A costing method, medium and system based on spare parts allocation for power grid engineering

Abstract

The present invention provides a cost estimation method, medium and system according to spare parts allocation of power grid projects, belonging to the technical field of power grid projects, and comprising: obtaining meteorological history records, electrical parameter history records and equipment replacement history records of the area where the power grid projects are located for many consecutive years, and establishing a replacement time list for each type of equipment; according to the replacement time list of each type of equipment, finding the replacement correspondence between the equipment and establishing an equipment replacement binding relationship set; classifying the equipment according to the equipment replacement binding relationship set; establishing a life model of the active replacement equipment and meteorological data and electrical parameter data; calculating the next replacement time of the corresponding active replacement equipment according to the most recent replacement time of each active replacement equipment of the power grid project and the life model; generating a spare parts allocation plan; and calculating the overall cost of spare parts procurement, storage and transportation according to the spare parts allocation plan.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid engineering, and in particular, relates to a cost estimation method, a medium and a system according to the allocation of spare parts for power grid engineering. Background Art

[0002] The power system is an indispensable infrastructure and an important pillar industry in modern society. To ensure the continuous, reliable and safe operation of power grid equipment, power grid companies need to formulate scientific equipment maintenance plans and reasonably allocate spare parts. Existing power grid companies mainly rely on manual experience to allocate spare parts, which has many problems: 1. Manual experience cannot fully consider complex influencing factors. The replacement cycle of power grid equipment is highly related to climate, load, operating parameters, etc. It is difficult to establish an accurate replacement cycle model based on experience alone. 2. Manual planning is inefficient. There are many types of spare parts and a large amount of equipment. Manually formulating allocation plans is time-consuming and labor-intensive. It is impossible to respond quickly to changes in allocation requirements. 3. It is difficult to find the inherent connection between equipment replacement. There are a large number of passive replacement equipment, and only focusing on active equipment cannot form a matching allocation plan. 4. It is difficult to accurately predict the life of equipment. Equipment aging is a complex time series process with many uncertain random factors. Manual estimation is prone to errors. 5. It is impossible to optimize loss costs. There are a lot of losses in procurement, warehousing, and transportation, and the empirical method cannot achieve the minimum cost.

[0003] In summary, the existing cost estimation process for spare parts allocation that relies on manual experience has technical problems such as low efficiency and large errors. How to use scientific optimization methods to improve spare parts allocation technology is an important issue facing power grid companies. Summary of the invention

[0004] In view of this, the present invention provides a cost estimation method, medium and system based on spare parts allocation for power grid projects, which can solve the technical problems of low efficiency and large errors in the existing cost estimation process for spare parts allocation relying on manual experience.

[0005] The present invention is achieved in that:

[0006] A first aspect of the present invention provides a cost estimation method based on the allocation of spare parts for power grid projects, which includes the following steps:

[0007] S10, obtaining the meteorological history records, electrical parameter history records, and equipment replacement history records of the area where the power grid project is located for many consecutive years, and establishing a replacement time list for each type of equipment;

[0008] S20, according to the replacement time list of each device, find the replacement correspondence between devices and establish a device replacement binding relationship set;

[0009] S30, classifying the devices according to the device replacement binding relationship set, including active replacement devices and passive replacement devices;

[0010] S40, establishing a life model of the active replacement equipment, meteorological data, and electrical parameter data;

[0011] S50, calculating the next replacement time of the corresponding active replacement device according to the most recent replacement time of each active replacement device of the power grid project and the life model;

[0012] S60, generating a spare parts allocation plan according to the obtained next replacement time of the active device and its corresponding passive replacement device;

[0013] S70. Calculate the overall cost of spare parts procurement, storage, and transportation according to the spare parts allocation plan.

