Power grid asset wall risk prediction algorithm and system based on KMP method

Abstract

The invention aims to provide a power grid asset wall risk prediction algorithm and system based on a KMP method. According to the method, the in-wall age limit of each type of equipment of the power grid is measured and calculated, the in-wall scale of assets is predicted, and the asset wall risk is predicted. Firstly, a K-M method is used for calculating a reference scrap probability curve, then Prophet is used for fitting external influence factors, the reference scrap probability curve is corrected, the in-wall age limit of various assets of a power grid is calculated in sequence, finally, the in-wall asset scale and value of a target year are calculated, and the future asset wall scale of the power grid is predicted. And determining the risk level of the asset wall. On the basis of considering the average life of each type of equipment of the power grid, the influence of external influence factors on the asset scrapping probability is also considered, the life distribution models are respectively established for each type of equipment, and the in-wall asset value of each type of equipment is predicted, so that the dynamic and refined prediction of asset wall risks is realized; and the risk prediction precision and timeliness of multiple types and long-life assets of the power grid assets are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of power grid asset wall risk prediction, specifically to a power grid asset wall risk prediction algorithm and system based on the KMP method. Background Technology

[0002] The asset wall model is a key asset management tool. By plotting the lifecycle distribution of a company's total assets on a coordinate system of asset age and asset value, it presents a "wall"-like shape, allowing for the visual prediction of future peak periods of asset aging and retirement. Asset wall risk arises when the amount of aging and retired assets exceeds the capacity for investment and renovation. Asset wall risk prediction provides a basis for risk control and investment strategy formulation by forecasting the lifespan of in-service assets, determining the timeframe for asset inclusion in the asset wall, and quantifying the future scale of the asset wall.

[0003] In recent decades, my country's power grid construction has experienced several investment peaks. Power grid assets put into operation intensively during the same period will eventually age rapidly, creating an asset wall risk. Failure to accurately predict this risk and promptly arrange asset upgrades will threaten the safe and stable operation of the power grid system. Therefore, establishing accurate predictive algorithms and systems is crucial for maintaining the long-term reliability of the power grid and is a practical requirement for the asset management departments of power grid companies.

[0004] Traditional asset wall prediction methods rely solely on simple average lifespan shifting, resulting in coarse-grained asset analysis that fails to account for the varying lifespans of different asset types, making accurate prediction of asset wall size difficult. To improve prediction accuracy, domestic scholars have conducted relevant research: Li Zhiwei et al. proposed a life-cycle management asset wall model for value prediction based on value chain management theory; Liu Yihe et al. constructed a future asset wall value prediction model by combining expected lifespan; and some scholars have introduced the Kaplan-Meier (KM) method into power equipment lifespan prediction, providing new insights for improving asset wall prediction.

[0005] The Kaplan-Miller (KM) method originated from survival analysis in statistics. Proposed in 1958 by Edward L. Kaplan and Paul Meier, it is used to process censored data and estimate survival probabilities. Its core principle is to construct survival curves through nonparametric estimation, making it suitable for reliability analysis scenarios such as equipment lifespan and disease prognosis. In my country's engineering field, Qi Haiming applied the KM method to predict the remaining lifespan of satellite attitude control systems in 2015; Lei Tong used it for performance degradation analysis of emergency light battery modules in 2021, validating its effectiveness in power equipment reliability assessment; Wang Kai first used the KM method in asset wall prediction in 2023; and Jiang Changtai proposed combining the KM method with bearing failure data to characterize equipment lifespan in 2024.

[0006] However, the KM method has limitations in the risk analysis of power grid asset walls: its results are static estimates, assuming that future scrapping patterns are consistent with historical patterns, and it is difficult to reflect the dynamic impact of external factors such as environmental impact, operating status, and maintenance strategies on asset life over time, resulting in the prediction accuracy failing to meet the refined needs of power grid asset management. Summary of the Invention

[0007] The purpose of this invention is to provide a power grid asset wall risk prediction algorithm and system based on the KMP method to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a power grid asset wall risk prediction algorithm based on the KMP method, comprising the following steps: Step 1: Collect the historical decommissioned and scrapped asset list and the current in-service asset list of the power grid. Calculate the baseline scrapping quantity based on the age of the decommissioned and scrapped assets of the power grid to obtain the baseline scrapping probability time series. Step 2: Collect historical data on external factors, including but not limited to environmental impact, operational status, and maintenance strategies. Predict external factors that fluctuate over time and affect the baseline scrapping probability to obtain the environmental stress index time series. Step 3: Use the environmental pressure index time series from Step 2 to correct the baseline scrapping probability time series from Step 1, and obtain the final dynamically adjusted asset scrapping risk. Step 4: Determine the in-wall age limit for various types of assets by setting pre-defined asset scrapping thresholds; Step 5: Traverse all active assets in the power grid, calculate the value of in-wall assets for a specified year, and assess the scale of asset wall risk. Step 6: Predict the power grid investment capacity for the specified year, calculate the ratio of asset wall risk scale to power grid investment capacity, and determine the asset wall risk level.

[0009] Preferably, the calculation process of the baseline scrapping probability time series in step 1 is as follows: Step 1.1: Collect a list of historically decommissioned and scrapped assets and a list of currently in-service assets of the power grid. The historical decommissioned and scrapped asset list includes, but is not limited to, fields such as asset code, asset name, asset classification, original asset value, technical scale, commissioning date, decommissioning and scrapping date, and reason for decommissioning and scrapping; the current in-service asset list includes, but is not limited to, fields such as asset code, asset name, asset classification, original asset value, technical scale, commissioning date, and operating status; among which, technical scale includes, but is not limited to, the length of line assets and the capacity of substation assets. When calculating the unit price of such assets, the original asset value per unit length or capacity needs to be used.

