A power project budget estimation method and system

Abstract

The application discloses a kind of power project budget estimation method and system, it is related to data analysis technical field;The historical data of target area are acquired and preprocessed to obtain target historical data;Characteristic data is obtained by extracting features to target historical data, obtain the spatial data of target area, and obtain target database according to characteristic data and spatial data;According to target database, the target model is obtained by updating preset model, and the power project task is obtained, and the final project scheme is obtained by inputting the power project task into target model;If final project scheme completes power project task, then determine the budget of power project.The load data of target area are acquired and preprocessed and extracted features, and the target database is constructed in combination with spatial data, and the preset model is updated and optimized;Final project scheme is obtained after inputting power project task into model, and the budget of power project is determined, which reduces data complexity, improves calculation efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data analysis, and particularly relates to a power project budget estimation method and system. BACKGROUND

[0002] With the rapid development of social economy, as a key field of infrastructure construction, power projects have experienced a long process from traditional power grid construction to smart grid transformation. The early budget estimation method of power projects is obtained by manual calculation, but as the number of power projects increases, manual calculation and existing estimation methods cannot meet the demand.

[0003] Application No. CN117993854A discloses a park power engineering project cost management method based on data analysis; investment budget is made for the main distribution network of the in-construction park power engineering project; investment accounting is made for the in-construction park power engineering project, and the accounting content includes real-time monitoring of the in-construction park power engineering project cost information and progress information, judging whether there is abnormal deviation data according to the project data and warning the deviation data, cost accounting according to the cost information of the in-construction park power engineering project and forming a project report; investment evaluation analysis is made for the in-construction park power engineering project.

[0004] In the prior art, the data dimension to be processed in the estimation process of the power project budget is large, resulting in low calculation efficiency, and the traditional model cannot cope with multiple variables, resulting in poor project simulation accuracy and large budget error. SUMMARY

[0005] The purpose of the present application is to solve the problem of the estimation process of the power project budget, which requires processing of data with multiple dimensions, resulting in low calculation efficiency, and the traditional model cannot cope with multiple variables, resulting in poor project simulation accuracy and large budget error, and a power project budget estimation method and system are proposed.

[0006] In the first aspect of the present application, a power project budget estimation method is first proposed, which comprises:

[0007] Obtain historical data of a target area, and pre-process the historical data to obtain target historical data;

[0008] Feature extraction is performed on the target historical data to obtain feature data, spatial data of the target area is obtained, and a target database is constructed according to the feature data and the spatial data;

[0009] A target model is obtained by updating a preset model according to the target database, and a power project task is obtained, which is input into the target model to obtain a final project scheme;

[0010] If the final project scheme completes the power project task, a power project budget is determined according to the final project scheme.

[0011] Optionally, feature extraction is performed on the target historical data to obtain feature data, including:

[0012] A sample set is determined according to the target historical data, and a weighted distance of each sample in the sample set is calculated; the sample set includes a plurality of samples; each sample is composed of a feature vector and a category label;

[0013] A target probability is calculated according to the weighted distance, and a classification probability is calculated according to the target probability; the target sample is any one of the samples in the sample set;

[0014] If the target probability is greater than a classification threshold, the target sample is retained, otherwise, the target sample is removed.

[0015] Optionally, a target model is obtained by updating a preset model according to the target database, including:

[0016] A parameter combination of the preset model is determined, and a target population is obtained by initializing a population; the parameter combination includes a penalty coefficient, a function parameter and a parameter search range;

[0017] The fitness of each chromosome in the target population is calculated, and an optimal chromosome is determined according to the numerical value of the fitness;

[0018] An individual optimal set is obtained by iteratively updating the target population according to the optimal chromosome, and a search region set is determined according to the individual optimal set; the individual optimal set includes the optimal individual of each chromosome in the iteration process; the search region set includes a plurality of search regions; the optimal individual and the search region correspond one by one;

[0019] Each search region is iteratively updated until a preset condition is met, and an optimal parameter combination is output, and a target model is obtained by updating the preset model according to the optimal parameter combination.

