences of the abnormal events corresponding to the respective task types; Determining the confidence level of abnormal event assessment of the target water conservancy project according to the weighted abnormal event prediction coefficients corresponding to the various task types; When the abnormal event assessment confidence level indicates that an abnormal event has occurred in the target water conservancy project, real-time monitoring data collection is performed on each subsystem and key equipment of the target water conservancy project, and preset governance strategies are pushed to each subsystem for optimization and adjustment based on the abnormal event assessment confidence level.

[0005] In a second aspect, an embodiment of the present invention provides a server system, including a server, wherein the server is used to execute the method described in the first aspect.

[0006] Compared with the prior art, the beneficial effects provided by the present invention include: using a water conservancy project construction operation parameter optimization and data management method and system disclosed by the present invention, by performing vector feature extraction and feature extraction on water volume scheduling parameters, multiple task types are determined and their abnormal event confidences are predicted respectively, and the weighted coefficients of the influencing factors are extracted and weighted to obtain the overall abnormal event assessment confidence of the project. If there is an abnormality, each subsystem and key equipment is monitored in real time and the preset management strategy is pushed for optimization and adjustment to improve the operation and management efficiency of the water conservancy project. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative work.

[0008] Figure 1 A schematic diagram of the steps of the method for optimizing the operation parameters and data management of water conservancy project construction provided by an embodiment of the present invention; Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0009] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0010] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0011] In order to solve the technical problems in the aforementioned background technology, Figure 1 This is a flow chart of the method for optimizing the operation parameters and data management of water conservancy project construction provided in the embodiment of the present disclosure. The method for optimizing the operation parameters and data management of water conservancy project construction is introduced in detail below.

[0012] Step S201, performing a vector feature extraction operation on the water scheduling parameters of the target water conservancy project to obtain a water scheduling vector representation corresponding to the water scheduling parameters; Step S202, performing a feature extraction operation on the water scheduling vector representation to obtain a target water scheduling feature corresponding to the water scheduling vector representation; Step S203, determining a plurality of task types corresponding to the water scheduling parameters, and performing abnormal event prediction processing on the target water scheduling characteristics according to each task type, to obtain abnormal event prediction confidence corresponding to each task type; Step S204, extracting influencing factors from the water scheduling vector representation to obtain influencing factor weighting coefficients corresponding to each task type; Step S205, weighting the abnormal event prediction confidences corresponding to the respective task types based on the weighted coefficients of the impact factors corresponding to the respective task types, to obtain weighted abnormal event prediction coefficients corresponding to the respective task types; the impact factor conversion value corresponding to a task type is used to weight the abnormal event prediction confidences corresponding to the corresponding task type; Step S206, determining the confidence level of abnormal event assessment of the target water conservancy project according to the weighted abnormal event prediction coefficients corresponding to the various task types; Step S207, when the abnormal event assessment confidence level indicates that an abnormal event exists in the target water conservancy project, real-time monitoring data collection is performed on each subsystem and key equipment of the target water conservancy project, and based on the abnormal event assessment confidence level, the preset governance strategies pushed to each subsystem are optimized and adjusted.

[0013] In an embodiment of the present invention, illustratively, the server first proceeds to collect data related to water scheduling parameters from various data sources of the target water conservancy project. The water conservancy hub includes multiple reservoirs, multiple river connection sections and numerous water conservancy facilities, and the data sources are rich and diverse. Real-time water level data can be obtained from the water level monitoring sensors of each reservoir, which are updated every hour and record the water level height at different times in detail, such as the water level is 100 meters at 8 am and 100.2 meters at 9 am in a specific period of time. The river flow monitoring station can provide the flow data of the river, such as the flow of a main river connection section at a certain time is 500 cubic meters per second. At the same time, data such as gate opening and pump station operating power can also be obtained from the operation control system of the water conservancy facility, such as the gate opening is 30% open in a certain period of time, and the pump station operating power is 200 kilowatts. Since these data types, formats, and dimensions are different, the server needs to organize and preprocess them, and unify the format for subsequent operations. For example, the water level data is unified in meters, and the flow data is standardized in the format of cubic meters per second. Afterwards, appropriate feature encoding methods are used, such as unique hot encoding to process gate openings in different states, and numerical encoding to process water level, flow, pump station operating power and other data. The pre-processed water scheduling parameters are converted into vector form, and water scheduling vector representations such as [100 (water level numerical encoding), 500 (flow numerical encoding), 0.3 (gate opening converted from unique hot encoding), 200 (pump station operating power numerical encoding)] are obtained, completing the conversion from original parameters to vector representation, laying the foundation for subsequent analysis. For water level data, taking the past 24 hours as an example, if the water level shows a trend of first slowly rising and then slowly falling, by calculating the water level difference at adjacent time points and analyzing its change law, it can be obtained that the water level rise rate is an average of 0.1 meters per hour in a few hours, and the drop rate is an average of 0.08 meters per hour in the following hours. For flow data, its seasonal and periodic characteristics are analyzed. The area where the water conservancy hub is located is rainy in summer. After analyzing the flow data for many years, it was found that the flow would rise periodically from June to August each year, which was about 30% higher than other months on average. This is the target water scheduling feature corresponding to the flow data. Relevant features can also be extracted from data such as gate opening and pump station operating power, such as the frequent adjustment frequency of gate opening and the stable operation duration of pump station operating power. According to the actual operation needs and functions of water conservancy projects, multiple task types such as flood control, irrigation and water supply, power generation, and ecological water replenishment were determined. For flood control tasks, the extracted target water scheduling features were used as input, and the decision tree algorithm was used to construct an abnormal event prediction model. The input features include water level change trend characteristics (water level rise rate is too fast may indicate an increased flood risk), seasonal and periodic characteristics of river flow (abnormally high flow in non-rainy season may be a flood hazard), etc. After training the model with historical flood control data, the probability of abnormal events under flood control tasks can be predicted based on the current input features.Similarly, the irrigation water supply task uses the support vector machine algorithm to build the corresponding model. The input features include water level data (too low water level may lead to insufficient water supply), the frequent adjustment frequency characteristics of the gate opening, etc. After inputting the current features, the prediction confidence of the abnormal event of insufficient water supply can be predicted. Other task types such as power generation tasks and ecological water replenishment tasks also build corresponding abnormal event prediction models respectively, and calculate the corresponding abnormal event prediction confidence according to the input target water scheduling characteristics, so as to comprehensively evaluate the abnormal event risks that may occur under each task type. The server further analyzes the water scheduling vector representation and extracts the corresponding weighted coefficients of the influencing factors of each task type. In flood control tasks, the water level change trend characteristics are extremely critical to flood control, and the weighted coefficient of the influencing factor may be 0.5; the seasonal characteristics of river flow have a slightly lower impact, and the coefficient may be 0.3; the frequent adjustment frequency characteristics of the gate opening have a relatively small impact, with a coefficient of 0.1; the operating power characteristics of the pump station have little impact, with a coefficient of 0.1. In the irrigation water supply task, the weighted coefficient of the water level data influencing factor may be high, such as 0.4; the frequent adjustment frequency of the gate opening is more important, with a coefficient of 0.3; the seasonal characteristics of the flow data have a relatively small impact, with a coefficient of 0.2; the pump station operation power characteristics have little impact, with a coefficient of 0.1. For other task types, the weighted coefficient of the influencing factor is also determined according to the importance of each feature to the corresponding task. Then, according to the weighted formula, the weighted coefficient of the influencing factor corresponding to each feature and the confidence of the abnormal event prediction are weighted to obtain the corresponding weighted abnormal event prediction coefficient of each task type. Finally, the confidence of the abnormal event assessment of the target water conservancy project is determined, and methods such as weighted average or simple average can be used. If the evaluation confidence is 0.6 through calculation, it indicates that there is a high possibility of abnormal events. For the reservoir subsystem, the frequency of data collection of reservoir water level, water temperature, water quality, etc. is increased from once an hour to once every half an hour. If the water level continues to rise and the confidence of the abnormal event assessment is high, push the preset governance strategy, such as increasing the gate opening to reduce the reservoir water level, and pay close attention to the water level changes. For the irrigation water supply subsystem, strengthen the data collection of irrigation channel flow, water quality and irrigation equipment operation status. If the risk of insufficient water supply is high, push strategies such as giving priority to water supply in important crop planting areas, reasonably adjusting water supply time and flow distribution, and checking equipment failures. For the power generation subsystem, increase the intensity of data collection on power, temperature, vibration and other data of the generator set. If the power generation efficiency may be affected, the push strategy is to conduct a comprehensive inspection of the generator set, adjust the power generation load, and ensure the best operating state. For the ecological water replenishment subsystem, strengthen the data collection of water replenishment river flow, water quality and ecological environment indicators. If there may be problems with ecological water replenishment, the push strategy includes adjusting the water replenishment time and flow, and strengthening the surrounding ecological environment monitoring. Through such comprehensive real-time monitoring data collection and strategy push optimization and adjustment, the server can effectively ensure the safe, stable and efficient operation of the target water conservancy project, and calmly deal with various abnormal events that may occur.In summary, through the systematic processing and analysis of water conservancy project data and the monitoring and strategy adjustment based on this, the operation status of water conservancy projects can be accurately controlled and effectively guaranteed.

[0014] In the embodiment of the present invention, the vector feature extraction operation is performed on the water scheduling parameters of the target water conservancy project to obtain the water scheduling vector representation corresponding to the water scheduling parameters, which can be implemented through the following examples.

[0015] Performing feature mapping processing on the water scheduling parameters to obtain mapping feature parameters of the water scheduling parameters; A water quantity dispatching vector representation is extracted from the mapping characteristic parameters to obtain a water quantity dispatching vector representation of the mapping characteristic parameters.

[0016] In an embodiment of the present invention, exemplarily, the server first receives raw data from various monitoring points and equipment, such as water level values ​​of reservoirs at different times, flow data at various locations of rivers, gate opening information, and pump station operating power data. These data have different formats and dimensions, and are difficult to be used directly for in-depth analysis. Therefore, the server carries out feature mapping processing. For water level data, a mapping interval (such as [0,10]) is set, and it is converted into comparable mapping feature parameters through a linear mapping function, such as a 120-meter water level can be mapped to 6. Due to the large fluctuation range of river flow data, a logarithmic mapping function is constructed based on the historical flow maximum value, such as a flow of 800 cubic meters per second can be mapped to 4.5. The gate opening information is normalized, and 40% of the opening is mapped to 0.4. The pump station operating power data is linearly mapped according to its historical power range, and 300 kilowatts of power can be mapped to 0.5. Through these processes, the original and diverse water scheduling parameters are converted into unified and comparable mapping feature parameters. After completing the feature mapping process, the server performs water scheduling vector representation extraction. Taking the water conservancy project system as an example, the mapping characteristic parameters of each parameter have been obtained, such as water level 6, flow 4.5, gate opening 0.4, and pump station operating power 0.5. First, determine the vector representation dimension. If the above four factors are mainly considered, the dimension is 4. Then put each mapping characteristic parameter into the vector in order to form a water scheduling vector representation of [6, 4.5, 0.4, 0.5]. In practice, more factors may be considered, such as precipitation in meteorological data. Perform feature mapping processing on the precipitation data. Assuming that the parameter after mapping is 3 and the extended vector dimension is 5, the water scheduling vector representation becomes [6, 4.5, 0.4, 0.5, 3]. If other factors such as water quality are considered, the mapping characteristic parameters are put into the vector after similar processing to further improve the vector representation. Through the above operations, the server can accurately extract the vector representation reflecting the water scheduling of the target water conservancy project, which facilitates subsequent analysis and processing and makes it more efficient.