[0010] Step 1.2: Calculate the age of all the above assets: The age of decommissioned and scrapped assets is determined by the time interval between the commissioning date and the decommissioning and scrapping date; the age of in-service assets is determined by the time interval between the commissioning date and the current date. Step 1.3: Use the KM method to calculate the survival probability of various assets at different ages: The KM survival probability curve is a continuous stepped curve plotted with equipment age on the horizontal axis and survival probability on the vertical axis. It is used to illustrate the relationship between equipment age and survival probability. Considering component-level survival probability analysis involves modeling survival probability curves for different types of components. The survival probability curve is calculated using the scrap ratio. Before the calculation, an assumption is made: when the equipment age is 0 years, its survival probability is 1. The iterative calculation formula for the survival probability curve is: ;in It is the first A point in time, namely age. Is The amount of assets that become obsolete upon reaching a certain age. Is The amount of assets still in active service at the age of [age missing]; Step 1.4: Calculate the baseline scrap probability time series: After obtaining the survival function, the survival difference between adjacent age groups is further calculated, and the formula for calculating the scrap probability curve is: .

[0011] Preferably, the environmental pressure index time series calculation steps in step 2 are as follows: Step 2.1 First, collect historical data on external factors. The environmental impact indicator used in the historical data on external factors is the corrosion coefficient of the asset at different ages. The operational status indicator used is the annual cumulative number of failures of the asset. The operation and maintenance strategy indicator used is the annual number of operation and maintenance operations for each age group. Then, the historical data of external factors are formatted into the ds format required by the Prophet model, outliers are removed using the Z-score method, and a logarithmic transformation is performed. The prediction structure of the Prophet model is represented as follows: ,in, for Time series values ​​at any given time; This is the trend term, used to simulate the non-periodic changes in time series. For periodic terms, it represents periodic changes in a time series; For a specific event term, it represents the external regression quantity, i.e., the abnormal impact of a specific event on the time series; This is the error term, which is usually assumed to follow a normal distribution; Step 2.2: Set up trend items and label specific events: Set the trend term as the cumulative scrap probability, and label specific events using the add_regressor function of the Prophet model to add the corrosion coefficient, number of failures, and number of maintenance operations as features; enable the Logistic Growth attribute of the Prophet model and set cap=1; Step 2.3: Train the Prophet model: By adding historical data of external factors using the add_regressor method, the Prophet model learns how historical data of external factors raises or lowers the baseline probability. Step 2.4: Determine the reasonableness of the model error: Calculate the root mean square error (RMSE) of the Prophet model. If RMSE > 0.10, recollect data or check for outliers until the error is reasonable. Step 2.5: Visualize and evaluate the prediction results. To visualize and evaluate the performance of the Prophet model on the probability curve of grid asset obsolescence, we first used Prophet's built-in plotting function for calendar time. The data is visualized; the virtual "dates" used internally by Prophet are remapped back to the "age" of the assets, and the baseline curve, model fitting curve, and scenario analysis curves under different external environments are highlighted; the rationality of the model is judged by checking whether the curve moves to the "upper left" when the external environment deteriorates (accelerating obsolescence) and whether it moves to the "lower right" when the environment improves (delaying obsolescence).

[0012] Preferably, the dynamically adjusted asset obsolescence risk in step 3 is calculated in the following way: Step 3.1: Calculate the adjustment factor: The Prophet method predicts the scrapping rates of various assets. Compared with the baseline expected scrap volume By comparison, an adjustment factor is obtained. ;for Annual adjustment factor: ; Step 3.2: Calculate the adjusted scrap probability: For a certain year in the future All assets in service Calculate the adjusted scrap probability: ,in Assets from KM exist The baseline scrap probability for the year is used to obtain the adjusted scrap probability. Step 3.3: Correct various scrap probability curves: The baseline scrap probability curve is corrected based on the adjusted scrap probability to complete the dynamic adjustment.

[0013] According to claim 1, the power grid asset wall risk prediction algorithm based on the KMP method is characterized in that: the specific implementation steps of step 4 are as follows: Step 4.1: Set the asset wall access threshold: Define asset wall access threshold , The threshold reflects the upper limit of a power grid company's tolerance for asset risk; according to power grid management requirements, it is set as follows: That is, when the predicted cumulative scrapping rate of a certain type of asset exceeds 80%, its corresponding age is the "lifespan" of that type of asset. Step 4.2: Calculate the dynamic cumulative scrap probability curve: Based on the dynamically adjusted asset obsolescence probability obtained in step 3 That is, in a specific year Below, different ages The probability density of asset obsolescence in the current year is converted into the cumulative obsolescence probability CDF; For a specific asset class Assuming its maximum design life or statistically observed maximum age is Then this type of asset at age Dynamic cumulative scrapping probability at time The calculation formula is: ,in, The age of the assets. ; Step 4.3: Determine the required years of occupancy for various types of assets: A comprehensive overview of the age of various asset classes Find the minimum age that satisfies the cumulative probability threshold condition; the incorporation period for various assets. Defined as: ; Step 4.4: Generate the asset inclusion criteria table: Based on the above calculations, a "Standard Table for Determining the Inclusion Period of Power Grid Assets" is generated for the target forecast year, which will serve as the basis for subsequent risk assessment.