[0020] Optionally, the power project task is input into the target model to obtain a final project scheme, including:

[0021] A task sequence is determined according to the power project task, a first task is input into the target model to obtain a first scheme; the task sequence includes a first task, a second task and a third task; the first task is global variable optimization, the second task is discrete variable optimization, and the third task is continuous variable optimization;

[0022] determining a first optimization target according to the first scheme, inputting the second task and the first optimization target into the target model to obtain a second scheme;

[0023] determining a second optimization target according to the second scheme, inputting the third task and the second optimization target into the target model to obtain a third scheme, and taking the third scheme as a final project scheme.

[0024] Optionally, if the final project scheme completes the power project task, the method further includes:

[0025] simulating according to the final project scheme to obtain a running state parameter, and determining that the final project scheme completes the power project task if the running state parameter meets a preset constraint condition;

[0026] Otherwise, determining that the final project scheme does not complete the power project task, and inputting the power project task into the target model for training until a preset training number is reached to stop the training.

[0027] In a second aspect of the embodiment of the present application, a power project budget estimation system is provided, which includes:

[0028] a data preprocessing module configured to acquire historical data of a target region, and preprocess the historical data to obtain target historical data;

[0029] a database construction module configured to extract features from the target historical data to obtain feature data, acquire spatial data of the target region, and construct a target database according to the feature data and the spatial data;

[0030] a model updating module configured to update a preset model according to the target database to obtain a target model, acquire a power project task, and input the power project task into the target model to obtain a final project scheme;

[0031] a budget determination module configured to determine a power project budget according to the final project scheme if the final project scheme completes the power project task.

[0032] Optionally, the database construction module includes:

[0033] a sample calculation module configured to determine a sample set according to the target historical data, and calculate a weighted distance of each sample in the sample set; the sample set includes a plurality of samples; and each sample is composed of a feature vector and a category label;

[0034] a probability calculation module configured to calculate a target probability of a target sample according to the weighted distance, and calculate a classification probability according to the target probability; the target sample is any one of the samples in the sample set.

[0035] A probability classification module is configured to retain the target sample if the target probability is greater than a classification threshold, and to reject the target sample otherwise.

[0036] Optionally, the model updating module comprises:

[0037] An initialization module is configured to determine a parameter combination of the preset model, and initialize a population to obtain a target population; the parameter combination comprises a penalty coefficient, a function parameter, and a parameter search range.

[0038] An adaptability calculation module is configured to calculate an adaptability of each chromosome in the target population, and determine an optimal chromosome according to a numerical value of the adaptability.

[0039] A local optimization module is configured to iteratively update the target population according to the optimal chromosome to obtain an individual optimal set, and determine a search region set according to the individual optimal set; the individual optimal set comprises optimal individuals of each chromosome in an iteration process; the search region set comprises a plurality of search regions; and the optimal individual and the search region are in one-to-one correspondence.

[0040] A global optimization module is configured to iteratively update each search region until a preset condition is met, and output an optimal parameter combination; and the preset model is updated according to the optimal parameter combination to obtain a target model.

[0041] Optionally, the model updating module further comprises:

[0042] A first execution module is configured to determine a task sequence according to the electric power project task, input a first task into the target model to obtain a first scheme; the task sequence comprises a first task, a second task, and a third task; the first task is global variable optimization, the second task is discrete variable optimization, and the third task is continuous variable optimization.

[0043] A second execution module is configured to determine a first optimization target according to the first scheme, and input the second task and the first optimization target into the target model to obtain a second scheme.

[0044] A third execution module is configured to determine a second optimization target according to the second scheme, input the third task and the second optimization target into the target model to obtain a third scheme, and take the third scheme as a final project scheme.

[0045] Optionally, the budget determination module further comprises:

[0046] The first judging module is used for obtaining a running state parameter by simulation according to the final project scheme, and determining that the final project scheme completes the power project task if the running state parameter meets a preset constraint condition.