[0014] On the basis of the above technical solution, a cost estimation method according to the allocation of spare parts for power grid projects of the present invention can also be improved as follows:

[0015] Among them, the steps of establishing the equipment replacement binding relationship set are specifically: based on the equipment replacement time list, using the Apriori algorithm to calculate the association rules of replacement time between equipment categories, using Support and Confidence as rule interestingness metrics; using the Lift indicator and rule coverage algorithm Pruning to screen non-redundant association rules; determining the strong correlation pairs of equipment replacement time according to the rule set to form equipment replacement binding relationships.

[0016] Furthermore, the step of classifying the equipment is specifically: using the K-means clustering algorithm to perform unsupervised classification learning on the equipment categories; selecting the optimal number of classifications based on the intra-class sample clustering Criterion evaluation index; repeating the experiment multiple times to determine the optimal equipment classification result, and obtaining the active and passive replacement equipment classification.

[0017] Furthermore, the method for establishing the life model is: collecting historical replacement data and related environmental characteristic parameters of actively replaced equipment; using a random forest regression model to model the remaining life of actively replaced equipment; adjusting parameters to determine the optimal structure of the random forest to prevent overfitting; establishing an LSTM neural network for sequence prediction, and integrating the random forest regression model and the LSTM neural network model to evaluate equipment life.

[0018] Furthermore, the step of calculating the next replacement time of the corresponding actively replaced device is specifically: for each actively replaced device, input its most recent replacement record and current environmental data; use the life assessment model to predict the current remaining life of the actively replaced device; and calculate the next replacement time of the actively replaced device in combination with the last replacement time.

[0019] Furthermore, the step of generating a spare parts allocation plan is specifically: based on the predicted replacement time of active replacement, the replacement time of the passive replacement equipment is expanded to obtain the replacement time of the passive replacement equipment; according to the replacement time of the active replacement equipment and the passive replacement equipment, a spare parts procurement plan is generated; and the procurement plan is integrated to obtain the final spare parts allocation plan.

[0020] Furthermore, the steps for calculating the overall cost of spare parts procurement, storage and transportation are specifically: counting the procurement quantity of each equipment in the deployment plan; inquiring the unit price to calculate the procurement cost; calculating the storage fee according to the storage time; estimating the transportation cost according to the transportation route; and summing up various expenses to obtain the overall spare parts deployment cost.

[0021] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned cost estimation method based on the allocation of spare parts for power grid projects.

[0022] A third aspect of the present invention provides a system for calculating the cost of spare parts deployment for power grid projects, which includes the above-mentioned computer-readable storage medium.

[0023] Compared with the prior art, the cost estimation method, medium and system provided by the present invention for the allocation of spare parts for power grid projects have the following beneficial effects: the technical solution of the present invention can establish an accurate time series prediction model by collecting and analyzing a large amount of equipment operation and environmental data, realize the estimation of the remaining life of power grid equipment, and overcome the limitations of pure manual experience prediction due to random factors. At the same time, the application of association rule learning and other technologies can discover the connection between equipment replacement behaviors, establish a driving mechanism for active equipment replacement on driven equipment, and complete the formulation of allocation plans step by step. It solves the technical problems of low efficiency and large errors in the existing cost estimation process for spare parts allocation that relies on manual experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0025] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0027] like Figure 1 As shown, it is a flowchart of a method for cost estimation based on spare parts allocation for power grid projects according to a first aspect of the present invention, and the present invention comprises the following steps:

[0028] S10, obtaining the meteorological history records, electrical parameter history records, and equipment replacement history records of the area where the power grid project is located for many consecutive years, and establishing a replacement time list for each type of equipment;

[0029] S20, according to the replacement time list of each device, find the replacement correspondence between devices and establish a device replacement binding relationship set;

[0030] S30, classifying the devices according to the device replacement binding relationship set, including active replacement devices and passive replacement devices;

[0031] S40, establishing a life model for active replacement equipment and meteorological data and electrical parameter data;

[0032] S50, calculating the next replacement time of the corresponding active replacement device according to the most recent replacement time of each active replacement device of the power grid project, as well as the life model and the device replacement binding relationship set;

[0033] S60, generating a spare parts allocation plan according to the obtained next replacement time of the active device and its corresponding passive replacement device; the spare parts include: the active replacement device and / or the passive replacement device, the passive replacement device is a separate device or a component in the active replacement device;

[0034] S70. Calculate the overall cost of spare parts procurement, storage, and / or transportation according to the spare parts allocation plan.