[0017] In the embodiment of the present invention, the feature mapping process is performed on the water scheduling parameter to obtain the mapping feature parameter of the water scheduling parameter, which can be implemented through the following example.

[0018] When the water volume dispatching parameters of the target water conservancy project are obtained, a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project is obtained; the target abnormal event assessment model includes a first abnormal event assessment sub-model; Determine demand information corresponding to the water scheduling parameter based on the first abnormal event assessment sub-model, and extract a target water demand parameter of the water scheduling parameter according to the demand information; Performing mapping processing on the target water demand parameter based on the first abnormal event assessment sub-model to obtain mapping processing data corresponding to the target water demand parameter; Based on the first abnormal event assessment sub-model, a standardization operation is performed on the mapping processing data corresponding to the target water demand parameter to obtain a mapping characteristic parameter of the water quantity scheduling parameter.

[0019] In an embodiment of the present invention, the target water conservancy project is a large-scale comprehensive hub, covering multiple reservoirs, complex irrigation channels and other facilities, serving many water needs. As the core of data processing, the server continuously collects various water scheduling parameters, such as real-time water level data of each reservoir updated every 15 minutes, irrigation channel flow data (fluctuating with seasons and demand), sluice opening information and pump station operation power data. At the same time, the server has a target abnormal event evaluation model that has been constructed for a long time, in which the first abnormal event evaluation submodel focuses on abnormal evaluation related to water demand. After the server obtains the relevant parameters and models, it is processed according to the first abnormal event evaluation submodel. Taking agricultural irrigation as an example, the model first analyzes the water scheduling parameters to determine the demand information, such as the approximate irrigation water demand of the main crops in the current growth stage. Then accurately extract the target water demand parameters, such as selecting corresponding data from the irrigation channel flow data, judging the water level height value that the reservoir needs to maintain according to the irrigation demand, and extracting the sluice opening value that affects the irrigation water allocation and regulation. After extracting the parameters, mapping processing is performed. For irrigation channel flow parameters, a mapping function is constructed based on the historical flow distribution law; reservoir water level parameters are mapped according to the design water level range and historical fluctuations; and sluice gate opening parameters are normalized. After the mapping process is completed to obtain the respective data, the standardization operation is performed. Taking the various parameters in the agricultural irrigation scenario as an example, according to the standardization rule of unifying the mapped data to a normal distribution interval with a mean of 0 and a standard deviation of 1, the mapping characteristic parameters of the final water scheduling parameters such as 0.67 (irrigation channel flow), 0.5 (reservoir water level), and 0.5 (sluice gate opening) are calculated through standardized formulas based on the overall mean and standard deviation of each parameter. These parameters are in a unified standard form to ensure the comparability and accuracy of subsequent analysis and processing.

[0020] In an embodiment of the present invention, the first abnormal event assessment sub-model includes a data access layer; the target water demand parameter includes a plurality of target water demand parameters, and the plurality of target water demand parameters include a target key water demand parameter; The mapping process of the target water demand parameter based on the first abnormal event assessment sub-model to obtain the corresponding mapping processing data of the target water demand parameter can be implemented through the following examples.

[0021] Determine a transformation domain corresponding to the target key water demand parameter according to the demand information corresponding to the target key water demand parameter; the transformation domain includes a plurality of conversion values, each conversion value having a corresponding target water demand value range; The target water demand numerical range corresponding to the target key water demand parameter is determined based on the data access layer, and the conversion value corresponding to the target water demand numerical range corresponding to the target key water demand parameter is used as the mapping processing data corresponding to the target key water demand parameter.

[0022] In an embodiment of the present invention, exemplarily, in daily operation, the server collects a large amount of water scheduling parameter data from various monitoring devices and systems. Among them, the target water demand parameters include multiple aspects, such as agricultural irrigation water demand parameters, urban residents' living water demand parameters, industrial water demand parameters, etc. Here we focus on the target key water demand parameters, assuming that they are irrigation water demand parameters for a large crop planting area. The server clarifies the demand information corresponding to the target key water demand parameters by connecting with the agricultural department's information and analyzing the historical water use data of the planting area. For example, this planting area is planted with rice, which is currently in the critical growth period. According to the growth characteristics and planting area of ​​rice, it is known that stable and sufficient irrigation water is needed every day at this stage, and the water quality of the irrigation water must meet certain standards. At the same time, the irrigation amount must be flexibly adjusted according to weather conditions (such as whether there is rainfall in the near future). Based on the above demand information, the server uses the relevant rules and algorithms in the first abnormal event evaluation submodel to determine the corresponding transformation domain of the target key water demand parameters. For the irrigation water demand parameters of this rice planting area, the determination of the transformation domain takes into account many factors. For example, from the perspective of water quantity, based on historical irrigation data and the water demand patterns of rice at different growth stages, multiple conversion values ​​and their corresponding target water demand ranges are determined. Assuming that the target water demand range corresponding to one of the conversion values ​​is between 5,000 cubic meters and 8,000 cubic meters of irrigation water per day, this means that when the irrigation water volume is in this range, it corresponds to a water use state represented by the conversion value; the other conversion value corresponds to a range of values ​​between 8,001 cubic meters and 10,000 cubic meters of irrigation water per day, representing another different water use situation, such as the need to increase irrigation volume during a period of relatively dry weather. From the perspective of water quality, there are also corresponding conversion values ​​and range settings. For example, when the pH of the water quality is between 6.5 and 7.5, and the oxygen content is between 5 mg / L and 8 mg / L, it corresponds to a specific conversion value, representing a state that meets the normal growth of rice and the quality of irrigation water meets the standard. The data access layer of the server continuously receives real-time data on irrigation water in this planting area from various monitoring points. For example, the flow data of the current channel can be obtained from the flow monitor of the irrigation channel, from which the actual amount of water irrigated to this planting area can be calculated; the pH value, oxygen content and other indicator data of the water quality can be obtained from the water quality monitoring equipment. Assume that the actual daily irrigation water volume of this planting area is 6,500 cubic meters, the pH value of the water quality is 7.0, and the oxygen content is 6 mg / L.According to the previously determined transformation domain, the server found that the 6,500 cubic meters of irrigation water is within the target water demand value range of "daily irrigation water volume between 5,000 cubic meters and 8,000 cubic meters", and the corresponding conversion value is A; the water quality also meets the numerical range corresponding to a certain conversion value set previously, and its corresponding conversion value is B. Then, the server will use the conversion value corresponding to the target water demand numerical range corresponding to the target key water demand parameter (here A and B) as the corresponding mapping processing data of the target key water demand parameter. These mapping processing data will be further used for subsequent analysis and processing to evaluate whether there are abnormal events and make corresponding scheduling decisions.

[0023] In an embodiment of the present invention, the target water scheduling characteristics are determined by performing a feature extraction operation on the water scheduling vector representation by a feature encoder in a first abnormal event assessment submodel; the first abnormal event assessment submodel is a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project; the embodiment of the present invention also provides the following implementation methods.

[0024] Acquire a first sample training array; the first training instance included in the first sample training array is a training instance that is not configured with an abnormal event target value; Acquire a plurality of training targets for the first original abnormal event evaluation sub-model; Processing the first training instance based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and determining multiple error parameters corresponding to the multiple training targets according to the sample water scheduling characteristics; one training target corresponds to one error parameter; A first target error parameter is determined according to the multiple error parameters, model parameters of the first original abnormal event assessment submodel are optimized according to the first target error parameter, and the optimized first original abnormal event assessment submodel is used as the first abnormal event assessment submodel.

[0025] In an embodiment of the present invention, exemplarily, the server uses the feature encoder in the first abnormal event assessment submodel to perform feature extraction operations on the acquired water scheduling vector representation. For example, the feature encoder will deeply analyze the vector value of information such as reservoir water level, river flow, gate opening, etc. in the vector representation, and extract target water scheduling features such as water level change trend, seasonal fluctuation characteristics of flow, and gate opening adjustment frequency. Then, the server carefully selects and sorts out the first sample training array from the historical operation data of the water conservancy project, in which the first training instance is not configured with the abnormal event target value, but only the water scheduling related data records of different time periods, covering the actual situation of water level, flow, gate opening, etc. at that time. At the same time, for the first original abnormal event assessment submodel, multiple training objectives such as accurate prediction of flooding in flood control scenarios, insufficient water supply in irrigation scenarios, and a significant decrease in power generation efficiency in power generation scenarios are determined. Subsequently, the first training instance in the first sample training array is sequentially input into the first original abnormal event assessment submodel for processing. Taking a certain instance as an example, the actual situation corresponding to its water scheduling vector representation is that the reservoir water level rises slowly, the river flow fluctuates normally, and the gate opening is adjusted less recently. After model processing, the corresponding sample water scheduling characteristics are obtained for each training target, such as the water level rise rate and other characteristics extracted under the flood control training target. Then, by comparing these sample water scheduling characteristics with the ideal situation of the corresponding training target, multiple error parameters are determined. Each training target has a corresponding error parameter. For example, under the flood control training target, when the actual water level rise rate is higher than the predicted normal rise rate, a deviation error parameter is generated. Finally, based on the multiple error parameters determined, the first target error parameter is calculated through a specific algorithm to comprehensively reflect the overall error of the model. The optimization algorithm is then used to adjust and optimize the parameters of the first original abnormal event assessment submodel. After multiple rounds of optimization to reduce the error, the optimized model is used as the first abnormal event assessment submodel, so that it can more accurately assess the abnormal events of the target water conservancy project.