[0014] Preferably, the specific implementation steps of step 5 are as follows: Step 5.1: Set the target year for forecasting and calculate the future service life of the assets: First, clarify the target year that needs to be predicted. (For example: predicting the size of the asset wall in 2028); then, iterate through the current list of in-service assets of the power grid, and calculate their expected service life in the target year based on the asset's commissioning date. The calculation formula is: ;in, Indicates the number in the list Item of assets, The year the asset was put into operation; Step 5.2: Filter the "In-wall Assets" list: The calculated expected service life is compared with the dynamic entry years for various assets determined in step 4 to screen out all risky assets that will enter the "asset wall" in the target year; let the first... The asset belongs to the following category: The dynamic in-wall lifespan for this category is... The decision logic is as follows: like If the asset is classified as an "in-wall asset," it will be included in the risk calculation; otherwise, the asset will be classified as a "safe asset" and will not be included in the current risk scale. Step 5.3: Calculate the overall risk scale of the asset wall: The values ​​of all selected "in-wall assets" are summed to obtain the target year's value. Total scale of power grid asset wall risk The calculation formula is: ,in, This refers to the set of all assets that meet the in-wall criteria for the target year. The desired asset wall size.

[0015] Preferably, the specific implementation steps of step 6 are as follows: Step 6.1: Predict the future investment capacity of the power grid: Collect data on the completed fixed asset investment in the power grid over the past N years, and use the compound annual growth rate formula: ,in Investment amount in the most recent year for Initial investment amount before the end of the year; Then predict investment capacity for the target year, based on the investment amount of the most recent year. Calculate the target year in the future Power grid investment capacity The calculation formula is: ,in The time interval between the current year and the target year, in years; Step 6.2: Calculate the funding requirements for the asset wall renovation, and consider the time value of money: The scale of the asset wall calculated in step 5 The calculation is based on the "original asset value." However, in actual equipment replacement and renovation, due to factors such as inflation and equipment technology upgrades, future replacement costs are usually higher than the historical original value. Therefore, it is necessary to then introduce interest or a cost growth factor to adjust the original value to obtain the actual asset renovation needs. The calculation formula is: ,in The preset average annual cost growth rate or interest rate is referenced to the Consumer Price Index (CPI). The time span from when the asset is put into operation to the target year; Step 6.3: Calculate the risk ratio: The revised asset wall renovation needs are compared with the projected grid investment capacity to calculate the asset wall risk ratio. The calculation formula is: ; Step 6.4: Determine the risk level of the asset wall: Based on the actual requirements of power grid asset management, risks are divided into three levels according to the ratio R, which is used to guide the formulation of investment strategies: Low risk The power grid investment capacity is greater than or equal to the asset wall renovation needs, the company's current investment plan covers the upcoming asset scrapping and replacement needs, the cash flow is healthy, and there is no pressure to "eliminate the wall"; Medium risk The demand for asset wall renovation exceeds investment capacity, but the excess is within 20%; a slight funding gap may occur in the future. It is recommended to alleviate the pressure by optimizing project priorities, appropriately increasing financing, or conducting life extension assessments for aging equipment; this is within a manageable range. High risk The demand for asset wall renovation far exceeds investment capacity, with a shortfall exceeding 20%. A severe lack of investment capacity will emerge in the future, unable to support large-scale asset upgrades. Without intervention, the power grid will face operational safety risks due to equipment exceeding its service life. A special "wall-removal" strategy must be launched immediately to secure additional dedicated funding or adjust power grid planning.

[0016] Preferably, a power grid asset wall risk prediction system based on the KMP method includes a baseline probability calculation module, an environmental pressure prediction module, a risk dynamic correction module, a wall entry age determination module, a risk scale calculation module, and a risk level assessment module. The baseline probability calculation module collects historical decommissioned and scrapped asset lists and current in-service asset lists for power grid assets, calculates the baseline scrapping quantity based on the age of the decommissioned and scrapped assets, and constructs a baseline scrapping probability time series. The environmental pressure prediction module acquires historical data on external factors, namely historical data on environmental impact, operating status, and maintenance strategies, and uses a time series model to predict future trends in external influencing factors, generating an environmental pressure index. The time series; the risk dynamic correction module: used to correct the benchmark scrapping probability time series using the environmental pressure index time series, generating a dynamically adjusted asset scrapping risk probability; the wall-mounting age determination module: used to determine the wall-mounting age of various assets based on a preset asset scrapping threshold and the dynamically adjusted asset scrapping risk probability; the risk scale calculation module: used to traverse all in-service assets in the power grid, calculate the value of wall-mounted assets in a specified year based on the wall-mounting age, and calculate the asset wall risk scale; the risk level assessment module: used to predict the power grid investment capacity in a specified year, calculate the ratio of the asset wall risk scale to the power grid investment capacity, and determine the asset wall risk level.

[0017] Compared with existing technologies, the advantages of this invention are as follows: This invention abandons the traditional method of prediction relying solely on static average lifespan or a single failure rate, and innovatively proposes a hybrid prediction model based on KM survival analysis and the Prophet time series algorithm. By establishing a baseline scrapping probability through the KM method and using the Prophet algorithm to dynamically correct the probability by incorporating external time-varying factors such as environment and operating status, it effectively solves the problem of traditional methods struggling to quantify the impact of external environmental changes on equipment lifespan. This makes the prediction results closer to the actual operating environment of the power grid, significantly improving prediction accuracy. This invention not only predicts the scale of asset scrapping but also further calculates the value of assets within the grid and correlates the risk scale with the power grid's investment capacity. By calculating the ratio of the asset risk scale to investment capacity, the risk level is quantitatively determined, providing scientific and quantitative data support for formulating precise technological upgrading investment plans and "wall elimination" strategies, avoiding operational risks caused by blind investment or insufficient funds. The system provided by this invention is based on microservices and AI computing power, breaking down data silos and realizing closed-loop management of the entire chain from multi-source data collection, cleaning, and fusion to risk analysis and disposal tracking, which greatly improves the efficiency and timeliness of power grid asset management. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2A schematic diagram of the KM survival probability curves for different asset types; Figure 3 This is a schematic diagram of the training process of the Prophet model in predicting the obsolescence trend of power grid assets. Figure 4 This is a schematic diagram of the system architecture of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides a technical solution: a power grid asset wall risk prediction algorithm based on the KMP method, comprising the following steps: Step 1: Collect the historical decommissioned and scrapped asset list and the current in-service asset list of the power grid. Calculate the baseline scrapping quantity based on the age of the decommissioned and scrapped assets of the power grid to obtain the baseline scrapping probability time series. Step 2: Collect historical data on external factors, including but not limited to environmental impact, operational status, and maintenance strategies. Predict external factors that fluctuate over time and affect the baseline scrapping probability to obtain the environmental stress index time series. Step 3: Use the environmental pressure index time series from Step 2 to correct the baseline scrapping probability time series from Step 1, and obtain the final dynamically adjusted asset scrapping risk. Step 4: Determine the in-wall age limit for various types of assets by setting pre-defined asset scrapping thresholds; Step 5: Traverse all active assets in the power grid, calculate the value of in-wall assets for a specified year, and assess the scale of asset wall risk. Step 6: Predict the power grid investment capacity for the specified year, calculate the ratio of asset wall risk scale to power grid investment capacity, and determine the asset wall risk level.