[0035] The specific implementation methods of the above steps are described in detail below:

[0036] For step S10, its purpose is to obtain necessary historical data and provide a data set for the subsequent establishment of an equipment replacement time model. Specific implementation methods may include: collecting meteorological data for many years in the area where the power grid project is located, such as temperature, humidity, wind speed, etc.; collecting historical records of electrical parameters of the power grid project, such as voltage, current, power, etc.; collecting replacement and maintenance records of power grid project equipment, and establishing a replacement time list for each type of equipment. The data that need to be collected here include meteorological data, electrical parameter data, and equipment maintenance data. These data record time series, and time series analysis methods can be used to mine data features. And these time series data can be stored in a database for subsequent query and modeling.

[0037] For step S20, the purpose is to find the internal correspondence between equipment replacements. The specific implementation method can adopt the following technical means:

[0038] 1) Collect the equipment replacement time list dataset, which records the historical replacement and maintenance time of various equipment. Preprocess the dataset, including cleaning null values ​​and duplicate data, and converting text descriptions into numerical codes.

[0039] 2) Construct a 0-1 matrix of equipment replacement time, where rows represent equipment categories and columns represent time periods. If the equipment has a replacement record in the time period, the corresponding position is 1, otherwise it is 0. This can intuitively reflect the corresponding relationship between the replacement time of different equipment categories.

[0040] 3) Calculate the correlation between the replacement time of different equipment types. The Apriori algorithm can be used to find the frequent replacement correspondence, that is, the association rules whose support and confidence exceed the threshold. This means that the equipment category pairs that meet the association rule are likely to have a replacement time binding relationship.

[0041] 4) The rules obtained in the previous step may contain redundant or similar rules. The Lift indicator Divide Prune algorithm can be used to filter and traverse the rules to obtain a non-redundant rule set. The Lift indicator reflects the strength of the rule, so that strongly associated equipment replacement rules can be filtered out.

[0042] 5) Visualize the rule set and draw an association rule network diagram, where each node is a type of equipment and the edge reflects the strength of its replacement time association. Network analysis algorithms can be used to analyze the degree centrality of nodes, that is, which equipment has the greatest influence.

[0043] 6) After the above steps, a strongly associated device category pair can be obtained as a device replacement binding relationship, forming a device replacement binding relationship set, which provides a basis for subsequent device classification.

[0044] The purpose of step S30 is to classify the devices. The specific implementation method may adopt the following technical means:

[0045] 1) Construct a device replacement time character matrix and encode the replacement time series of the device category with 0-1. This is the input data for unsupervised learning.

[0046] 2) Perform principal component analysis (PCA) on the sample matrix to reduce feature dimensionality, reduce feature redundancy, and improve the efficiency of subsequent cluster analysis. You can draw a Scree Plot of the sample principal component scores to determine the number of principal components.

[0047] 3) Use K-means algorithm to cluster samples. First, we need to determine the number of clusters K. We can select an effective K value by observing the PCA results. Randomly initialize K cluster centers and iterate and optimize to cluster samples of the same type and separate samples of different types.

[0048] 4) Use the elbow rule to find the optimal number of clusters K for K-means. Traverse different K values, plot a curve of K and the sum of squared deviations within the class, and find the K value corresponding to the elbow position.

[0049] 5) Repeat the K-means clustering experiment multiple times and test different initial centers to prevent falling into the local optimal solution. Record the clustering quality of each experiment and select the best classification result, that is, the clustering label of the device category.

[0050] 6) According to the clustering label results and combined with the equipment replacement binding relationship, determine which categories are active replacement equipment and which categories are passive replacement equipment, and obtain the final equipment classification results.