[0026] In an embodiment of the present invention, the plurality of training targets include randomly deleting training targets; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics, including: Perform feature mapping processing on the first training instance based on the first original abnormal event assessment sub-model to obtain a first mapping feature parameter instance of the first training instance; the first mapping feature parameter instance includes mapping processing data instances on multiple index tags; Randomly selecting a first index marker to be randomly deleted from a plurality of index markers corresponding to the first mapping feature parameter instance; Performing random deletion processing on the mapping processing data instance on the first index mark in the first mapping feature parameter instance according to the random deletion information to obtain a first feature parameter instance; Based on the first original abnormal event assessment sub-model, the first feature parameter instance is subjected to water scheduling vector representation extraction to obtain a first water scheduling vector representation instance of the first feature parameter instance, and a feature extraction operation is performed on the first water scheduling vector representation instance to obtain a first sample water scheduling feature corresponding to the first water scheduling vector representation instance; the first sample water scheduling feature belongs to the sample water scheduling feature; Based on the execution unit corresponding to the random deletion training target, performing random deletion inference processing on the first sample water scheduling feature to obtain inference mapping processing data corresponding to the random deletion information in the first feature parameter instance; An error parameter corresponding to the randomly deleted training target is determined based on the mapping processing data instance on the first index mark in the first mapping feature parameter instance and the inferred mapping processing data.

[0027] In an embodiment of the present invention, illustratively, the server selects a first training instance from the historical operation data of the water conservancy project. This first training instance contains rich information related to water scheduling, such as the water level values ​​of each reservoir at a specific time, the flow data of each irrigation channel, the opening of different sluices, and the operating power of the pump station. The server inputs this first training instance into the first original abnormal event assessment sub-model, and the model first performs feature mapping processing on it. Taking the reservoir water level data as an example, assuming that the water level of a reservoir at this moment is 120 meters, the model maps the 120-meter water level to a value in a specific interval (such as [0,1]) through a specific mapping function (such as a linear mapping function) based on the range of historical water level data (such as the lowest water level of 80 meters and the highest water level of 150 meters), assuming that it is 0.6 after mapping. Similarly, for the flow data of the irrigation channel, the opening of the sluice, and the operating power of the pump station, the model also uses the corresponding mapping method to process them respectively, and finally obtains the first mapping feature parameter instance of the first training instance. This first mapping characteristic parameter instance contains mapping processing data instances on multiple index tags, such as the mapping processing data instance with the index tag "water level" is 0.6, the mapping processing data instance with the index tag "flow" is another mapped value, and so on. After obtaining the first mapping characteristic parameter instance, the server randomly selects the first index tag to be randomly deleted from the multiple index tags corresponding to the instance. Assume that the index tags of the first mapping characteristic parameter instance are "water level", "flow", "sluice gate opening", "pump station operating power", etc. The server randomly selects the index tag "sluice gate opening" as the first index tag through tools such as a random number generator. After determining that the first index tag is "sluice gate opening", the server randomly deletes the mapping processing data instance on the index tag "sluice gate opening" in the first mapping characteristic parameter instance according to the random deletion information. For example, the original mapping processing data instance of "sluice gate opening" is 0.4 (assuming the value obtained after the previous mapping process). Now, according to the rules of random deletion processing, this value is deleted from the first mapping characteristic parameter instance to obtain the first characteristic parameter instance. At this point, the first characteristic parameter instance no longer contains the mapping processing data instance corresponding to the index tag "sluice opening". The server then inputs the first characteristic parameter instance into the first original abnormal event assessment sub-model again. The model first extracts the water scheduling vector representation of the first characteristic parameter instance, and combines the mapping processing data instances corresponding to the remaining index tags into a vector in a certain order to obtain the first water scheduling vector representation instance of the first characteristic parameter instance. For example, if the remaining index tags are "water level" and "flow", and their corresponding mapping processing data instances are 0.6 and another value, respectively, then the first water scheduling vector representation instance can be [0.6, that value].Then, the model performs feature extraction operations on the first water scheduling vector representation instance, and extracts features that can reflect the characteristics and laws of water scheduling from this vector representation instance. For example, the water level change trend features (such as rising rate, falling rate, etc.) are extracted from the numerical values ​​corresponding to the water level, and the seasonal characteristics and periodic characteristics of the flow are extracted from the numerical values ​​corresponding to the flow, so as to obtain the first sample water scheduling features corresponding to the first water scheduling vector representation instance. There is an execution unit corresponding to the random deletion training target in the server. This execution unit will receive the first sample water scheduling feature and perform random deletion inference processing on it according to the pre-set rules and algorithms. Assume that the first sample water scheduling feature contains water level change trend features (such as a rising rate of 0.1 m / h) and seasonal characteristics of flow (such as summer flow is 30% higher than winter flow). According to the requirements of the random deletion training target, the execution unit may perform random deletion inference on one of the features. For example, according to a certain probability (assuming that this probability is set based on historical data and model training requirements), the execution unit decides to perform random deletion inference processing on the seasonal characteristics of the flow, and obtains the inferred mapping processing data corresponding to the random deletion information in the first feature parameter instance through a specific inference algorithm (determined by the structure and training method of the model). Assume that after the inference processing, the inferred mapping processing data obtained indicates that the change of the flow in a certain season has a new inferred value (this value is obtained based on the inference and data processing of the model). The server finally determines the error parameter corresponding to the random deletion training target based on the mapping processing data instance on the first index mark in the first mapping feature parameter instance (which has been deleted before, here refers to the original value, that is, the mapping processing data instance 0.4 of "sluice opening") and the inferred mapping processing data (such as the inferred value of the change of the flow in a certain season obtained before). Specifically, the server will compare the actual situation represented by the original "sluice opening" mapping processing data instance with the inferred situation represented by the inferred mapping processing data obtained after the random deletion inference processing. If there is a large difference between the two, for example, the original sluice opening has a certain impact on water scheduling, but the flow situation has changed significantly after inference, resulting in the overall assessment of water scheduling not being consistent with the actual situation, then a large error parameter will be generated. This error parameter reflects the degree of deviation between the result of the model processing the first training instance and the actual situation under the random deletion training target, so that the first original abnormal event assessment sub-model can be further optimized based on this error parameter, so that it can be more accurately evaluated and predicted when facing similar situations.

[0028] In an embodiment of the present invention, the plurality of training targets include a local permutation training target; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics. The implementation can be performed through the following examples.

[0029] Perform feature mapping processing on the first training instance based on the first original abnormal event assessment sub-model to obtain a first mapping feature parameter instance of the first training instance; the first mapping feature parameter instance includes mapping processing data instances on multiple index tags; Randomly selecting a second index marker to be partially replaced from a plurality of index markers corresponding to the first mapping feature parameter instance; Performing local replacement processing on the mapping processing data instance on the second index mark in the first mapping feature parameter instance according to preset element information to obtain a second feature parameter instance; the preset element information is different from the mapping processing data instance on the second index mark; Based on the first original abnormal event assessment sub-model, the water scheduling vector representation of the second feature parameter instance is extracted to obtain a second water scheduling vector representation instance of the second feature parameter instance, and a feature extraction operation is performed on the second water scheduling vector representation instance to obtain a second sample water scheduling feature corresponding to the second water scheduling vector representation instance; the second sample water scheduling feature belongs to the sample water scheduling feature; Based on the execution unit corresponding to the local replacement training target, performing local replacement inference processing on the second sample water quantity scheduling feature to obtain local replacement inference results corresponding to the multiple index marks; According to the local permutation inference results corresponding to the second index mark and the multiple index marks, an error parameter corresponding to the local permutation training objective is determined.

[0030] In an embodiment of the present invention, exemplarily, the server selects a first training instance from historical operation data, covering water scheduling data such as reservoir water level, river flow, sluice opening, and pump station operating power. For example, at a certain moment, the reservoir water level is 130 meters, the river flow is 800 cubic meters per second, the sluice opening is 30%, and the pump station operating power is 250 kilowatts. After this instance is input into the first original abnormal event assessment sub-model, feature mapping processing is performed. Taking the reservoir water level as an example, according to the historical water level range (80 meters to 150 meters), it is linearly mapped to the [0,1] interval. The 130-meter water level can be mapped to obtain a mapping processing data instance of 0.6. Data such as river flow are also processed according to their respective mapping rules to obtain a first mapping feature parameter instance, including mapping processing data instances corresponding to multiple index tags such as "water level" and "flow". Then, the server randomly selects a second index tag to be partially replaced from these index tags, such as selecting "flow". According to the preset element information, the mapping processing data instance on the "flow" index tag is locally replaced. The original numerical value (assuming 0.4) corresponding to the flow rate of 800 cubic meters per second is replaced by the mapping data instance (assuming 0.3) obtained by the same mapping rule according to the historical typical flow value (such as 500 cubic meters per second), thereby obtaining the second characteristic parameter instance. Subsequently, the second characteristic parameter instance is input into the model again. First, the water scheduling vector representation is extracted, and the mapping data instances corresponding to each index mark are sequentially combined into vectors to obtain the second water scheduling vector representation instance, such as [0.6, 0.3, 0.3, 0.25]. Then, the feature extraction operation is performed to extract the water level change trend, the flow-related features after replacement, the water gate opening change influence features, the pump station operation power stability features, etc., to obtain the second sample water scheduling features. The execution unit corresponding to the local replacement training target in the server receives the second sample water scheduling features and performs local replacement inference processing. For each index mark related feature, such as "water level" infers its change trend under the replaced flow, "flow" infers the subsequent changes, etc., to obtain multiple local replacement inference results. Finally, the error parameter is determined based on the local permutation inference results corresponding to the second index mark "flow" and each index mark. Before permutation, the actual operation data has its own rules. If there is a significant upward trend in the water level, the sluice gate opening and the pump station operation power adjustment after permutation, the difference between the actual and permutation inference processing results is quantified as an error parameter, which is used to reflect the deviation between the model's processing results of the first training instance and the actual situation, so as to subsequently optimize the first original abnormal event assessment sub-model. Through such a data processing and analysis process, the model is continuously improved so that it can more accurately serve the operation evaluation and management of water conservancy projects.

[0031] In an embodiment of the present invention, the plurality of training targets include an interference addition training target; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics. The implementation can be performed through the following examples.

[0032] randomly inserting interference data into the target water demand parameter of the first training instance to obtain an interference training instance, and performing feature mapping processing on the interference training instance to obtain interference mapping feature parameters of the interference training instance; Performing feature mapping processing on the first training instance to obtain a first mapping feature parameter instance of the first training instance; Based on the first original abnormal event assessment sub-model, the interference mapping feature parameter is subjected to water scheduling vector representation extraction to obtain a third water scheduling vector representation instance of the interference mapping feature parameter, and a feature extraction operation is performed on the third water scheduling vector representation instance to obtain a third sample water scheduling feature corresponding to the third water scheduling vector representation instance; the third sample water scheduling feature belongs to the sample water scheduling feature; Based on the first original abnormal event assessment sub-model, the first mapping feature parameter instance is subjected to water scheduling vector representation extraction to obtain a fourth water scheduling vector representation instance of the first mapping feature parameter instance, and a feature extraction operation is performed on the fourth water scheduling vector representation instance to obtain a fourth sample water scheduling feature corresponding to the fourth water scheduling vector representation instance; the fourth sample water scheduling feature belongs to the sample water scheduling feature; According to the third sample water volume scheduling characteristic and the fourth sample water volume scheduling characteristic, an error parameter corresponding to the interference addition training target is determined.