[0047] The second judging module is used for determining that the final project scheme does not complete the power project task if not, and inputting the power project task into the target model for training until a preset training number is reached to stop the training.

[0048] The present application has the following beneficial effects:

[0049] The present application provides a power project budget estimation method, historical data of a target region are obtained and preprocessed to obtain target historical data, feature data are obtained by feature extraction on the target historical data, spatial data of the target region are obtained, and a target database is constructed according to the feature data and the spatial data, a target model is obtained by updating a preset model according to the target database, a power project task is obtained, the power project task is input into the target model to obtain a final project scheme, and the power project budget is determined if the final project scheme completes the power project task. The target region load data are obtained and preprocessed and feature extracted, the target database is constructed in combination with the spatial data, the preset model is updated and optimized, the final project scheme is obtained after the power project task is input into the model, and the power project budget is determined, so that the data complexity is reduced, and the calculation efficiency and accuracy are improved. BRIEF DESCRIPTION OF DRAWINGS

[0050] The present application will be further described below in combination with the drawings.

[0051] Figure 1 A flowchart of a power project budget estimation method provided for the embodiment of the present application is shown in the figure.

[0052] Figure 2 A framework diagram of another power project budget estimation system provided for the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. The term "and / or" in this document is only used to describe the associated relationship of associated objects, and can represent three relationships, for example, A and B can represent three cases of A alone, A and B together, and B alone. In addition, the description of "first", "second" and the like in the present application is only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the fact that a person skilled in the art can realize it. When the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, and is not within the protection scope of the present application.

[0054] Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative labor are within the protection scope of the present application.

[0055] The embodiments of the present application provide a power project budget estimation method. Referring to Figure 1 , Figure 1 A flowchart of a power project budget estimation method provided by the embodiments of the present application is provided. The method comprises the following steps:

[0056] S101, obtaining historical data of a target area, and pre-processing the historical data to obtain target historical data;

[0057] S102, performing feature extraction on the target historical data to obtain feature data, obtaining spatial data of the target area, and constructing a target database according to the feature data and the spatial data;

[0058] S103, updating a preset model according to the target database to obtain a target model, obtaining a power project task, and inputting the power project task into the target model to obtain a final project scheme;

[0059] S104, if the final project scheme completes the power project task, determining a power project budget according to the final project scheme.

[0060] Based on the power project budget estimation method provided by the embodiments of the present application, by obtaining target area load data and performing preprocessing and feature extraction, combining spatial data to construct a target database, updating and optimizing a preset model; inputting the power project task into the model to obtain a final project scheme, and determining a power project budget, the data complexity is reduced, and the calculation efficiency and accuracy are improved.

[0061] In an implementation, the target area can be an area for construction of power generation facilities such as thermal power, hydroelectric power, and wind and solar power; the historical data includes historical load data of the target area, weather characteristics such as season, temperature, and date characteristics such as weekdays / holidays; preprocessing operations such as data cleaning to handle missing values such as deletion, filling, and outliers such as identification and correction or removal to ensure data integrity and accuracy; and data conversion including standardization / normilization to make different dimension characteristics comparable, discretization to segment continuous values, and function transformation such as logarithmic conversion.

[0062] In an implementation, core features such as load fluctuation, temperature factor, and geographical distance are screened through neighborhood component analysis to reduce data dimension, remove redundant information, reduce computational complexity, and improve the model's focus on key factors, thereby improving the model's accuracy and efficiency.

[0063] In an implementation, a database associating attribute data and spatial data is established through a geographic information system (spatial data) to associate feature data with spatial data of power facilities such as location coordinates, user distribution area, and power supply node layout, so that the database can support optimization analysis of power facility layout, realize reasonable layout planning of power projects, avoid resource waste and operation risks caused by unreasonable layout, and improve the overall operation efficiency and reliability of the power system.