[0021] Furthermore, in step 1, power grid assets are the smallest asset management unit determined based on the company's asset management objectives, typically referring to tangible fixed assets used to support power grid operation. Professionally, power grid assets include five major categories: transmission, substation, distribution, communication, and automation. Due to the high value, diverse types, and large quantity of power grid assets, companies usually use financial software or specialized asset management software to establish asset ledgers. Both historical decommissioned and scrapped power grid asset lists and current in-service asset lists can be obtained through these asset ledgers. The calculation process for the benchmark scrapping probability time series is as follows: Step 1.1: Collect a list of historically decommissioned and scrapped assets and a list of currently in-service assets of the power grid. The historical decommissioned and scrapped asset list includes, but is not limited to, fields such as asset code, asset name, asset classification, original asset value, technical scale, commissioning date, decommissioning and scrapping date, and reason for decommissioning and scrapping; the current in-service asset list includes, but is not limited to, fields such as asset code, asset name, asset classification, original asset value, technical scale, commissioning date, and operating status; among which, technical scale includes, but is not limited to, the length of line assets and the capacity of substation assets. When calculating the unit price of such assets, the original asset value per unit length or capacity needs to be used.

[0022] Step 1.2: Calculate the age of all the above assets: The age of decommissioned and scrapped assets is determined by the time interval between the commissioning date and the decommissioning and scrapping date; the age of in-service assets is determined by the time interval between the commissioning date and the current date. Step 1.3: Use the KM method to calculate the survival probability of various assets at different ages: The KM survival probability curve is a continuous stepped curve plotted with equipment age on the horizontal axis and survival probability on the vertical axis. It is used to illustrate the relationship between equipment age and survival probability. Considering component-level survival probability analysis involves modeling survival probability curves for different types of components. The survival probability curve is calculated using the scrap ratio. Before the calculation, an assumption is made: when the equipment age is 0 years, its survival probability is 1. The iterative calculation formula for the survival probability curve is: ;in It is the first A point in time, namely age. Is The amount of assets that become obsolete upon reaching a certain age. Is The amount of assets still in active service at the age of [age missing]; Taking the conductors and towers, a subcategory of power transmission assets, as an example, a survival probability curve is plotted, such as... Figure 2 As shown in the figure, the expected lifespan of the conductor is approximately 25 to 30 years, while the expected lifespan of the tower is approximately 40 years.

[0023] Step 1.4: Calculate the baseline scrap probability time series: After obtaining the survival function, the survival difference between adjacent age groups is further calculated, and the formula for calculating the scrap probability curve is: .

[0024] Furthermore, the time series calculation steps for the environmental pressure index in step 2 are as follows: The primary reason for the decommissioning of power grid assets is the increased defect and failure rates caused by aging. The main causes of equipment aging include environmental impacts, operating conditions, and maintenance strategies. These differences influence the probability of decommissioning, thus requiring the transformation of these external factors into a time series of environmental stress indices. This step mainly utilizes the Prophet method, an open-source forecasting model specifically designed for time series prediction. It can be directly obtained from public function libraries in development environments such as Python. Prophet adopts a top-down modeling approach; its core purpose is not to analyze the failure probability of individual assets, but rather to directly model the temporal evolution of the total decommissioning probability at the power grid level. Its advantage lies in capturing dynamic trends and fitting the trend terms and specificities in the data well. The training and prediction process of the Prophet model is as follows: Figure 3 As shown, the process includes steps such as data preparation, trend modeling, defining external regression parameters, model training, judging the reasonableness of errors, and outputting and visualizing the results. This process allows us to obtain the total scrap volume for each future year. This refers to the macroeconomic failure trend of assets over time; the detailed calculation process is as follows: Step 2.1 First, collect historical data on external factors. The environmental impact indicator used in the historical data on external factors is the corrosion coefficient of the asset at different ages. The operational status indicator used is the annual cumulative number of failures of the asset. The operation and maintenance strategy indicator used is the annual number of operation and maintenance operations for each age group. Then, the historical data of external factors is formatted into the ds format required by the Prophet model; because the horizontal axis of the scrap probability curve is the equipment age, and the ds format of Prophet is date, the asset's "age (years)" needs to be mapped to virtual "calendar time". For example: 1 year old -> 2000-01-01, 2 years old -> 2000-01-02, and so on; Outliers are removed using the Z-score method and then a logarithmic transformation is performed. The prediction structure of the Prophet model is represented as follows: ,in, for Time series values ​​at any given time; This is the trend term, used to simulate the non-periodic changes in time series. For periodic terms, it represents periodic changes in a time series; For a specific event term, it represents the external regression quantity, i.e., the abnormal impact of a specific event on the time series; This is the error term, which is usually assumed to follow a normal distribution; Step 2.2: Set up trend items and label specific events: Set the trend term as the cumulative scrap probability, and label specific events using the add_regressor function of the Prophet model to add the corrosion coefficient, number of failures, and number of maintenance operations as features; enable the Logistic Growth attribute of the Prophet model and set cap=1; Table 1-1 shows an example of Prophet model settings written in Python: Table 1-1 Examples of Prophet Model Settings import pandas as pd import numpy as np from prophet import Prophet #1. Data Preparation (Ages 0-50) ages = np.arange(0, 51) #Constructing a virtual date: Starting from 2000-01-01, each day represents one year of age. dummy_dates=pd.to_datetime('2000-01-01') + pd.to_timedelta(ages, unit='D') #Read the baseline scrapping probability curve, taking transmission lines as an example, omitting the middle 40 years in the set. base_prob=[0.0, 0.0, 0.002, 0.003,0.003…0.997,0.998,1,1,1] #2. Incorporate external influencing factors (Regressors) #The data represent: corrosion level (corrosion_level, 0-10), cumulative failures (failure_count, times), and maintenance frequency (maint_freq, times / year). df['corrosion_level']=[0.0,0.0,0.0, 0.1,0.21…0.8,0.9,0.9,0.9,1] df['failure_count']= [0,0,1,2,4…127,130,136,139,140] df['maint_freq']=np.random.choice(0,0,1,1,1…3,4,4,4,4) Step 2.3: Train the Prophet model: By adding historical data of external factors using the add_regressor method, the Prophet model learns how historical data of external factors raises or lowers the baseline probability; Table 1-2 shows the Python implementation code.