[0051] Regarding step S40, its purpose is to establish a life model for active replacement of equipment. The specific implementation method can adopt the following technical means:

[0052] 1) Collect historical replacement records of actively replaced equipment as training data, including feature information such as replacement time, location, and equipment working parameters.

[0053] 2) Collect natural environment data, such as temperature, humidity, corrosive gases, etc., as characteristic variables that affect equipment life.

[0054] 3) Preprocess the data, including outlier processing and normalization, to improve the quality of model training. Use PCA and other methods to reduce the dimension of features to prevent overfitting.

[0055] 4) Build a regression model based on random forest ensemble learning. Randomly sample features and training data sets to generate multiple decision trees, and combine the prediction results of all decision trees to obtain the random forest model output.

[0056] 5) Adjust the parameters of random forest, including the number of decision trees, maximum depth, feature sample ratio, etc., and select appropriate parameters through cross-validation to prevent overfitting.

[0057] 6) Use recursive neural network models such as LSTM to establish an end-to-end sequence prediction framework. Input the device history record sequence and model the device status.

[0058] 7) Adjust the LSTM model structure, including the number of LSTM layers, number of nodes, and dropout, and optimize the model through Early Stopping and other methods to prevent overfitting.

[0059] 8) Finally, the random forest model and LSTM model are integrated to build an efficient equipment remaining life prediction framework to provide a basis for active equipment life assessment.

[0060] 9) Continue to optimize the model and use the data from new equipment replacements for incremental learning to improve model robustness and prediction accuracy.

[0061] For step S50, its purpose is to calculate the next replacement time of the active device. The specific implementation can be:

[0062] 1) For each active device, query the record of its last replacement and obtain the replacement date. This is the input data.

[0063] 2) Based on the geographical location of the equipment, query relevant historical meteorological data and power grid electrical parameters as model input.

[0064] 3) The above data is fed into the random forest model trained in step S40 to obtain the current remaining life prediction result of the equipment.

[0065] 4) Add the predicted remaining life value to the known last replacement date to calculate the next predicted replacement time for the equipment.

[0066] 5) Repeat the above operation for each active device to obtain the next predicted replacement time of all active devices and form a replacement schedule.

[0067] 6) If there is insufficient historical replacement data for the device and it is not possible to directly model and predict, the K nearest neighbor algorithm can be used to make a comprehensive prediction based on the replacement records of the K closest similar devices.

[0068] 7) In order to improve the reliability of the prediction, the prediction operation can be repeated multiple times, multiple versions of the prediction schedule can be integrated, and the median or average can be used as the final prediction value to reduce the error rate.

[0069] 8) Visualize the prediction results and mark the equipment that will be replaced sooner or is time-sensitive, so as to prioritize the maintenance plan for these equipment.

[0070] The purpose of step S60 is to generate a spare parts allocation plan. The specific implementation can be:

[0071] 1) The active equipment replacement schedule predicted in step S50 is used as input data.

[0072] 2) According to the device replacement binding relationship table, the replacement time of the active device is expanded to obtain the replacement time of the slave device.

[0073] 3) Arrange all equipment in order of replacement time to form a preliminary equipment replacement schedule.

[0074] 4) Optimize the constraints of the schedule. Determine the constraints such as spare parts supplier response time and spare parts transportation time. Use optimization algorithms such as linear programming to optimize the replacement schedule while satisfying the constraints.

[0075] 5) Based on the optimized equipment replacement schedule, a spare parts procurement plan is gradually generated according to the parameters such as the type, quantity, and procurement cycle of spare parts required for each equipment.

[0076] 6) Integrate the spare parts procurement plan according to the time sequence and delivery batches to generate a spare parts allocation plan, including the allocation time, spare parts list, allocation path, etc. of each delivery batch.

[0077] 7) Optionally, a vehicle routing optimization algorithm is used to determine the best transportation route for each delivery batch to minimize transportation distance and cost.

[0078] 8) Output a complete spare parts allocation plan to provide spare parts guarantee for equipment maintenance. At the same time, generate a visual display of the plan and mark the key procurement and allocation time nodes.