[0064] In an implementation, a support vector machine model (preset model) is constructed, the features in the database are taken as input, and the load rate optimization, operation error minimization, and classification or regression results are taken as target output; a particle swarm optimization algorithm is used to optimize key parameters of the preset model such as penalty coefficient C and kernel function parameter γ, to minimize prediction error through swarm intelligence search and improve model accuracy.

[0065] In an implementation, when the final project scheme is determined, the power project budget can be calculated according to the scheme, including the cost of engineering and the like.

[0066] In an embodiment, feature extraction is performed on the target historical data to obtain feature data, including:

[0067] A sample set is determined according to the target historical data, and a weighted distance of each sample in the sample set is calculated; the sample set includes multiple samples; each sample is composed of a feature vector and a category label;

[0068] A target probability is calculated according to the weighted distance, and a classification probability is calculated according to the target probability; the target sample is any one of the samples in the sample set;

[0069] If the target probability is greater than the classification threshold, the target sample is retained, otherwise, the target sample is rejected.

[0070] In an implementation, a sample set S={x i ,y i}(i=1,2,…,N), wherein xi is a feature vector, containing load data, weather quantization value, date type and other multi-dimensional features; y i is a class label of the sample, used to identify the class to which the sample belongs, such as “high load”, “low load”, “normal operation”, “abnormal operation” and the like; the importance of different features is quantified by a feature weight vector w, and the weighted distance between any two samples x i and x j is calculated, and the calculation formula of the weighted distance is: wherein D w (x i ,x j ) is the weighted distance between the samples x i and x j , w k is the weight of the kth feature, K is the total number of features, x ik is the kth feature value of the sample x i , and x jk is the kth feature value of the sample x j ; if the weight w k of a certain feature is larger, it means that the feature accounts for a higher proportion in the distance calculation, i.e., the difference of the feature has a more significant influence on the “distance” between samples; if w k is close to 0, the influence of the difference of the feature on the distance can be ignored, and the feature is regarded as a redundant feature.

[0071] In an implementation, the target sample is calculated according to the weighted distance to obtain a target probability, and the probability (target probability) that any sample x j is selected as the reference point of the target sample x i is calculated. wherein P ij is the probability that x j is the reference point of x i , (k(z)=exp(-z / τ) is a kernel function (decreasing function of the weighted distance D w ), k(D w (x i ,x j )) is the kernel function value of x i and x j , is the sum of the kernel function values of all other samples and x i , the correlation between samples is measured by the probability distribution, and the classification accuracy is improved; when two samples x i and xj weighted distance D of w The smaller the distance D, that is, the more similar the core features, the greater the value of k(D w ), and the higher the probability P ij (same-class samples are more likely to be reference points); when D w is larger, the core features are more different, k(D w ) is smaller, and P ij is lower, and the probability of selecting different-class samples as reference points is low.

[0072] In an implementation, the classification probability is calculated according to the target probability, that is, the probability of correctly classifying the sample x i (classification probability) is calculated, where y i =1 when x j and x ij belong to the same class, otherwise y ij =0; the greater the value of p i , the stronger the relevance of the features of x i to the same-class samples, and the more reliable the classification; by optimizing the feature vector, the features with high weights w k are retained, that is, the features that have a greater impact on p i , and the redundant features with weights close to 0 are removed, to realize feature dimension reduction.

[0073] In an embodiment, the target model is obtained by updating the preset model according to the target database, including:

[0074] determining the parameter combination of the preset model, initializing the population to obtain the target population; the parameter combination includes: a penalty coefficient, a function parameter, and a parameter search range;

[0075] calculating the fitness of each chromosome in the target population, and determining the optimal chromosome according to the numerical value of the fitness;

[0076] updating the target population iteratively according to the optimal chromosome to obtain an individual optimal set, and determining a search region set according to the individual optimal set; the individual optimal set contains the optimal individual of each chromosome in the iteration process; the search region set contains multiple search regions; the optimal individual and the search region correspond one-to-one;

[0077] iteratively updating according to each search region, and outputting the optimal parameter combination when a preset condition is met, and updating the preset model to obtain the target model according to the optimal parameter combination.