[0025] Table 1-2 Prophet Model Training #3. Initialize the model #Key Settings: #growth='logistic': Ensures the curve has an upper limit. #seasonality_mode='additive': Linear superposition of external factors #daily / weekly / yearly_seasonality=False: Turns off all periodicity because age does not have a weekly / yearly cycle. m=Prophet( growth='logistic', daily_seasonality=False, weekly_seasonality=False, yearly_seasonality=False, changepoint_prior_scale=0.05 # Controls curve smoothness ) #4. Add external factors (correction item) #mode='additive' indicates that these factors will linearly increase or decrease the probability of scrapping. m.add_regressor('corrosion_level') m.add_regressor('failure_count') m.add_regressor('maint_freq') #5. Fitting the baseline curve m.fit(df) Step 2.4: Determine the reasonableness of the model error: Calculate the root mean square error (RMSE) of the Prophet model. If RMSE > 0.10, recollect data or check for outliers until the error is reasonable. Step 2.5: Visualize and evaluate the prediction results. To visualize and evaluate the performance of the Prophet model on the probability curve of grid asset obsolescence, we first used Prophet's built-in plotting function for calendar time. The model is visualized graphically; the virtual "dates" used internally by Prophet are remapped back to the "age" of assets, with a focus on displaying baseline curves, model fitting curves, and scenario analysis curves under different external environments; the model's rationality is assessed by examining whether the curve shifts "upper left" (accelerating obsolescence) when the external environment deteriorates and whether it shifts "lower right" (delaying obsolescence) when the environment improves. Furthermore, the dynamically adjusted asset obsolescence risk in step 3 is calculated in the following way: Specifically, in step 3, the baseline scrap probability has been corrected using the environmental stress index, resulting in a corrected scrap probability curve that better reflects the current power grid operating environment. The core of this step lies in defining the boundaries of the "asset wall," that is, determining at what age various assets are considered to have entered a high-risk scrapping period, i.e., the "wall entry age." According to the power grid's safety management requirements, when the cumulative scrap probability of an asset reaches 80%, it means that the asset has entered its theoretical scrapping period, and maintenance costs and failure risks increase significantly; thus, the asset is determined to have entered the "asset wall." Step 3.1: Calculate the adjustment factor: The Prophet method predicts the scrapping rates of various assets. Compared with the baseline expected scrap volume By comparison, an adjustment factor is obtained. ;for Annual adjustment factor: ; Step 3.2: Calculate the adjusted scrap probability: For a certain year in the future All assets in service Calculate the adjusted scrap probability: ,in Assets from KM exist The baseline scrap probability for the year is used to obtain the adjusted scrap probability. Step 3.3: Correct various scrap probability curves: The baseline scrap probability curve is corrected based on the adjusted scrap probability to complete the dynamic adjustment.

[0026] According to claim 1, the specific implementation steps of step 4 of the power grid asset wall risk prediction algorithm based on the KMP method are as follows: Step 4.1: Set the asset wall access threshold: Define asset wall access threshold , The threshold reflects the upper limit of a power grid company's tolerance for asset risk; according to power grid management requirements, it is set as follows: That is, when the predicted cumulative scrapping rate of a certain type of asset exceeds 80%, its corresponding age is the "lifespan" of that type of asset. Step 4.2: Calculate the dynamic cumulative scrap probability curve: Based on the dynamically adjusted asset obsolescence probability obtained in step 3 That is, in a specific year Below, different ages The probability density of asset obsolescence in the current year is converted into the cumulative obsolescence probability CDF; For a specific asset class Assuming its maximum design life or statistically observed maximum age is Then this type of asset at age Dynamic cumulative scrapping probability at time The calculation formula is: ,in, The age of the assets. ; Step 4.3: Determine the required years of occupancy for various types of assets: A comprehensive overview of the age of various asset classes Find the minimum age that satisfies the cumulative probability threshold condition; the incorporation period for various assets. Defined as: ; Step 4.4: Generate the asset inclusion criteria table: Based on the above calculations, a "Standard Table for Determining the Inclusion Period of Power Grid Assets" is generated for the target forecast year, which will serve as the basis for subsequent risk assessment.