[0079] The purpose of step S70 is to calculate the overall cost. The specific implementation can be:

[0080] 1) According to the spare parts allocation plan, count the purchase quantity of each spare part.

[0081] 2) Check the unit price of spare parts and calculate the spare parts purchase cost based on the quantity.

[0082] 3) The storage time of spare parts can be extracted from the plan table, and the storage cost can be calculated based on the storage cost rate.

[0083] 4) Extract the transportation batch and route information from the dispatch plan, and estimate the transportation cost based on the transportation cost calculation formula.

[0084] 5) Sum the purchasing cost, storage cost, and transportation cost generated above to obtain the overall cost of spare parts allocation.

[0085] 6) Optionally, mark the main cost components to assist in cost verification. Also provide sensitivity analysis to assess the impact of changes in key parameters on overall costs.

[0086] 7) Optionally, generate an intuitive cost composition chart to visualize the results. Pay attention to whether the budget constraints are met and the optimization direction of cost control.

[0087] 8) Optionally, after implementation according to the plan, continue to track the actual expenses of procurement and deployment, evaluate the deviation between actual cost and predicted cost, and continuously optimize the cost calculation model.

[0088] The following is a specific embodiment of the present invention. The variables used in this embodiment are explained as follows:

[0089] Step S10

[0090] The purpose of this step is to collect the historical data required for modeling. First, collect the meteorological parameters of the power grid area for N consecutive years, using the matrix X∈R N*M Indicates, where M is the number of meteorological factors, such as temperature, humidity, etc.

[0091] At the same time, the electrical parameters of the power grid in the past N years are collected, and the matrix Y∈R N*K Indicates, where K is the number of electrical parameters, such as voltage, current, etc.

[0092] In addition, the replacement and maintenance records of various types of power grid equipment are collected. For each device i, the vector T i ∈R N Represents its N-year replacement time series.

[0093] Collect relevant data according to the above method, and use relational database to store and manage these data sets to provide data support for subsequent modeling and analysis.

[0094] Step S20

[0095] This step finds the intrinsic correspondence between equipment replacements by mining the association rules of equipment replacement time.

[0096] First, the frequent item set F of replacement time association rules is constructed, which is composed of frequent combinations of equipment replacement time. Then, the Apriori algorithm is used to calculate the association rules between replacement time item sets.

[0097] For rule R: A→B, its support is defined as support(R)=P(A∪B), which is the proportion of records containing both change time A and B in the total records. The confidence of the rule is defined as confidence(R)=P(B|A), which is the probability that a record containing change time A also contains B.

[0098] Finally, according to the thresholds of minimum support minsup and minimum confidence minconf, the association rules that meet the two conditions are screened as strong association rules of equipment replacement time, and then the equipment replacement binding relationship is determined.

[0099] Step S30

[0100] This step classifies the equipment through clustering algorithm. First, construct the replacement time feature matrix Z∈R of the equipment category M*N , where M is the number of devices and N is the length of the time series.

[0101] Then, the PCA dimensionality reduction method is used to obtain the reduced feature matrix Z′. The number of clusters K is set, and the K-means algorithm is used for clustering:

[0102]

[0103] Where S i represents the i-th cluster sample set, u i is the class center of the i-th cluster. The optimization goal is to aggregate samples of the same type and separate samples of different types.

[0104] Repeat K-means multiple times and select the optimal number of clusters K through the elbow rule * , and obtain the final clustering label of the device category. Then, according to the clustering results, the active and passive replacement devices can be distinguished.

[0105] Step S40

[0106] This step establishes a life assessment model for active replacement equipment. The historical replacement record data of active equipment is used as the training set D = (x 1 ,y 1 ), (x 2 ,y 2 ),...,(x N ,y N ), where x i Indicates sample characteristics, such as environmental parameters, electrical parameters, etc., y i Indicates Devices lifetime.