[0078] In an implementation, the preset model can be a support vector machine; the parameter combination includes: 1. a penalty coefficient C: controlling the overfitting degree of the SVM model, the greater the value, the heavier the penalty for incorrect classification; 2. a kernel function parameter γ: controlling the nonlinearity degree of the model, affecting the mapping complexity of the sample in the high-dimensional space; 3. setting the search range of the parameter, for example: C ∈ [0.01, 100], γ ∈ [0.001, 10], as the search space boundary of the particle swarm.

[0079] In an implementation, the individual optimal set includes the optimal individual of each chromosome in the iteration process, the target population includes a plurality of chromosomes, and each iteration records the historical optimal position of each particle, that is, the parameter combination with the highest fitness so far; the optimal chromosome: the position with the highest fitness is selected from the individual optimal of all particles as the current optimal solution of the entire particle swarm; the fitness formula is the prior art, for example:

[0080]

[0081] wherein w is a weight vector of the target model, ||w|| is a modulus square of the weight vector, C is a penalty coefficient, ε is a relaxation variable, and n is the number of training samples. 2 i

[0082] In an implementation, in each iteration process, the particle with the highest fitness is selected as the optimal chromosome according to the fitness, and the optimal chromosome is reselected. Each particle randomly explores the search space in the early stage of iteration, and the possible combination of the preset model parameters is evaluated by the fitness function. The individual optimal records the “best solution so far” found by the particle, so that even if the subsequent particle deviates from the region due to global guidance, the historical high-quality solution can still be traced back to the individual optimal to prevent loss of high-quality information; the optimal chromosome guides the particle to concentrate in the optimal region recognized by the group (global exploration), and the individual optimal promotes the particle to search in the vicinity of the historical high-quality region (local development).

[0083] In an implementation, the preset condition is that the number of iterations reaches a preset value (for example, 100 times), and the fitness of the optimal chromosome does not obviously improve for a plurality of rounds (for example, 10 rounds) in succession; until the preset condition is met, the optimal parameter combination is output, that is, the parameter combination corresponding to the optimal chromosome at this time is updated as the optimal parameter combination of the preset model.

[0084] In an embodiment, inputting the power project task into the target model to obtain the final project scheme includes:

[0085] ​​​According to the power project task, a task sequence is determined, a first task is input into the target model to obtain a first scheme; the task sequence includes a first task, a second task and a third task; the first task is global variable optimization, the second task is discrete variable optimization, and the third task is continuous variable optimization.

[0086] According to the first scheme, a first optimization target is determined, the second task and the first optimization target are input into the target model to obtain a second scheme;

[0087] According to the second scheme, a second optimization target is determined, the third task and the second optimization target are input into the target model to obtain a third scheme, and the third scheme is taken as a final project scheme.

[0088] In an implementation manner, according to the power project task, a task sequence is determined, the first task is pre-optimization, the second task is coarse optimization, and the third task is fine optimization; the pre-optimization is that in a pre-optimization stage, variable types (discrete or continuous) are not distinguished, all core variables (circuit breakers, on-load tap changers and distributed power sources) affecting the migration planning are globally optimized, and the purpose is to quickly lock the approximate scheduling range of each device (for example, how much reactive power the distributed power source needs to output, and how many circuit breakers need to be put into operation), but the operation frequency limit of the discrete device (for example, the circuit breaker) is not considered.

[0089] In an implementation manner, the coarse optimization focuses on discrete variables, the distributed power source reactive power output in the pre-optimization is fixed, and the scheduling plan of the circuit breaker and the on-load tap changer is separately corrected, for example, the pre-optimization may require that the circuit breaker switches 5 times within 1 hour, but only 3 times are actually allowed, the coarse optimization adjusts the switching time or frequency to ensure that the operation constraint is met, and the deviation from the pre-optimization result is minimized.