[0027] Furthermore, the specific implementation steps of step 5 are as follows: Step 5.1: Set the target year for forecasting and calculate the future service life of the assets: First, clarify the target year that needs to be predicted. (For example: predicting the size of the asset wall in 2028); then, iterate through the current list of in-service assets of the power grid, and calculate their expected service life in the target year based on the asset's commissioning date. The calculation formula is: ;in, Indicates the number in the list Item of assets, The year the asset was put into operation; Step 5.2: Filter the "In-wall Assets" list: The calculated expected service life is compared with the dynamic entry years for various assets determined in step 4 to screen out all risky assets that will enter the "asset wall" in the target year; let the first... The asset belongs to the following category: The dynamic in-wall lifespan for this category is... The decision logic is as follows: like If the asset is classified as an "in-wall asset," it will be included in the risk calculation; otherwise, the asset will be classified as a "safe asset" and will not be included in the current risk scale. Step 5.3: Calculate the overall risk scale of the asset wall: The values ​​of all selected "in-wall assets" are summed to obtain the target year's value. Total scale of power grid asset wall risk The calculation formula is: ,in, This refers to the set of all assets that meet the in-wall criteria for the target year. The desired asset wall size.

[0028] Furthermore, the specific implementation steps of step 6 are as follows: Step 6.1: Predict the future investment capacity of the power grid: Collect data on the completed fixed asset investment in the power grid over the past N years, and use the compound annual growth rate formula: ,in Investment amount in the most recent year for Initial investment amount before the end of the year; Then predict investment capacity for the target year, based on the investment amount of the most recent year. Calculate the target year in the future Power grid investment capacity The calculation formula is: ,in The time interval between the current year and the target year, in years; Step 6.2: Calculate the funding requirements for the asset wall renovation, and consider the time value of money: The scale of the asset wall calculated in step 5 The calculation is based on the "original asset value." However, in actual equipment replacement and renovation, due to factors such as inflation and equipment technology upgrades, future replacement costs are usually higher than the historical original value. Therefore, it is necessary to then introduce interest or a cost growth factor to adjust the original value to obtain the actual asset renovation needs. The calculation formula is: ,in The preset average annual cost growth rate or interest rate is referenced to the Consumer Price Index (CPI). The time span from when the asset is put into operation to the target year; Step 6.3: Calculate the risk ratio: The revised asset wall renovation needs are compared with the projected grid investment capacity to calculate the asset wall risk ratio. The calculation formula is: ; Step 6.4: Determine the risk level of the asset wall: Based on the actual requirements of power grid asset management, risks are divided into three levels according to the ratio R, which is used to guide the formulation of investment strategies: Low risk The power grid investment capacity is greater than or equal to the asset wall renovation needs, the company's current investment plan covers the upcoming asset scrapping and replacement needs, the cash flow is healthy, and there is no pressure to "eliminate the wall"; Medium risk The demand for asset wall renovation exceeds investment capacity, but the excess is within 20%; a slight funding gap may occur in the future. It is recommended to alleviate the pressure by optimizing project priorities, appropriately increasing financing, or conducting life extension assessments for aging equipment; this is within a manageable range. High risk The demand for asset wall renovation far exceeds investment capacity, with a shortfall exceeding 20%. A severe lack of investment capacity will emerge in the future, unable to support large-scale asset upgrades. Without intervention, the power grid will face operational safety risks due to equipment exceeding its service life. A special "wall-removal" strategy must be launched immediately to secure additional dedicated funding or adjust power grid planning.

[0029] Furthermore, a power grid asset wall risk prediction system based on the KMP method includes a baseline probability calculation module, an environmental pressure prediction module, a risk dynamic correction module, a wall entry age determination module, a risk scale calculation module, and a risk level assessment module. The baseline probability calculation module collects historical decommissioned and scrapped asset lists and current in-service asset lists for power grid assets, calculates the baseline scrapping quantity based on the age of the decommissioned and scrapped assets, and constructs a baseline scrapping probability time series. The environmental pressure prediction module acquires historical data on external factors, namely environmental impact, operating status, and maintenance strategies, and uses a time series model to predict future trends in external influencing factors, generating an environmental pressure index. The time series; the risk dynamic correction module: used to correct the benchmark scrapping probability time series using the environmental pressure index time series, generating a dynamically adjusted asset scrapping risk probability; the wall-mounting age determination module: used to determine the wall-mounting age of various assets based on a preset asset scrapping threshold and the dynamically adjusted asset scrapping risk probability; the risk scale calculation module: used to traverse all in-service assets in the power grid, calculate the value of wall-mounted assets in a specified year based on the wall-mounting age, and calculate the asset wall risk scale; the risk level assessment module: used to predict the power grid investment capacity in a specified year, calculate the ratio of the asset wall risk scale to the power grid investment capacity, and determine the asset wall risk level.

[0030] In one specific embodiment of the present invention, the power grid asset wall risk prediction system adopts a layered architecture design (e.g., B / S architecture), which includes, from bottom to top, a data layer, an algorithm layer, a business layer, and an application layer: Data and Computing Power Foundation Layer: As the underlying support of the system, this layer is responsible for connecting external data platforms and multi-source databases. This layer can use graph database engines (such as Neo4j) to build a knowledge graph of power grid assets and configure intelligent algorithm containers (such as PyAILAB) to provide computing power support.

[0031] Platform Service Layer: Built on a microservice architecture (such as Spring Cloud), this layer integrates data processing frameworks (such as Python) and intelligent proxy engines (such as Dify). It is responsible for providing standardized AI capability interfaces and data analysis services to upper-layer applications, enabling elastic scheduling of computing resources.

[0032] Business Logic Layer: This layer incorporates the core KMP hybrid prediction algorithm model of this invention. It specifically executes core business logic such as data cleaning, multi-source data fusion, asset wall simulation testing, and asset evaluation within the asset wall, supporting risk analysis and strategy formulation for provincial power companies.