[0107] Build a random forest regression model that integrates multiple decision trees to train samples:

[0108]

[0109] Among them, h k (x) represents the prediction of the kth decision tree, α k The weight is used to adjust the number of decision trees, maximum depth and other parameters to optimize the model.

[0110] At the same time, an LSTM neural network is established to predict the equipment life sequence. Finally, the two models are integrated to predict equipment failure.

[0111] Step S50

[0112] This step predicts the next replacement time for each active device.

[0113] For device j, it is known that its last replacement time is t j , collect its current environment characteristics x j The information is input into the evaluation model established by S40 to obtain the current remaining life prediction of the equipment. Then the predicted replacement time of equipment j is:

[0114]

[0115] Repeat this operation to obtain the predicted replacement schedule for all active devices. Where n is the number of active devices.

[0116] Step S60

[0117] This step develops a spare parts deployment plan. The replacement schedule will be predicted As input data, connect the equipment replacement association relationship and expand to obtain the replacement schedule T′ of the driven equipment.

[0118] Then, all equipment are arranged in chronological order to generate a preliminary deployment plan P. The linear programming model is used to optimize the plan P under constraints such as transportation time and cost to obtain the optimal deployment plan P*.

[0119] Step S70

[0120] This step calculates the total cost of the deployment plan. First, extract the purchase list from the optimal plan P*, count the number of each spare part, and calculate the purchase cost C based on the unit price. 1 :

[0121]

[0122] Among them, m is the number of spare parts, q i , p i are the purchase quantity and unit price of the i-th type spare parts respectively.

[0123] Then according to the spare parts storage time ti And storage rate r, calculate storage cost:

[0124]

[0125] Finally, the transportation cost is calculated based on the transportation mileage L and the ton-kilometer freight c:

[0126] C 3 =L·c

[0127] The total cost of the deployment plan is the sum of the following items:

[0128] C total =C 1 +C 2 +C 3

[0129] The following are the specific steps of the embodiment:

[0130] Variable explanation:

[0131] N: The time span of the collected historical data, that is, the length of the time series of the data

[0132] M: Number of meteorological factors

[0133] X∈R N*M : Meteorological data matrix, behavior time, columns for each meteorological factor

[0134] K: Number of electrical parameters

[0135] Y∈R N*K : Electrical parameter data matrix

[0136] T i ∈R N : Historical replacement time series of equipment i

[0137] F: Frequent itemsets of equipment replacement time

[0138] minsup: minimum support threshold

[0139] minconf: minimum confidence threshold

[0140] M: Number of device categories

[0141] N: time series length

[0142] Z∈R M*N : Replacement time characteristic matrix of equipment category

[0143] K: number of clusters

[0144] S i : The i-th cluster

[0145] u i : The class center of the i-th cluster

[0146] D: Active device historical data sample set

[0147] x i : Sample characteristics

[0148] y i : Equipment life value

[0149] n: number of active devices

[0150] t j : The last replacement time of device j

[0151] x j : Current environment characteristics of device j

[0152] Equipment j predicted life

[0153] All active equipment predicted replacement schedule

[0154] T′: Extended driven equipment replacement schedule

[0155] P:Preliminary deployment plan

[0156] P * : The optimal deployment plan after optimization

[0157] m: number of spare parts types

[0158] q i : The quantity of spare parts purchased of type i

[0159] p i : Unit price of the i-th spare part

[0160] t i : Storage time of the i-th spare part

[0161] r: storage rate

[0162] L: Transportation mileage

[0163] c: ton-kilometer freight

[0164] C total : Total cost of deployment plan

[0165] Formula explanation:

[0166] Support (R): reflects the frequency of occurrence of the rule

[0167] Confidence (R): reflects the prediction accuracy of the rule

[0168] K-means objective function: minimize the distance between samples within a cluster

[0169] Random Forest Model: y: predicted value, α k : Decision tree weight

[0170] Proactive equipment prediction replacement time:

[0171] Purchase cost: C 1 =∑q i ·p i : The product of the purchase quantity and the unit price

[0172] Storage cost: C 2 =∑q i ·t i r: product of rate and time

[0173] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned cost estimation method based on the allocation of spare parts for power grid projects.