[0090] In an implementation manner, the fine optimization focuses on continuous variables, the scheduling plan of the circuit breaker and the on-load tap changer after the coarse optimization is fixed, only the reactive power output (continuous control variable) of the distributed power generation is adjusted, the system network loss is minimized as the target, the deviation after the adjustment of the discrete variable is compensated, for example, the voltage deviation caused by the operation frequency limit of the on-load tap changer, and the final distributed power generation reactive power output plan is output, and the complete scheduling scheme is obtained in combination with the discrete variable plan of the coarse optimization.

[0091] In an implementation manner, the pre-optimization only retains basic operation constraints of the device (for example, the reactive power output of the distributed power source does not exceed the rated value, and the voltage does not exceed the limit), and the global optimal space is quickly explored; the coarse optimization adds the operation frequency constraint of the discrete device, and solves the problem that the pre-optimization is theoretically feasible but practically inoperable; and the fine optimization adds the boundary of the distributed power source reactive power output and the node voltage constraint, and ensures the accuracy of the continuous variable adjustment.

[0092] In an embodiment, if the final project scheme completes the power project task, further comprising:

[0093] According to the simulation of the final project scheme, the running state parameters are obtained, and if the running state parameters meet the preset constraint condition, it is determined that the final project scheme completes the power project task;

[0094] Otherwise, it is determined that the final project scheme does not complete the power project task, and the power project task is input into the target model for training until the preset training times are reached, and the training is stopped.

[0095] In an implementation manner, the running state parameters are, for example, voltage and power: if all node voltages are within the range and the line power is not overloaded, the system stability meets the expectation; if the voltage deviates or is overloaded, the distributed power reactive compensation needs to be increased in the fine optimization stage; device operation: if the operation times of the circuit breaker and the on-load tap changer do not exceed the preset upper limit (for example, the on-load tap changer is switched less than 3 times per day), the device loss constraint is met; if the upper limit is exceeded, the operation time limit needs to be tightened in the coarse optimization stage; load rate: if the load rate in more than 90% of the time period in the simulation output load rate curve is within 70%-90%, the expectation is met; if the range is exceeded (for example, continuously >90% or <70%), the load distribution strategy needs to be optimized.

[0096] In an implementation manner, when the constraint condition is met at the first simulation, the validity of the scheme is directly confirmed, and potential risks in actual execution are avoided; if the constraint is not met, the target model is trained in a targeted manner through the task data, and the adaptation ability of the model to the task scene can be continuously optimized; by limiting the training times, the resource waste caused by overtraining is prevented while the optimization effect of the model is ensured, and the flexibility of the training and the scheme quality are improved.

[0097] Based on the same inventive concept, the embodiments of the present application also provide a power project budget estimation system. Referring to Figure 2 , Figure 2 A framework diagram of a power project budget estimation system provided by the embodiments of the present application comprises:

[0098] The data preprocessing module is configured to obtain historical data of a target region, preprocess the historical data to obtain target historical data, and obtain the target historical data.

[0099] The database construction module is configured to extract features from the target historical data to obtain feature data, obtain spatial data of the target region, and construct a target database according to the feature data and the spatial data.

[0100] The model updating module is configured to update a preset model according to the target database to obtain a target model, obtain a power project task, and input the power project task into the target model to obtain a final project scheme.

[0101] The budget determining module is configured to determine the power project budget according to the final project scheme if the final project scheme completes the power project task.

[0102] The power project budget estimation system provided by the embodiment of the application can obtain target area load data, pre-process and extract features, construct a target database in combination with spatial data, update and optimize a preset model, input a power project task into the model to obtain a final project scheme, and determine a power project budget, thereby reducing data complexity, improving calculation efficiency and accuracy.