[0033] In an embodiment of the present invention, it is assumed that the target water conservancy project is a large-scale comprehensive water conservancy hub, including multiple reservoirs, a complex irrigation channel network, numerous sluice gates, pump stations and other facilities. The server is responsible for processing and analyzing the relevant data of the water conservancy project to ensure its normal operation and accurately evaluate abnormal events. The server selects a first training instance from the historical operation data of the water conservancy project. This first training instance contains data related to water scheduling in many aspects such as reservoir water level, river flow, sluice opening, and target water demand parameters (such as specific values ​​of agricultural irrigation water demand in a certain area, specific values ​​of industrial water demand, etc.). Taking the agricultural irrigation water demand parameter as an example, it is assumed that the normal irrigation water demand in a certain area during a specific time period is 5,000 cubic meters per day. The server randomly inserts interference data on this target water demand parameter according to the requirements of the interference addition training target. For example, a fluctuation value is randomly added to make the irrigation water demand in the area become 6,000 cubic meters per day, thereby obtaining an interference training instance. Then, the server performs feature mapping processing on this interference training instance. For the reservoir water level, if its actual value is 120 meters, according to the set linear mapping rule based on the historical water level range (assuming the lowest water level is 80 meters and the highest water level is 150 meters), it is mapped to the [0,1] interval, and the mapped value may be 0.6. Similarly, the data such as river flow, sluice opening, and new irrigation water demand (6000 cubic meters) are also processed according to their respective mapping rules, and finally the interference mapping feature parameters of the interference training instance are obtained. At the same time, the server also performs feature mapping processing on the original first training instance. Still taking the data mentioned above as an example, for the reservoir water level of 120 meters, the corresponding mapping processing data instance is obtained according to the same mapping rule (assuming it is also 0.6, which may actually be different due to different mapping rule details). The data such as river flow, sluice opening, and the original agricultural irrigation water demand (5000 cubic meters) are processed by their respective mapping methods to obtain the first mapping feature parameter instance of the first training instance. The server inputs the obtained interference mapping feature parameters into the first original abnormal event assessment sub-model. First, the model extracts the water dispatch vector representation of the interference mapping feature parameters. The mapping processing data instances corresponding to each data are combined into a vector in a certain order to obtain the third water dispatch vector representation instance of the interference mapping feature parameters. For example, if after the previous processing, the reservoir water level mapping processing data instance is 0.6, the river flow mapping processing data instance is another value (assuming it is 0.4), the sluice opening mapping processing data instance is 0.3, and the new irrigation water demand mapping processing data instance is 0.5 (corresponding to 6000 cubic meters after mapping), then the third water dispatch vector representation instance can be [0.6, 0.4, 0.3, 0.5]. Then, the model performs feature extraction operations on the third water dispatch vector representation instance.Extract the water level change trend characteristics, seasonal characteristics of flow, influence characteristics of sluice opening change, and influence characteristics of new irrigation water demand change from this vector representation instance to obtain the third sample water scheduling characteristics corresponding to the third water scheduling vector representation instance. Similarly, the server inputs the first mapping feature parameter instance into the first original abnormal event assessment sub-model. First, perform water scheduling vector representation extraction, combine the mapping processing data instances corresponding to each data in a certain order, and obtain the fourth water scheduling vector representation instance of the first mapping feature parameter instance. Assume that each data mapping processing data instance is similar to the previous one (but the irrigation water demand is the original mapping value corresponding to 5000 cubic meters), such as [0.6, 0.4, 0.3, 0.4] (the last value here corresponds to the original irrigation water demand after mapping). Then, the model performs feature extraction operations on the fourth water scheduling vector representation instance, extracting water level change trend characteristics, seasonal characteristics of flow, influence characteristics of sluice opening change, and influence characteristics related to the original irrigation water demand, and obtain the fourth sample water scheduling characteristics corresponding to the fourth water scheduling vector representation instance. The server finally determines the error parameter corresponding to the interference addition training target based on the third sample water scheduling feature and the fourth sample water scheduling feature. For example, due to the insertion of interference data in the third sample water scheduling feature, the impact characteristics brought about by the new change in irrigation water demand may show that it is necessary to increase the water release from the reservoir to meet the irrigation demand, while the fourth sample water scheduling feature shows that the demand can be met by following the normal water release rhythm based on the original data. This difference between the two in responding to irrigation water demand, as well as the different performances that may be caused by changes in irrigation water demand in terms of water level, flow, sluice opening, etc., will be quantified as an error parameter to reflect the deviation between the result of the model processing the first training instance and the actual situation (i.e., the situation when no interference data is added) under the interference addition training target, so as to optimize and adjust the first original abnormal event assessment sub-model later.

[0034] In an embodiment of the present invention, the plurality of training targets include an approximate target value training target; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics. The implementation can be performed through the following examples.

[0035] Performing feature mapping processing on the first training instance to obtain a first mapping feature parameter instance of the first training instance; Based on the first original abnormal event assessment sub-model, the first mapping feature parameter instance is subjected to water scheduling vector representation extraction to obtain a fourth water scheduling vector representation instance of the first mapping feature parameter instance, and a feature extraction operation is performed on the fourth water scheduling vector representation instance to obtain a fourth sample water scheduling feature corresponding to the fourth water scheduling vector representation instance; the fourth sample water scheduling feature belongs to the sample water scheduling feature; According to the execution unit corresponding to the approximate target value training target, the fourth sample water quantity scheduling feature is subjected to abnormal type recognition processing to obtain the training abnormal type recognition confidence corresponding to the first training instance; Acquire a label generation model corresponding to the first original abnormal event assessment sub-model, perform abnormal type recognition processing on the first training instance according to the label generation model, and obtain an approximate target value abnormal event prediction confidence corresponding to the first training instance; According to the training abnormality type recognition confidence and the approximate target value abnormal event prediction confidence, the error parameter corresponding to the approximate target value training target is determined.

[0036] In an embodiment of the present invention, exemplarily, the server selects a first training instance from the historical operation data of the water conservancy project, which contains a lot of water scheduling related information such as reservoir water level, river flow, sluice opening, pump station operating power and water demand in different regions. For example, at a certain moment, the reservoir water level is 120 meters, the flow of a certain section of the river is 800 cubic meters per second, the sluice opening is 30%, the pump station operating power is 250 kilowatts, and the agricultural irrigation water demand in a certain area is 5,000 cubic meters per day. This instance is input into the first original abnormal event assessment sub-model, and feature mapping processing is performed first. Taking the reservoir water level as an example, referring to the historical water level range (assuming a minimum of 80 meters and a maximum of 150 meters) and the set linear mapping rules, the 120-meter water level is mapped to the [0,1] interval, and the value may be 0.6. Other data such as river flow, sluice opening, etc. are also processed according to their respective rules, and finally the first mapping feature parameter instance of the first training instance is obtained. Next, the instance is input into the model again to extract the vector representation of water scheduling, and the mapped data instances are combined into vectors in order, such as [0.6, 0.4 (assuming that it corresponds to 800 cubic meters per second after mapping), 0.3, 0.25, 0.5 (assuming that it corresponds to 5000 cubic meters per day after mapping)], which is the fourth water scheduling vector representation instance. Subsequently, the model performs feature extraction operations on it, extracting features such as water level change trend, flow seasonality, the impact of sluice opening changes, pump station operation power stability and the impact of irrigation water demand changes, and obtains the fourth sample water scheduling feature. The execution unit of the server receives these features and performs abnormal type identification processing according to specific algorithms and rules. For example, if the water level has a slow rising trend, the current dry season but the flow is slightly higher, the frequency of sluice adjustment increases, etc., the execution unit determines whether there are abnormal types such as flood control risk and insufficient irrigation water supply based on this and the preset standards, and obtains the corresponding training abnormal type identification confidence, such as 0.3 for flood control risk confidence and 0.2 for insufficient irrigation water supply confidence. At the same time, the server inputs the first training instance into the label generation model trained based on a large amount of historical data and known annotation information, which also makes judgments based on data features. For example, for flood control risk, the label generation model obtains a prediction confidence of 0.25 for abnormal events of approximate target value; for insufficient irrigation water supply, the confidence is 0.18. Finally, the error parameter corresponding to the training target of the approximate target value is determined based on the obtained training abnormal type recognition confidence and the prediction confidence of abnormal events of approximate target value. Taking flood control risk as an example, the difference between the two is 0.3 - 0.25 = 0.05, which reflects the degree of deviation between the model processing result and the actual situation (assuming that the label generation model is closer to the actual situation). The difference is also calculated for other abnormal types, and the error parameters are combined to facilitate the subsequent optimization and adjustment of the first original abnormal event evaluation sub-model.

[0037] In an embodiment of the present invention, the target water scheduling feature is determined by a feature encoder in a first abnormal event assessment submodel performing a feature extraction operation on the water scheduling vector representation; the first abnormal event assessment submodel is a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project; the target abnormal event assessment model also includes a second abnormal event assessment submodel arranged after the feature encoder; The abnormal event prediction processing is performed on the target water volume scheduling characteristics according to each task type to obtain the abnormal event prediction confidence corresponding to each task type, which can be implemented through the following examples.

[0038] Determining a second abnormal event assessment sub-model from the target abnormal event assessment model; the second abnormal event assessment sub-model includes a knowledge fusion network corresponding to each task type; Based on the knowledge fusion network corresponding to each task type, the target water volume scheduling characteristics are processed for abnormal event prediction to obtain the abnormal event prediction confidence corresponding to each task type; a knowledge fusion network is used to determine the abnormal event prediction confidence corresponding to a task type.