[0033] Application Interaction Layer: Provides users with a visual operation interface, covering functions such as data source configuration, provincial / multi-province data management, asset wall risk visualization analysis and system permission management, realizing closed-loop tracking from data input to asset disposal.

[0034] See system architecture Figure 4 As shown.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power grid asset wall risk prediction algorithm based on the KMP method, characterized in that, Includes the following steps: Step 1: Collect the historical decommissioned and scrapped asset list and the current in-service asset list of the power grid. Calculate the baseline scrapping quantity based on the age of the decommissioned and scrapped assets of the power grid to obtain the baseline scrapping probability time series. Step 2: Collect historical data on external factors, including but not limited to environmental impact, operational status, and maintenance strategies. Predict external factors that fluctuate over time and affect the baseline scrapping probability to obtain the environmental stress index time series. Step 3: Use the environmental pressure index time series from Step 2 to correct the baseline scrapping probability time series from Step 1, and obtain the final dynamically adjusted asset scrapping risk. Step 4: Determine the in-wall age limit for various types of assets by setting pre-defined asset scrapping thresholds; Step 5: Traverse all active assets in the power grid, calculate the value of in-wall assets for a specified year, and assess the scale of asset wall risk. Step 6: Predict the power grid investment capacity for the specified year, calculate the ratio of asset wall risk scale to power grid investment capacity, and determine the asset wall risk level.

2. The power grid asset wall risk prediction algorithm based on the KMP method according to claim 1, characterized in that: The calculation process of the baseline scrapping probability time series in step 1 is as follows: Step 1.1: Collect a list of historically decommissioned and scrapped assets and a list of currently in-service assets of the power grid. The historical decommissioned and scrapped asset list includes, but is not limited to, fields such as asset code, asset name, asset classification, original asset value, technical scale, commissioning date, decommissioning and scrapping date, and reason for decommissioning and scrapping; the current in-service asset list includes, but is not limited to, fields such as asset code, asset name, asset classification, original asset value, technical scale, commissioning date, and operating status; among which, technical scale includes, but is not limited to, the length of line assets and the capacity of substation assets. Step 1.2: Calculate the age of all the above assets: The age of decommissioned and scrapped assets is determined by the time interval between the commissioning date and the decommissioning and scrapping date; the age of in-service assets is determined by the time interval between the commissioning date and the current date. Step 1.3: Use the KM method to calculate the survival probability of various assets at different ages: The KM survival probability curve is a continuous stepped curve plotted with equipment age on the horizontal axis and survival probability on the vertical axis. It illustrates the relationship between equipment age and survival probability. The survival probability curve is calculated using the scrap ratio. Before the calculation, an assumption is made: when the equipment age is 0 years, its survival probability is 1. The iterative calculation formula for the survival probability curve is: ;in It is the first A point in time, namely age. Is The amount of assets that become obsolete upon reaching a certain age. Is The amount of assets still in active service at the age of [age missing]; Step 1.4: Calculate the baseline scrap probability time series: After obtaining the survival function, the survival difference between adjacent age groups is further calculated, and the formula for calculating the scrap probability curve is: .

3. The power grid asset wall risk prediction algorithm based on the KMP method according to claim 1, characterized in that: The time series calculation steps for the environmental pressure index in step 2 are as follows: Step 2.1 First, collect historical data on external factors. The environmental impact indicator used in the historical data on external factors is the corrosion coefficient of the asset at different ages. The operational status indicator used is the annual cumulative number of failures of the asset. The operation and maintenance strategy indicator used is the annual number of operation and maintenance operations for each age group. Then, the historical data of external factors are formatted into the ds format required by the Prophet model, outliers are removed using the Z-score method, and a logarithmic transformation is performed. The prediction structure of the Prophet model is represented as follows: ,in, for Time series values ​​at any given time; This is the trend term, used to simulate the non-periodic changes in time series. For periodic terms, it represents periodic changes in a time series; For a specific event term, it represents the external regression quantity, i.e., the abnormal impact of a specific event on the time series; This is the error term, which is usually assumed to follow a normal distribution; Step 2.2: Set up trend items and label specific events: Set the trend term as the cumulative scrap probability, and label specific events using the add_regressor function of the Prophet model to add the corrosion coefficient, number of failures, and number of maintenance operations as features; enable the Logistic Growth attribute of the Prophet model and set cap=1; Step 2.3: Train the Prophet model: By adding historical data of external factors using the add_regressor method, the Prophet model learns how historical data of external factors raises or lowers the baseline probability. Step 2.4: Determine the reasonableness of the model error: Calculate the root mean square error (RMSE) of the Prophet model. If RMSE > 0.10, recollect data or check for outliers until the error is reasonable. Step 2.5: Visualize and evaluate the prediction results. To visualize and evaluate the performance of the Prophet model on the probability curve of grid asset obsolescence, we first used Prophet's built-in plotting function for calendar time. It provides a graphical representation; it remaps the virtual "dates" used internally by Prophet back to the "age" of the assets, and highlights the baseline curves, model fit curves, and scenario analysis curves under different external environments.

4. The power grid asset wall risk prediction algorithm based on the KMP method according to claim 1, characterized in that: The dynamically adjusted asset obsolescence risk in step 3 is calculated using the following method: Step 3.1: Calculate the adjustment factor: The Prophet method predicts the scrapping rates of various assets. Compared with the baseline expected scrap volume By comparison, an adjustment factor is obtained. ;for Annual adjustment factor: ; Step 3.2: Calculate the adjusted scrap probability: For a certain year in the future All assets in service Calculate the adjusted scrap probability: ,in Assets from KM exist The adjusted scrap probability is obtained from the baseline scrap probability for the year. Step 3.3: Correct various scrap probability curves: The baseline scrap probability curve is corrected based on the adjusted scrap probability to complete the dynamic adjustment.