[0103] In one embodiment, the database construction module comprises:

[0104] The sample calculation module is configured to determine a sample set according to target historical data and calculate the weighted distance of each sample in the sample set; the sample set comprises a plurality of samples; each sample comprises a feature vector and a category label;

[0105] The probability calculation module is configured to calculate a target sample according to the weighted distance to obtain a target probability and calculate a classification probability according to the target probability; the target sample is any one of the samples in the sample set;

[0106] The probability classification module is configured to retain the target sample if the target probability is greater than a classification threshold, or to eliminate the target sample.

[0107] In one embodiment, the model updating module comprises:

[0108] The initialization module is configured to determine a parameter combination of the preset model, initialize a population to obtain a target population; the parameter combination comprises a penalty coefficient, a function parameter and a parameter search range;

[0109] The fitness calculation module is configured to calculate the fitness of each chromosome in the target population and determine an optimal chromosome according to the numerical value of the fitness;

[0110] The local optimization module is configured to iteratively update the target population according to the optimal chromosome to obtain an individual optimal set, determine a search region set according to the individual optimal set; the individual optimal set comprises the optimal individual of each chromosome in the iteration process; the search region set comprises a plurality of search regions; the optimal individual and the search region are in one-to-one correspondence;

[0111] The global optimization module is configured to iteratively update each search region until a preset condition is met to output an optimal parameter combination, and update the preset model according to the optimal parameter combination to obtain a target model.

[0112] In one embodiment, the model updating module further comprises:

[0113] The first execution module is configured to determine a task sequence according to the power project task, input a first task into the target model, and obtain a first scheme; the task sequence includes the first task, a second task, and a third task; the first task is global variable optimization, the second task is discrete variable optimization, and the third task is continuous variable optimization;

[0114] The second execution module is configured to determine a first optimization target according to the first scheme, input the second task and the first optimization target into the target model, and obtain a second scheme;

[0115] The third execution module is configured to determine a second optimization target according to the second scheme, input the third task and the second optimization target into the target model, obtain a third scheme, and take the third scheme as a final project scheme.

[0116] In an embodiment, the budget determination module further includes:

[0117] The first judgment module is configured to perform simulation according to the final project scheme to obtain a running state parameter, and determine that the final project scheme completes the power project task if the running state parameter meets a preset constraint condition.

[0118] The second judgment module is configured to determine that the final project scheme does not complete the power project task if the running state parameter does not meet the preset constraint condition, and input the power project task into the target model for training until a preset training number is reached to stop the training.

[0119] The above describes one embodiment of the present application in detail, but the content is only a preferred embodiment of the present application, and cannot be considered as limiting the implementation range of the present application. Any equivalent changes and improvements made according to the scope of the present application should still belong to the scope of the claims of the present application.

Claims

1. A method for estimating a budget of a power project, characterized by, The method comprises: acquiring historical data of a target region, and preprocessing the historical data to obtain target historical data; extracting features from the target historical data to obtain feature data, acquiring spatial data of the target region, and constructing a target database according to the feature data and the spatial data; updating a preset model according to the target database to obtain a target model, acquiring a power project task, inputting the power project task into the target model to obtain a final project scheme; if the final project scheme completes the power project task, determining a power project budget according to the final project scheme; inputting the power project task into the target model to obtain a final project scheme comprises: determining a task sequence according to the power project task, inputting a first task into the target model to obtain a first scheme; the task sequence comprises a first task, a second task and a third task; the first task is global variable optimization, the second task is discrete variable optimization, and the third task is continuous variable optimization; determining a first optimization target according to the first scheme, inputting the second task and the first optimization target into the target model to obtain a second scheme; determining a second optimization target according to the second scheme, inputting the third task and the second optimization target into the target model to obtain a third scheme, and taking the third scheme as the final project scheme; if the final project scheme completes the power project task, further comprising: performing simulation according to the final project scheme to obtain a running state parameter, and if the running state parameter meets a preset constraint condition, determining that the final project scheme completes the power project task; otherwise, determining that the final project scheme does not complete the power project task, and inputting the power project task into the target model for training until a preset training number is reached to stop training.