[0039] In an embodiment of the present invention, exemplarily, when processing water conservancy project related data, the server operates according to the existing target abnormal event assessment model. The model is intended to evaluate abnormal events that may occur in water conservancy projects, and includes a feature encoder and a second abnormal event assessment sub-model set thereafter. When abnormal event prediction processing is to be performed, the server first accurately determines the second abnormal event assessment sub-model from the target abnormal event assessment model. This sub-model internally constructs corresponding knowledge fusion networks for different task types, such as the knowledge fusion network corresponding to the flood control task type, the knowledge fusion network corresponding to the irrigation and water supply task type, and the knowledge fusion network corresponding to the power generation task type. Taking the flood control task type as an example, the server extracts the target water scheduling features determined by performing feature extraction operations on the water scheduling vector representation through the feature encoder. These features may include the changing trend of reservoir water level (such as the recent water level rise rate), the seasonal fluctuation of river flow (such as the large increase in flow in the rainy season), the adjustment frequency of the sluice opening, etc. The server inputs these target water scheduling features into the knowledge fusion network corresponding to the flood control task type. This knowledge fusion network is constructed based on a large amount of historical flood control data and relevant expert knowledge. It will conduct in-depth analysis and fusion processing on the input features. For example, the water level rise rate is compared with the water level rise during historical floods, combined with the current river flow fluctuations and the flow change pattern during previous floods, and then considers multiple factors such as the impact of sluice opening adjustment on flood control. Through these comprehensive analyses and processing, the knowledge fusion network finally outputs the corresponding abnormal event prediction confidence of the flood control task type. Assume that after calculation and analysis, it is concluded that under the current target water scheduling characteristics, the abnormal event prediction confidence of the flood control task is 0.6, that is, there is a 60% probability that flood control-related abnormal events will occur. Similarly, for the irrigation water supply task type, the server inputs the target water scheduling characteristics into its corresponding knowledge fusion network. The network will consider factors such as the matching of reservoir water level and irrigation water supply demand, the impact of sluice opening on water supply distribution, and whether river flow can meet irrigation demand, and output the abnormal event prediction confidence of the irrigation water supply task after processing. For other task types such as power generation task types, a similar process is followed to input the target water volume scheduling characteristics into their corresponding knowledge fusion networks. The knowledge fusion network analyzes and processes the target water volume scheduling characteristics based on its own construction logic and data basis, thereby obtaining the corresponding abnormal event prediction confidence of each task type, providing an important basis for the subsequent operation decisions of water conservancy projects.

[0040] In the embodiments of the present invention, the following implementation modes are also provided.

[0041] Acquire a second sample training array; the second training instance included in the second sample training array is associated with a task type target value; the task type target value includes a plurality of task type target value values ​​corresponding to the plurality of task types, one task type corresponds to a task type target value value, and the plurality of task type target values ​​are determined according to the task type corresponding to the second training instance; Acquire a first abnormal event assessment submodel that has completed training, perform a vector feature extraction operation on the second training instance based on the first abnormal event assessment submodel to obtain a fifth water scheduling vector representation instance of the second training instance, perform a feature extraction operation on the fifth water scheduling vector representation instance to obtain a fifth sample water scheduling feature corresponding to the fifth water scheduling vector representation instance; Acquire multiple basic knowledge fusion networks corresponding to the multiple task types, perform task type analysis on the fifth sample water quantity scheduling characteristics according to the multiple basic knowledge fusion networks, and obtain multiple task type analysis confidences corresponding to the multiple basic knowledge fusion networks; one task type corresponds to one basic knowledge fusion network, and one basic knowledge fusion network is used to determine one task type analysis confidence; A second target error parameter is determined according to the analysis confidences of the multiple task types and the magnitudes of the target values ​​of the multiple task types. According to the second target error parameter, the model parameters of the multiple basic knowledge fusion networks are optimized, the optimized multiple basic knowledge fusion networks are used as the multiple knowledge fusion networks, and the second abnormal event assessment sub-model is determined according to the multiple knowledge fusion networks.

[0042] In an embodiment of the present invention, exemplarily, the server carefully selects and sorts out the second sample training array from the historical operation data of the water conservancy project. Each second training instance here is associated with a task type target value. For example, for the flood control task type, if a second training instance corresponds to the water conservancy project related data before a flood (such as reservoir water level, river flow, sluice opening, etc.), the value of its task type target value may be set to 1, indicating that the possibility of abnormal events (flood occurrence) in the flood control task under this condition is high; for the irrigation water supply task type, if another second training instance is data from a period of time when irrigation water is tight, the value of its task type target value may be set to 0.5, indicating that the possibility of abnormal events such as insufficient water supply in the irrigation water supply task is medium. The server obtains the first abnormal event evaluation sub-model that has been trained, and inputs the second training instances in the second sample training array into the model in sequence. First, the model performs a vector feature extraction operation on the second training instance, such as converting the reservoir water level, river flow, sluice opening, etc. in the instance into a vector form according to certain rules, and obtains the fifth water volume scheduling vector representation instance of the second training instance. Next, a feature extraction operation is performed on the fifth water scheduling vector representation instance to extract the fifth sample water scheduling features that can reflect the characteristics of water scheduling, such as water level change trend, flow seasonality, and sluice opening adjustment frequency. The server obtains multiple basic knowledge fusion networks corresponding to multiple task types (such as flood control, irrigation water supply, power generation, etc.). The extracted fifth sample water scheduling features are input into the corresponding basic knowledge fusion network respectively. Taking the basic knowledge fusion network corresponding to the flood control task type as an example, it will combine historical flood control data and relevant professional knowledge to conduct an in-depth analysis of the input fifth sample water scheduling features. For example, according to the water level change trend, it is judged whether it conforms to the water level rise pattern before the historical flood, and the possibility of flood occurrence is analyzed in combination with the flow characteristics. After analysis, a task type analysis confidence is output. Assume that for this flood control task type, the output task type analysis confidence is 0.7, indicating that the model believes that the possibility of abnormal events in the flood control task under the current input features is high. Similarly, for the basic knowledge fusion networks corresponding to other task types such as irrigation water supply and power generation, the fifth sample water scheduling characteristics will be processed according to their respective analysis logics, and their respective task type analysis confidences will be output respectively. The server determines the second target error parameter based on the values ​​of multiple task type analysis confidences and the corresponding multiple task type target values. For example, for the flood control task type, the task type analysis confidence is 0.7, while the value of its task type target value is 1. The difference between the two (such as by calculating the absolute value of the difference) reflects the degree of prediction deviation of the current basic knowledge fusion network on this task type. By combining the deviations of each task type, the second target error parameter can be determined.Then, according to the second target error parameter, the server uses a suitable optimization algorithm (such as gradient descent, etc.) to optimize the model parameters of multiple basic knowledge fusion networks. After optimization, these basic knowledge fusion networks are used as multiple knowledge fusion networks, and the second abnormal event evaluation sub-model is determined based on them, so that abnormal events can be predicted more accurately in subsequent task type analysis.

[0043] In an embodiment of the present invention, the water scheduling vector representation is determined by extracting the water scheduling vector representation of the mapping feature parameters by the vector representation layer in the first abnormal event assessment submodel; the mapping feature parameters are determined by the data access layer in the first abnormal event assessment submodel performing feature mapping processing on the water scheduling parameters; the first abnormal event assessment submodel is a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project; the target abnormal event assessment model also includes a third abnormal event assessment submodel arranged after the vector representation layer; The influencing factor extraction processing is performed on the water scheduling vector representation to obtain the corresponding influencing factor weighting coefficients of each task type, which can be implemented through the following example.

[0044] Determining a third abnormal event assessment sub-model from the target abnormal event assessment model; Based on the third abnormal event assessment sub-model, the water quantity scheduling vector representation is subjected to influencing factor extraction processing to obtain corresponding influencing factor weighting coefficients of each task type.

[0045] In an embodiment of the present invention, exemplarily, when processing water conservancy project data, the server works according to the existing target abnormal event assessment model. This model is intended to comprehensively assess various abnormal events that may occur in water conservancy projects, and includes multiple components such as a data access layer, a vector representation layer, and a third abnormal event assessment submodel set after the vector representation layer. When it is necessary to extract the influencing factor of the water scheduling vector representation, the server first accurately determines the third abnormal event assessment submodel from the target abnormal event assessment model. This submodel is a key part specifically used to analyze the degree of influence of various factors on water scheduling under different task types, and then determine the weighted coefficient of the influencing factor. Taking the flood control task type as an example, the server obtains the water scheduling vector representation obtained through a series of processing before. This vector representation contains a lot of information related to water scheduling, such as the vector value corresponding to the reservoir water level, the vector value corresponding to the river flow, the vector value corresponding to the sluice opening, and the vector value corresponding to the pump station operating power. The server inputs this water scheduling vector representation into the third abnormal event assessment submodel. Based on a large amount of historical flood control data and relevant professional knowledge, this sub-model will conduct an in-depth analysis of the input vector representation to extract the weighted coefficient of the influencing factor. For the factor of reservoir water level, the model will analyze its importance in the flood control task. For example, if historical data shows that whenever the reservoir water level rises rapidly and exceeds a certain critical value, the risk of flooding will increase significantly. Then in the current water volume dispatch vector representation, the water level situation reflected by the vector value corresponding to the reservoir water level will be given a higher weighted coefficient of the influencing factor. Assume that according to the analysis, the weighted coefficient of the influencing factor of the reservoir water level under the flood control task type is determined to be 0.5. For river flow, if in the flood control scenario, the size and change trend of the flow have an important impact on the formation and development of floods. When the river flow continues to increase and exceeds the normal flood discharge capacity, it is easy to cause floods. Therefore, according to the vector value corresponding to the river flow in the current water volume dispatch vector representation and the comparative analysis of historical data, the weighted coefficient of the influencing factor of the river flow under the flood control task type may be determined to be 0.3. Similarly, for factors such as sluice gate opening and pump station operating power, the third abnormal event assessment sub-model will also analyze them according to their actual impact on flood control tasks. For example, the reasonable adjustment of sluice gate opening is crucial to controlling the reservoir discharge and regulating river flow to cope with floods. If the current opening shows that it has a certain impact on flood control, it may be given an impact factor weighting coefficient of 0.1 after analysis; the pump station operating power has a relatively small impact on flood control tasks, and its impact factor weighting coefficient may be determined to be 0.1.Similarly, for other task types such as irrigation water supply task types and power generation task types, the server will also input the corresponding water scheduling vector representation into the third abnormal event assessment sub-model, and determine the weighted coefficients of the influencing factors corresponding to each factor under each task type by analyzing the importance of each factor in the task according to the characteristics and requirements of each task, so as to comprehensively and accurately evaluate the influence of each factor on water scheduling under different task types.

[0046] In an embodiment of the present invention, the target abnormal event assessment model also includes a second abnormal event assessment sub-model for determining the abnormal event prediction confidence corresponding to each task type; the embodiment of the present invention also provides the following implementation manner.