5. The power grid asset wall risk prediction algorithm based on the KMP method according to claim 1, characterized in that: The specific implementation steps of step 4 are as follows: Step 4.1: Set the asset wall access threshold: Define asset wall access threshold , The threshold reflects the upper limit of a power grid company's tolerance for asset risk; According to power grid management requirements, the following settings are made: That is, when the predicted cumulative scrapping rate of a certain type of asset exceeds 80%, its corresponding age is the "lifespan" of that type of asset. Step 4.2: Calculate the dynamic cumulative scrap probability curve: Based on the dynamically adjusted asset obsolescence probability obtained in step 3 That is, in a specific year Below, different ages The probability density of asset obsolescence in the current year is converted into the cumulative obsolescence probability CDF; For a specific asset class Assuming its maximum design life or statistically observed maximum age is Then this type of asset at age Dynamic cumulative scrapping probability at time The calculation formula is: ,in, The age of the assets. ; Step 4.3: Determine the required years of occupancy for various types of assets: A comprehensive overview of the age of various asset classes Find the minimum age that satisfies the cumulative probability threshold condition; the incorporation period for various assets. Defined as: ; Step 4.4: Generate the asset inclusion criteria table: Based on the above calculations, a "Standard Table for Determining the Inclusion Period of Power Grid Assets" is generated for the target forecast year, which will serve as the basis for subsequent risk assessment.

6. The power grid asset wall risk prediction algorithm based on the KMP method according to claim 1, characterized in that: The specific implementation steps of step 5 are as follows: Step 5.1: Set the target year for forecasting and calculate the future service life of the assets: First, clarify the target year that needs to be predicted. Then, iterate through the current list of in-service assets of the power grid, and calculate the expected service life of each asset in the target year based on its commissioning date. The calculation formula is as follows: ;in, Indicates the number in the list Item of assets, The year the asset was put into operation; Step 5.2: Filter the "In-wall Assets" list: The calculated expected service life is compared with the dynamic entry years for various assets determined in step 4 to screen out all risky assets that will enter the "asset wall" in the target year; let the first... The asset belongs to the following category: The dynamic in-wall lifespan for this category is... The decision logic is as follows: like If the asset is classified as an "in-wall asset," it will be included in the risk calculation; otherwise, it will be classified as a "safe asset" and will not be included in the current risk scale. Step 5.3: Calculate the overall risk scale of the asset wall: The values ​​of all selected "in-wall assets" are summed to obtain the target year's value. Total scale of power grid asset wall risk The calculation formula is: ,in, This refers to the set of all assets that meet the in-wall criteria for the target year. The desired asset wall size.

7. The power grid asset wall risk prediction algorithm based on the KMP method according to claim 1, characterized in that: The specific implementation steps of step 6 are as follows: Step 6.1: Predict the future investment capacity of the power grid: Collect data on the completed fixed asset investment in the power grid over the past N years, and use the compound annual growth rate formula: ,in Investment amount in the most recent year for Initial investment amount before the end of the year; Then predict investment capacity for the target year, based on the investment amount of the most recent year. Calculate the target year in the future Power grid investment capacity The calculation formula is: ,in The time interval between the current year and the target year, in years; Step 6.2: Calculate the funding requirements for the asset wall renovation, and consider the time value of money: The scale of the asset wall calculated in step 5 It is calculated based on the "original value of the asset," and then interest or construction cost growth factors are introduced to adjust the original value to obtain the actual asset wall renovation needs. The calculation formula is: ,in The preset average annual cost growth rate or interest rate is referenced to the Consumer Price Index (CPI). The time span from when the asset is put into operation to the target year; Step 6.3: Calculate the risk ratio: The revised asset wall renovation needs are compared with the projected grid investment capacity to calculate the asset wall risk ratio. The calculation formula is: ; Step 6.4: Determine the risk level of the asset wall: Based on the actual requirements of power grid asset management, risks are divided into three levels according to the ratio R, which is used to guide the formulation of investment strategies: Low risk The power grid investment capacity is greater than or equal to the asset wall renovation needs, the company's current investment plan covers the upcoming asset scrapping and replacement needs, the cash flow is healthy, and there is no pressure to "eliminate the wall"; Medium risk The demand for asset wall renovation exceeds the investment capacity, but the excess is within 20%. High risk The demand for asset wall renovation far exceeds the investment capacity, with a gap of more than 20%.

8. A power grid asset wall risk prediction system based on the KMP method according to any one of claims 1-7, characterized in that, The system includes a baseline probability calculation module, an environmental pressure prediction module, a risk dynamic correction module, a wall-mounted age determination module, a risk scale calculation module, and a risk level assessment module. The baseline probability calculation module collects historical decommissioned and scrapped asset lists and current in-service asset lists for power grid assets, calculates the baseline scrapping quantity based on the age of the decommissioned and scrapped assets, and constructs a baseline scrapping probability time series. The environmental pressure prediction module acquires historical data on external factors, namely environmental impact, operational status, and maintenance strategies, and uses a time series model to predict future trends in external influencing factors, generating an environmental pressure index time series. The risk dynamic correction module uses the environmental pressure index time series to correct the baseline scrapping probability time series, generating a dynamically adjusted asset scrapping risk probability. The in-wall age determination module is used to determine the in-wall age of various assets based on a preset asset scrapping threshold and a dynamically adjusted asset scrapping risk probability. The risk scale calculation module is used to traverse all in-service assets in the power grid, calculate the in-wall asset value for a specified year based on the in-wall age, and calculate the asset wall risk scale. The risk level assessment module is used to predict the power grid investment capacity for a specified year, calculate the ratio of the asset wall risk scale to the power grid investment capacity, and determine the asset wall risk level.

Images