2. The method of claim 1, wherein, extracting features from the target historical data to obtain feature data comprises: determining a sample set according to the target historical data, and calculating weighted distances of samples in the sample set; the sample set comprises a plurality of samples; any sample is composed of a feature vector and a category label; calculating a target probability of a target sample according to the weighted distances, and calculating a classification probability according to the target probability; the target sample is any one of the samples in the sample set; if the target probability is greater than a classification threshold, the target sample is retained, otherwise, the target sample is removed.

3. The method of claim 1, wherein, updating a preset model according to the target database to obtain a target model comprises: determining a parameter combination of the preset model, initializing a population to obtain a target population; the parameter combination comprises a penalty coefficient, a function parameter and a parameter search range; calculating the fitness of each chromosome in the target population, and determining an optimal chromosome according to the numerical value of the fitness. According to the optimal chromosome, the target population is iteratively updated to obtain an individual optimal set, and a search region set is determined according to the individual optimal set; the individual optimal set contains the optimal individual of each chromosome in the iteration process; the search region set contains a plurality of search regions; the optimal individual and the search region are one-to-one corresponding; According to each search region, iteratively update until the preset condition is met, and output the optimal parameter combination; according to the optimal parameter combination, update the preset model to obtain the target model.

4. A power project budget estimation system characterized by, The system comprises: A data preprocessing module is configured to obtain historical data of a target region, and preprocess the historical data to obtain target historical data; A database construction module is configured to extract features from the target historical data to obtain feature data, obtain spatial data of the target region, and construct a target database according to the feature data and the spatial data; A model updating module is configured to update a preset model according to the target database to obtain a target model, obtain a power project task, and input the power project task into the target model to obtain a final project scheme; A budget determination module is configured to determine a power project budget according to the final project scheme if the final project scheme completes the power project task; The model updating module further comprises: A first execution module is configured to determine a task sequence according to the power project task, and input a first task into the target model to obtain a first scheme; the task sequence comprises a first task, a second task and a third task; the first task is global variable optimization, the second task is discrete variable optimization, and the third task is continuous variable optimization; A second execution module is configured to determine a first optimization target according to the first scheme, and input the second task and the first optimization target into the target model to obtain a second scheme; A third execution module is configured to determine a second optimization target according to the second scheme, input the third task and the second optimization target into the target model to obtain a third scheme, and take the third scheme as the final project scheme; The budget determination module further comprises: A first judgment module is configured to obtain a running state parameter by simulating the final project scheme, and determine that the final project scheme completes the power project task if the running state parameter meets a preset constraint condition; A second judgment module is configured to determine that the final project scheme does not complete the power project task if not, and input the power project task into the target model for training until a preset training number is reached to stop training.

5. The power project budget estimation system according to claim 4, wherein, The database construction module comprises: A sample calculation module is configured to determine a sample set according to the target historical data, and calculate the weighted distance of each sample in the sample set; the sample set contains a plurality of samples; any sample is composed of a feature vector and a class label; A probability calculation module is configured to calculate a target probability of a target sample according to the weighted distance, and calculate a classification probability according to the target probability; the target sample is any one in the sample set. A probability classification module is configured to retain the target sample if the target probability is greater than a classification threshold, and to reject the target sample otherwise.

6. The power project budget estimation system according to claim 4, wherein, The model updating module comprises: An initialization module is configured to determine a parameter combination of the preset model, initialize a population to obtain a target population, and determine a penalty coefficient, a function parameter, and a parameter search range. An adaptability calculation module is configured to calculate the adaptability of each chromosome in the target population, and determine an optimal chromosome according to the numerical value of the adaptability. A local optimization module is configured to iteratively update the target population according to the optimal chromosome to obtain an individual optimal set, determine a search region set according to the individual optimal set, and one-to-one correspond the optimal individual in the individual optimal set to the search region in the search region set. A global optimization module is configured to iteratively update each search region until a preset condition is met, output an optimal parameter combination, and update the preset model according to the optimal parameter combination to obtain a target model.

Images