[0047] Acquire a third sample training array; the third training instance included in the third sample training array is associated with an abnormal event target value; Acquire a first abnormal event assessment submodel that has completed training, perform a vector feature extraction operation on the third training instance based on the first abnormal event assessment submodel to obtain a sixth water scheduling vector representation instance of the third training instance, perform a feature extraction operation on the sixth water scheduling vector representation instance to obtain a sixth sample water scheduling feature corresponding to the sixth water scheduling vector representation instance; Acquire the trained second abnormal event assessment sub-model, perform abnormal event prediction processing on the sixth sample water quantity scheduling feature based on multiple knowledge fusion networks in the second abnormal event assessment sub-model, and obtain the training abnormal event prediction confidence corresponding to each task type; a knowledge fusion network is used to determine the training abnormal event prediction confidence corresponding to a task type; Based on the third original abnormal event assessment sub-model, the sixth water quantity scheduling vector representation is subjected to influence factor extraction processing to obtain training influence factor weighting coefficients corresponding to each task type; Determine the training abnormal event assessment confidence of the virtual water conservancy project corresponding to the third training instance according to the training abnormal event prediction confidence corresponding to each task type and the training influence factor weighting coefficient corresponding to each task type; A third target error parameter is determined according to the abnormal event target value and the training abnormal event assessment confidence; model parameters of the third original abnormal event assessment sub-model are optimized according to the third target error parameter; the optimized third original abnormal event assessment sub-model is used as the third abnormal event assessment sub-model; and the target abnormal event assessment model is determined according to the first abnormal event assessment sub-model, the second abnormal event assessment sub-model and the third abnormal event assessment sub-model.

[0048] In an embodiment of the present invention, exemplarily, the server carefully selects and sorts out the third sample training array from the massive historical operation data of the water conservancy project, wherein each third training instance is associated with an abnormal event target value. For flood control tasks, if a third training instance corresponds to relevant data when a flood occurs, such as reservoir water level, river flow, sluice opening, etc., its abnormal event target value is set to 1, which means that a serious abnormality (flood occurs) has occurred in the flood control task; and for irrigation water supply tasks, if the instance is data from the period of insufficient irrigation water supply, the abnormal event target value may be set to 0.5, indicating that a more obvious abnormality has occurred. The server first obtains the first abnormal event evaluation sub-model that has been trained, and inputs the third training instances in the third sample training array in sequence. The model first performs a vector feature extraction operation on the instance, and converts the data such as reservoir water level, river flow, sluice opening, etc. into vector form according to specific rules to obtain the sixth water scheduling vector representation instance. Then, the sixth sample water scheduling features that can reflect the characteristics of water scheduling, such as water level change trend, flow seasonality, and sluice opening adjustment frequency, are extracted from it. Subsequently, the server obtains the trained second abnormal event assessment sub-model, which contains multiple knowledge fusion networks corresponding to different task types. The extracted sixth sample water scheduling features are input into the corresponding knowledge fusion networks respectively. Taking the knowledge fusion network corresponding to the flood control task as an example, it will combine historical flood control data with professional knowledge to deeply analyze the input features. For example, based on the trend of water level changes, it is judged whether it fits the water level rise pattern before the historical flood, and the possibility of flood occurrence is analyzed in combination with flow characteristics, etc., and then the training abnormal event prediction confidence of the flood control task type is output, which is assumed to be 0.8, that is, the model believes that the possibility of abnormal flood control tasks under the current input features is high. The knowledge fusion networks of other task types also process features according to their own logic and output the corresponding training abnormal event prediction confidence. The server extracts the influencing factors of the sixth water scheduling vector representation based on the third original abnormal event assessment sub-model. Taking the irrigation water supply task as an example, the importance of each factor in the task is analyzed. The water level of the reservoir is crucial. Too low a water level may lead to insufficient water supply, and its training influence factor weighting coefficient may be 0.4; reasonable adjustment of the sluice opening affects the water supply distribution, and the coefficient may be 0.3; the river flow is related to the water supply, and the coefficient may be 0.2; the operating power of the pump station has a small impact, and the coefficient may be 0.1. The coefficients of other task types are determined in the same way. Then, according to the training abnormal event prediction confidence and training influence factor weighting coefficient of each task type, the training abnormal event assessment confidence of the virtual water conservancy project corresponding to the third training instance is determined. For example, for the flood control task type, the relevant data is known, and its contribution value to the overall assessment confidence is obtained through weighted calculation, and then the contribution values ​​of other task types are combined to obtain the training abnormal event assessment confidence of the instance.Then, the third target error parameter is determined based on the abnormal event target value of the third training instance and the confidence level of the training abnormal event assessment. For example, the difference between the two can be used as a part of it, and the final parameter is determined by comprehensively considering the conditions of each task type. Finally, the server optimizes the parameters of the third original abnormal event assessment sub-model based on the third target error parameter using a suitable optimization algorithm (such as gradient descent, etc.), and uses the optimized model as the third abnormal event assessment sub-model, and then combines the trained first and second abnormal event assessment sub-models to determine the complete target abnormal event assessment model, so that it can play a more accurate role in the subsequent abnormal event assessment of water conservancy projects.

[0049] In an embodiment of the present invention, the third abnormal event assessment sub-model includes a decision component and a weight allocation component; The influencing factor extraction processing of the water quantity scheduling vector representation based on the third abnormal event assessment sub-model to obtain the corresponding influencing factor weighting coefficients of each task type can be implemented through the following examples.

[0050] Performing a decision feature extraction operation on the water scheduling vector representation based on the decision component to obtain a decision extraction feature corresponding to the water scheduling vector representation; Based on the weight allocation component, the decision extraction features are subjected to influencing factor identification to obtain corresponding influencing factor weighting coefficients of the various task types.

[0051] In an embodiment of the present invention, exemplarily, the server obtains the water scheduling vector representation after pre-processing, including relevant information such as reservoir water level, river flow, sluice opening, pump station operating power, etc., and inputs it into the decision component of the third abnormal event assessment sub-model. The decision component extracts decision significance features according to the built-in rule algorithm in-depth analysis. Taking the flood control task as an example, attention is paid to features such as the rising trend of reservoir water level, abnormal changes in river flow, and abnormal adjustment of sluice opening. The server then inputs the decision-extracted features into the weight allocation component, which identifies the influencing factors according to preset rules, historical data and professional knowledge, and determines the weighted coefficients of the influencing factors of each task type. For example, in the flood control task, the weighted coefficient of the water level rising trend feature may be 0.5, the abnormal flow change feature may be 0.3, and the abnormal sluice opening adjustment feature may be 0.1; in the irrigation and water supply task, the reservoir water level, sluice opening, river flow, pump station operating power, etc. correspond to different weighted coefficients. In this way, the weight allocation component can provide a basis for evaluating the status of water conservancy projects.

[0052] In an embodiment of the present invention, the multiple task types include a target task type; the embodiment of the present invention also provides the following implementation manner.

[0053] Acquire a target abnormal event assessment model for performing abnormal event assessment on a target water conservancy project, and acquire a basic integrated model for transfer learning from the target abnormal event assessment model; Acquire a fourth sample training array corresponding to the target task type; the fourth sample training array includes a fourth training instance; Processing the fourth training instance based on the target abnormal event assessment model to obtain an abnormal event assessment confidence corresponding to the fourth training instance, and using the abnormal event assessment confidence corresponding to the fourth training instance as an approximate target value instance for training the basic integrated model; Processing the fourth training instance based on the basic integrated model to obtain an integrated abnormal event assessment confidence level corresponding to the fourth training instance; The model parameters of the basic integrated model are optimized according to the integrated abnormal event assessment confidence and the approximate target value instance, and the optimized basic integrated model is used as the target integrated model of the target abnormal event assessment model; the target integrated model is used to perform abnormal event assessment on the water scheduling parameters under the target task type.

[0054] In an embodiment of the present invention, exemplarily, the server first obtains a target abnormal event assessment model for abnormal event assessment of a target water conservancy project. The model has a certain assessment capability after preliminary construction and training. At the same time, a basic integrated model for transfer learning from the target abnormal event assessment model is obtained. This basic integrated model can be obtained based on data training of other similar water conservancy projects or different stages of the same water conservancy project. It has a certain versatility but needs to be further optimized for the current target task type. Next, the server obtains a fourth sample training array corresponding to the target task type, which contains multiple fourth training instances. Taking the irrigation and water supply task type as the target task type as an example, the fourth training instance can be a water conservancy project data record related to irrigation and water supply in different time periods and seasons, such as specific data such as reservoir water level, river flow, sluice opening, and water demand in the irrigation area during a certain period of time. The server inputs the fourth training instances in the fourth sample training array into the target abnormal event assessment model one by one for processing. Still taking the irrigation water supply task type as an example, for a specific fourth training instance, it is assumed that it records data for a certain period of time in summer, the reservoir water level is 120 meters, the river flow is 500 cubic meters per second, the sluice opening is 30%, and the water demand in the irrigation area is 3,000 cubic meters per day. The target abnormal event evaluation model will analyze and evaluate these input data based on its internal algorithm and existing data model to obtain the abnormal event evaluation confidence corresponding to the fourth training instance. For example, after evaluation, it is believed that under the current data situation, the abnormal event evaluation confidence of insufficient irrigation water supply is 0.3, that is, there is a 30% probability that insufficient irrigation water supply will occur. Then, the server uses this abnormal event evaluation confidence as an approximate target value instance for training the basic integrated model. The server then inputs the same fourth training instance into the basic integrated model for processing. The basic integrated model analyzes the input fourth training instance data based on its own structure and algorithm. For the fourth training instance of the above-mentioned irrigation and water supply task type, the basic integrated model may comprehensively consider multiple factors such as reservoir water level, river flow, sluice opening, and water demand, and obtain the corresponding integrated abnormal event assessment confidence of the fourth training instance through its internal integration mechanism (such as comprehensive judgment of multiple sub-models, etc.). Assume that after processing, the obtained integrated abnormal event assessment confidence is 0.25, that is, there is a 25% probability that abnormal events related to irrigation and water supply will occur. The server optimizes the model parameters of the basic integrated model based on the obtained integrated abnormal event assessment confidence and the approximate target value instance. Continuing to take the irrigation and water supply task type as an example, the approximate target value instance is 0.3 (from the assessment result of the target abnormal event assessment model), and the integrated abnormal event assessment confidence is 0.25 (from the processing result of the basic integrated model).There is a certain difference between the two, indicating that there is a deviation between the evaluation results of the basic integrated model and the target abnormal event evaluation model. The server uses a specific optimization algorithm (such as gradient descent, etc.) to adjust the model parameters of the basic integrated model based on this deviation to make its evaluation results closer to the evaluation results of the target abnormal event evaluation model. After multiple rounds of such optimization process, the optimized basic integrated model is used as the target integrated model of the target abnormal event evaluation model. This target integrated model is specifically used to evaluate abnormal events of water scheduling parameters under target task types (such as irrigation water supply task types). It can more accurately judge the abnormal events that may occur under this task type, and provide a more reliable basis for the reasonable scheduling and operation of water conservancy projects.

[0055] In an embodiment of the present invention, the fourth training instance includes a task training instance generated under the target task type; The obtaining of the fourth sample training array corresponding to the target task type may be implemented through the following example.