[0174] A third aspect of the present invention provides a system for calculating the cost of spare parts deployment for power grid projects, which includes the above-mentioned computer-readable storage medium.

[0175] It should be noted that the active replacement equipment and passive replacement equipment mentioned in this solution include both equipment and equipment accessories. Correspondingly, spare parts can also include both equipment and equipment accessories.

Claims

1. A cost estimation method based on the allocation of spare parts for power grid projects, characterized in that: The following steps are involved: S10, obtaining the meteorological history records, electrical parameter history records, and equipment replacement history records of the area where the power grid project is located for many consecutive years, and establishing a replacement time list for each type of equipment; S20, according to the replacement time list of each device, find the replacement correspondence between devices and establish a device replacement binding relationship set; S30, classifying the devices according to the device replacement binding relationship set, including active replacement devices and passive replacement devices; S40, establishing a life model of the active replacement equipment, meteorological data, and electrical parameter data; S50, calculating the next replacement time of the corresponding active replacement device according to the most recent replacement time of each active replacement device of the power grid project, as well as the life model and the device replacement binding relationship set; S60, generating a spare parts allocation plan according to the obtained next replacement time of the active device and its corresponding passive replacement device; the spare parts include: the active replacement device and / or the passive replacement device, the passive replacement device is a separate device or a component in the active replacement device; S70, calculating the overall cost of spare parts procurement, storage and transportation according to the spare parts allocation plan; The steps of establishing the equipment replacement binding relationship set are as follows: based on the equipment replacement time list, the Apriori algorithm is used to calculate the association rules of the replacement time between equipment categories, and the Support degree and Confidence degree are used as the rule interestingness measurement; the Lift indicator and the rule coverage algorithm Pruning are used to screen non-redundant association rules; according to the rule set, the strong association pairs of equipment replacement time are determined to form equipment replacement binding relationships; The step of classifying the equipment specifically comprises: using the K-means clustering algorithm to perform unsupervised classification learning on the equipment categories; selecting the optimal number of classifications according to the intra-class sample clustering Criterion evaluation index; repeating the experiment multiple times to determine the optimal equipment classification result, and obtaining the classification of active and passive replacement equipment; The method for establishing the life model is as follows: collecting historical replacement data of active replacement equipment and related environmental characteristic parameters; using a random forest regression model to model the remaining life of active replacement equipment; adjusting parameters to determine the optimal structure of the random forest to prevent overfitting; establishing an LSTM neural network for sequence prediction, and integrating the random forest regression model and the LSTM neural network model to evaluate equipment life; The step of calculating the next replacement time of the corresponding actively replaced device is specifically: for each actively replaced device, inputting its most recent replacement record and current environmental data; using the life assessment model to predict the current remaining life of the actively replaced device; and calculating the next replacement time of the actively replaced device in combination with the last replacement time; Among them, the steps of generating a spare parts allocation plan are specifically: based on the predicted replacement time of active replacement, the replacement time of the passive replacement equipment is expanded; according to the replacement time of the active replacement equipment and the passive replacement equipment, a spare parts procurement plan is generated; and the procurement plan is integrated to obtain the final spare parts allocation plan.

2. A cost estimation method based on spare parts allocation for power grid projects according to claim 1, characterized in that: The steps of calculating the overall cost of spare parts procurement, storage and transportation are specifically: counting the procurement quantity of each type of equipment in the deployment plan; inquiring the unit price to calculate the procurement cost; calculating the storage fee according to the storage time; estimating the transportation cost according to the transportation route; Add up all the costs to get the overall spare parts allocation cost.

3. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, which, when executed, are used to execute a cost estimation method for allocating spare parts for power grid projects as described in any one of claims 1 to 2.

4. A cost system based on the allocation of spare parts for power grid projects, characterized in that: The computer-readable storage medium of claim 3 is included.

Images