[0056] Obtaining a task training instance under the target task type; Determine a pending training instance from a set of training instances for instance selection; Matching and associating the pending training instance with the task training instance based on the training instance expansion model to obtain a matching coefficient between the task training instance and the pending training instance; If the matching coefficient meets the instance expansion condition, the undetermined training instance is used as an expanded training instance corresponding to the target task type; The augmented training instance and the task training instance are used as fourth training instances corresponding to the target task type.

[0057] In an embodiment of the present invention, illustratively, the server first specifically selects task training instances related to the irrigation water supply task type from the historical operation data of the water conservancy project. These task training instances record the data conditions of each key link in the actual irrigation water supply process. For example, during the peak irrigation period in spring, a task training instance records that the water level of a reservoir is 110 meters, and the river flow at the entrance of the irrigation channel is 400 cubic meters per second. The corresponding sluice opening is set to 40%, and the water demand in the irrigation area is 2,500 cubic meters per day. Such detailed data records can reflect the actual situation of irrigation water supply at that time, and provide a real and targeted data basis for subsequent model training. The server then determines the pending training instance from the training instance set used for instance selection. This training instance set contains various data records of water conservancy projects in different operation scenarios and different time periods. For example, a data record is selected from the set, which records that the reservoir water level is 105 meters in a certain period, the river flow is 350 cubic meters per second at a certain section, the sluice opening is 35%, and there are some other water demand conditions (such as industrial water, etc.) in the surrounding area during this period as a pending training instance. Although this instance is not completely for the irrigation water supply task type, it contains some factors that may be related to irrigation water supply, so it is selected as an object for further analysis. The server inputs the obtained pending training instance and the task training instance under the previously selected irrigation water supply task type into the training instance expansion model. This training instance expansion model will conduct a comprehensive analysis of the two instances according to a series of set rules and algorithms to determine the matching coefficient between them. For example, for the task training instance mentioned above (the situation during the peak period of spring irrigation) and the pending training instance (the situation that includes multiple water demands in a certain period of time), the model will compare their reservoir water level data, analyze the potential impact of water level differences and water level change trends on irrigation water supply; at the same time, compare river flow data, consider the role of flow size and fluctuations on irrigation water supply; and pay attention to factors such as the similarities and differences in sluice openings and the impact of other water demand situations on irrigation water supply resource allocation. After complex calculations and analysis, it is assumed that the matching coefficient between the two instances is 0.6. This matching coefficient reflects the similarity between the pending training instance and the task training instance in terms of factors related to the irrigation water supply task. The server then determines whether this matching coefficient meets the instance expansion condition. Assume that the instance expansion condition set is that the matching coefficient is greater than or equal to 0.5. Since the matching coefficient obtained above is 0.6, which meets the instance expansion condition, the server uses the pending training instance as the expanded training instance corresponding to the target task type (irrigation water supply task type). Finally, the server uses the expanded training instance and the original task training instance together as the fourth training instance corresponding to the target task type.The fourth training instance set composed in this way includes both actual data records under pure irrigation and water supply task types (task training instances), and other data records with certain relevance after screening and matching (expanded training instances). It can provide a richer and more comprehensive data basis for subsequent model training based on the target task type, and help improve the accuracy of the model's assessment of abnormal events under irrigation and water supply task types.

[0058] In an embodiment of the present invention, determining the abnormal event assessment confidence of the target water conservancy project according to the abnormal event prediction confidence corresponding to each task type and the impact factor weighting coefficient corresponding to each task type can be implemented through the following example.

[0059] Based on the weighted coefficients of the impact factors corresponding to the respective task types, the prediction confidences of the abnormal events corresponding to the respective task types are weighted to obtain the weighted abnormal event prediction coefficients corresponding to the respective task types; the conversion value of the impact factor corresponding to a task type is used to weight the prediction confidences of the abnormal events corresponding to the respective task types; The confidence level of abnormal event assessment of the target water conservancy project is determined according to the weighted abnormal event prediction coefficients corresponding to the various task types.

[0060] In the embodiment of the present invention, for example, in the water conservancy project, multiple task types such as flood control, irrigation and water supply, and power generation are set. The server has obtained the abnormal event prediction confidence and influencing factor weighting coefficient of each task type through models and algorithms. Taking the flood control task as an example, its abnormal event prediction confidence is 0.6, and the influencing factor weighting coefficients of factors such as reservoir water level, river flow, sluice opening, and pump station operating power are 0.4, 0.3, 0.2, and 0.1 respectively. After weighted calculation, the weighted abnormal event prediction coefficient is 0.6. Similarly, the irrigation and water supply task type assumes that the abnormal event prediction confidence is 0.4, and the weighting coefficients of each factor are different, and the weighted abnormal event prediction coefficient is 0.4; the power generation task type and the like also calculate the corresponding coefficients in this way. Afterwards, the server determines the abnormal event assessment confidence of the target water conservancy project, and can use weighted average or simple average methods. If only the above three types of tasks are considered, their weighted abnormal event prediction coefficients are known, and the simple average gives an abnormal event assessment confidence of about 0.43, and the weighted average (assuming the weight of each task type) gives 0.45. Through these calculations, considering the abnormal prediction of each task type and the importance of factors, the overall abnormal event assessment confidence of the target water conservancy project is obtained, which is used to evaluate and decide the overall operation status of the project.

[0061] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned method for optimizing the construction and operation parameters of a water conservancy project and for data management. Figure 2 As shown, Figure 2 The block diagram of the computer device 100 provided in the embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112 and a communication unit 113. To achieve data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are directly or indirectly electrically connected to each other. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0062] For illustrative purposes, the foregoing description is made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise form disclosed. Numerous modifications and variations are possible in accordance with the above teachings. These embodiments are selected and described in order to best illustrate the principles of the present disclosure and its practical application, so that those skilled in the art can best utilize the present disclosure and utilize various embodiments with different modifications to suit the intended specific application.

Claims

1. A method for optimizing operation parameters and managing data for water conservancy project construction, characterized in that: include: Performing a vector feature extraction operation on the water scheduling parameters of the target water conservancy project to obtain a water scheduling vector representation corresponding to the water scheduling parameters; Performing a feature extraction operation on the water scheduling vector representation to obtain a target water scheduling feature corresponding to the water scheduling vector representation; Determine a plurality of task types corresponding to the water scheduling parameters, perform abnormal event prediction processing on the target water scheduling characteristics according to each task type, and obtain abnormal event prediction confidence corresponding to each task type; Extracting influencing factors from the water quantity dispatching vector representation to obtain influencing factor weighting coefficients corresponding to each task type; Based on the weighted coefficients of the impact factors corresponding to the respective task types, the prediction confidences of the abnormal events corresponding to the respective task types are weighted to obtain the weighted abnormal event prediction coefficients corresponding to the respective task types; the conversion value of the impact factor corresponding to a task type is used to weight the prediction confidences of the abnormal events corresponding to the respective task types; Determining the confidence level of abnormal event assessment of the target water conservancy project according to the weighted abnormal event prediction coefficients corresponding to the various task types; When the abnormal event assessment confidence level indicates that an abnormal event has occurred in the target water conservancy project, real-time monitoring data collection is performed on each subsystem and key equipment of the target water conservancy project, and preset governance strategies are pushed to each subsystem for optimization and adjustment based on the abnormal event assessment confidence level.

2. The method according to claim 1, characterized in that The performing of a vector feature extraction operation on the water scheduling parameters of the target water conservancy project to obtain a water scheduling vector representation corresponding to the water scheduling parameters includes: When the water volume dispatching parameters of the target water conservancy project are obtained, a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project is obtained; the target abnormal event assessment model includes a first abnormal event assessment sub-model; Determine the demand information corresponding to the water scheduling parameter based on the first abnormal event assessment submodel, and extract the target water demand parameter of the water scheduling parameter according to the demand information; the first abnormal event assessment submodel includes a data access layer; the target water demand parameter includes multiple target water demand parameters, and the multiple target water demand parameters include a target key water demand parameter; Determine a transformation domain corresponding to the target key water demand parameter according to the demand information corresponding to the target key water demand parameter; the transformation domain includes a plurality of conversion values, each conversion value having a corresponding target water demand value range; Determine the target water demand value range corresponding to the target key water demand parameter based on the data access layer, and use the conversion value corresponding to the target water demand value range corresponding to the target key water demand parameter as the mapping processing data corresponding to the target key water demand parameter; Based on the first abnormal event assessment sub-model, a standardization operation is performed on the mapping processing data corresponding to the target water demand parameter to obtain a mapping characteristic parameter of the water quantity scheduling parameter; A water quantity dispatching vector representation is extracted from the mapping characteristic parameters to obtain a water quantity dispatching vector representation of the mapping characteristic parameters.

3. The method according to claim 1, characterized in that: The target water quantity dispatching feature is determined by a feature encoder in a first abnormal event assessment submodel performing a feature extraction operation on the water quantity dispatching vector representation; the first abnormal event assessment submodel is a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project; The method further comprises: Acquire a first sample training array; the first training instance included in the first sample training array is a training instance that is not configured with an abnormal event target value; Acquire a plurality of training targets for the first original abnormal event evaluation sub-model; Processing the first training instance based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and determining multiple error parameters corresponding to the multiple training targets according to the sample water scheduling characteristics; one training target corresponds to one error parameter; A first target error parameter is determined according to the multiple error parameters, model parameters of the first original abnormal event assessment submodel are optimized according to the first target error parameter, and the optimized first original abnormal event assessment submodel is used as the first abnormal event assessment submodel.

4. The method according to claim 3, characterized in that The multiple training targets include randomly deleting training targets; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics, including: Performing feature mapping processing on the first training instance based on the first original abnormal event assessment sub-model to obtain a first mapping feature parameter instance of the first training instance; the first mapping feature parameter instance includes mapping processing data instances on multiple index tags; Randomly selecting a first index marker to be randomly deleted from a plurality of index markers corresponding to the first mapping feature parameter instance; Performing random deletion processing on the mapping processing data instance on the first index mark in the first mapping feature parameter instance according to the random deletion information to obtain a first feature parameter instance; Based on the first original abnormal event assessment sub-model, the first feature parameter instance is subjected to water scheduling vector representation extraction to obtain a first water scheduling vector representation instance of the first feature parameter instance, and a feature extraction operation is performed on the first water scheduling vector representation instance to obtain a first sample water scheduling feature corresponding to the first water scheduling vector representation instance; the first sample water scheduling feature belongs to the sample water scheduling feature; Based on the execution unit corresponding to the random deletion training target, performing random deletion inference processing on the first sample water scheduling feature to obtain inference mapping processing data corresponding to the random deletion information in the first feature parameter instance; Determining an error parameter corresponding to the random deletion training target according to the mapping processing data instance on the first index mark in the first mapping feature parameter instance and the inferred mapping processing data; The multiple training targets also include a local permutation training target; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics, including: Performing feature mapping processing on the first training instance based on the first original abnormal event assessment sub-model to obtain a first mapping feature parameter instance of the first training instance; the first mapping feature parameter instance includes mapping processing data instances on multiple index tags; Randomly selecting a second index marker to be partially replaced from a plurality of index markers corresponding to the first mapping feature parameter instance; Performing local replacement processing on the mapping processing data instance on the second index mark in the first mapping feature parameter instance according to preset element information to obtain a second feature parameter instance; the preset element information is different from the mapping processing data instance on the second index mark; Based on the first original abnormal event assessment sub-model, the water scheduling vector representation of the second feature parameter instance is extracted to obtain a second water scheduling vector representation instance of the second feature parameter instance, and a feature extraction operation is performed on the second water scheduling vector representation instance to obtain a second sample water scheduling feature corresponding to the second water scheduling vector representation instance; the second sample water scheduling feature belongs to the sample water scheduling feature; Based on an execution unit corresponding to the local replacement training target, performing local replacement inference processing on the second sample water quantity scheduling feature to obtain local replacement inference results corresponding to the multiple index marks; Determining an error parameter corresponding to the local permutation training objective according to the local permutation inference results corresponding to the second index mark and the multiple index marks; The multiple training targets also include an interference addition training target; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics, including: randomly inserting interference data into the target water demand parameter of the first training instance to obtain an interference training instance, and performing feature mapping processing on the interference training instance to obtain interference mapping feature parameters of the interference training instance; Performing feature mapping processing on the first training instance to obtain a first mapping feature parameter instance of the first training instance; Based on the first original abnormal event assessment sub-model, the interference mapping feature parameter is subjected to water scheduling vector representation extraction to obtain a third water scheduling vector representation instance of the interference mapping feature parameter, and a feature extraction operation is performed on the third water scheduling vector representation instance to obtain a third sample water scheduling feature corresponding to the third water scheduling vector representation instance; the third sample water scheduling feature belongs to the sample water scheduling feature; Based on the first original abnormal event assessment sub-model, the first mapping feature parameter instance is subjected to water scheduling vector representation extraction to obtain a fourth water scheduling vector representation instance of the first mapping feature parameter instance, and a feature extraction operation is performed on the fourth water scheduling vector representation instance to obtain a fourth sample water scheduling feature corresponding to the fourth water scheduling vector representation instance; the fourth sample water scheduling feature belongs to the sample water scheduling feature; Determining an error parameter corresponding to the interference addition training target according to the third sample water volume scheduling feature and the fourth sample water volume scheduling feature; The multiple training targets also include an approximate target value training target; The first training instance is processed based on the first original abnormal event assessment sub-model to obtain sample water scheduling characteristics corresponding to each training target, and multiple error parameters corresponding to the multiple training targets are determined according to the sample water scheduling characteristics, including: Performing feature mapping processing on the first training instance to obtain a first mapping feature parameter instance of the first training instance; Based on the first original abnormal event assessment sub-model, the first mapping feature parameter instance is subjected to water scheduling vector representation extraction to obtain a fourth water scheduling vector representation instance of the first mapping feature parameter instance, and a feature extraction operation is performed on the fourth water scheduling vector representation instance to obtain a fourth sample water scheduling feature corresponding to the fourth water scheduling vector representation instance; the fourth sample water scheduling feature belongs to the sample water scheduling feature; According to the execution unit corresponding to the approximate target value training target, the fourth sample water quantity scheduling feature is subjected to abnormal type recognition processing to obtain the training abnormal type recognition confidence corresponding to the first training instance; Acquire a label generation model corresponding to the first original abnormal event assessment sub-model, perform abnormal type recognition processing on the first training instance according to the label generation model, and obtain an approximate target value abnormal event prediction confidence corresponding to the first training instance; According to the training abnormality type recognition confidence and the approximate target value abnormal event prediction confidence, the error parameter corresponding to the approximate target value training target is determined.

5. The method according to claim 1, characterized in that The target water quantity scheduling feature is determined by a feature encoder in a first abnormal event assessment submodel performing a feature extraction operation on the water quantity scheduling vector representation; the first abnormal event assessment submodel is a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project; the target abnormal event assessment model also includes a second abnormal event assessment submodel arranged after the feature encoder; The abnormal event prediction processing is performed on the target water volume scheduling characteristics according to each task type to obtain the abnormal event prediction confidence corresponding to each task type, including: Determining a second abnormal event assessment sub-model from the target abnormal event assessment model; the second abnormal event assessment sub-model includes a knowledge fusion network corresponding to each task type; Based on the knowledge fusion network corresponding to each task type, the target water volume scheduling characteristics are processed for abnormal event prediction to obtain the abnormal event prediction confidence corresponding to each task type; a knowledge fusion network is used to determine the abnormal event prediction confidence corresponding to a task type.

6. The method according to claim 5, characterized in that The method further comprises: Acquire a second sample training array; the second training instance included in the second sample training array is associated with a task type target value; the task type target value includes a plurality of task type target value values ​​corresponding to the plurality of task types, one task type corresponds to a task type target value value, and the plurality of task type target values ​​are determined according to the task type corresponding to the second training instance; Acquire a first abnormal event assessment submodel that has completed training, perform a vector feature extraction operation on the second training instance based on the first abnormal event assessment submodel to obtain a fifth water scheduling vector representation instance of the second training instance, perform a feature extraction operation on the fifth water scheduling vector representation instance to obtain a fifth sample water scheduling feature corresponding to the fifth water scheduling vector representation instance; Acquire multiple basic knowledge fusion networks corresponding to the multiple task types, perform task type analysis on the fifth sample water quantity scheduling characteristics according to the multiple basic knowledge fusion networks, and obtain multiple task type analysis confidences corresponding to the multiple basic knowledge fusion networks; one task type corresponds to one basic knowledge fusion network, and one basic knowledge fusion network is used to determine one task type analysis confidence; A second target error parameter is determined according to the analysis confidences of the multiple task types and the values ​​of the target values ​​of the multiple task types. According to the second target error parameter, the model parameters of the multiple basic knowledge fusion networks are optimized, and the optimized multiple basic knowledge fusion networks are used as multiple knowledge fusion networks. According to the multiple knowledge fusion networks, the second abnormal event assessment sub-model is determined.

7. The method according to claim 1, characterized in that The water scheduling vector representation is determined by extracting the water scheduling vector representation of the mapping feature parameters by the vector representation layer in the first abnormal event assessment submodel; the mapping feature parameters are determined by the data access layer in the first abnormal event assessment submodel performing feature mapping processing on the water scheduling parameters; the first abnormal event assessment submodel is a target abnormal event assessment model for performing abnormal event assessment on the target water conservancy project; the target abnormal event assessment model also includes a third abnormal event assessment submodel arranged after the vector representation layer; The extracting of influencing factors from the water quantity scheduling vector representation to obtain the corresponding weighted coefficients of influencing factors of each task type includes: Determining a third abnormal event assessment sub-model from the target abnormal event assessment model; the third abnormal event assessment sub-model includes a decision component and a weight allocation component; Performing a decision feature extraction operation on the water scheduling vector representation based on the decision component to obtain a decision extraction feature corresponding to the water scheduling vector representation; Based on the weight allocation component, the decision extraction features are subjected to influencing factor identification to obtain corresponding influencing factor weighting coefficients of the various task types.

8. The method according to claim 7, characterized in that The target abnormal event assessment model also includes a second abnormal event assessment sub-model for determining the abnormal event prediction confidence corresponding to each task type; The method further comprises: Acquire a third sample training array; the third training instance included in the third sample training array is associated with an abnormal event target value; Acquire a first abnormal event assessment submodel that has completed training, perform a vector feature extraction operation on the third training instance based on the first abnormal event assessment submodel to obtain a sixth water scheduling vector representation instance of the third training instance, perform a feature extraction operation on the sixth water scheduling vector representation instance to obtain a sixth sample water scheduling feature corresponding to the sixth water scheduling vector representation instance; Acquire the trained second abnormal event assessment sub-model, perform abnormal event prediction processing on the sixth sample water quantity scheduling feature based on multiple knowledge fusion networks in the second abnormal event assessment sub-model, and obtain the training abnormal event prediction confidence corresponding to each task type; a knowledge fusion network is used to determine the training abnormal event prediction confidence corresponding to a task type; Based on the third original abnormal event assessment sub-model, the sixth water quantity scheduling vector representation is subjected to influence factor extraction processing to obtain training influence factor weighting coefficients corresponding to each task type; Determine the training abnormal event assessment confidence of the virtual water conservancy project corresponding to the third training instance according to the training abnormal event prediction confidence corresponding to each task type and the training influence factor weighting coefficient corresponding to each task type; A third target error parameter is determined according to the abnormal event target value and the training abnormal event assessment confidence; model parameters of the third original abnormal event assessment sub-model are optimized according to the third target error parameter; the optimized third original abnormal event assessment sub-model is used as the third abnormal event assessment sub-model; and the target abnormal event assessment model is determined according to the first abnormal event assessment sub-model, the second abnormal event assessment sub-model and the third abnormal event assessment sub-model.

9. The method according to claim 1, characterized in that: The plurality of task types include a target task type; The method further comprises: Acquire a target abnormal event assessment model for performing abnormal event assessment on a target water conservancy project, and acquire a basic integrated model for transfer learning from the target abnormal event assessment model; Obtaining a task training instance under the target task type; Determine a pending training instance from a set of training instances for instance selection; Matching and associating the pending training instance with the task training instance based on the training instance expansion model to obtain a matching coefficient between the task training instance and the pending training instance; If the matching coefficient meets the instance expansion condition, the undetermined training instance is used as an expanded training instance corresponding to the target task type; The expanded training instance and the task training instance are used as a fourth training instance corresponding to the target task type; the fourth training instance includes the task training instance generated under the target task type; Processing the fourth training instance based on the target abnormal event assessment model to obtain an abnormal event assessment confidence corresponding to the fourth training instance, and using the abnormal event assessment confidence corresponding to the fourth training instance as an approximate target value instance for training the basic integrated model; Processing the fourth training instance based on the basic integrated model to obtain an integrated abnormal event assessment confidence level corresponding to the fourth training instance; The model parameters of the basic integrated model are optimized according to the integrated abnormal event assessment confidence and the approximate target value instance, and the optimized basic integrated model is used as the target integrated model of the target abnormal event assessment model; the target integrated model is used to perform abnormal event assessment on the water scheduling parameters under the target task type.

10. A server system, characterized in that: The method comprises a server, wherein the server is used to execute the method described in any one of claims 1 to 9.

Images