A method, system, storage medium and program product for predicting operation and maintenance costs

Abstract

A method, system, storage medium and program product for predicting operation and maintenance costs, in which historical data are divided into a high-cost data group and a low-cost data group; a pre-sign sequence of a high-cost event is determined; a feature segment whose similarity to the pre-sign sequence is greater than a preset threshold is retrieved in the low-cost data group; a bifurcation point data table is generated based on similar feature segments; a feature distance is calculated between the real-time acquired operation data and the bifurcation point data table to obtain a feature distance matrix; the probability weights of the real-time acquired operation data on different cost evolution paths in the bifurcation point data table are calculated based on the feature distance matrix; several cost evolution branches are generated according to the probability weights; the degree of fit between each cost evolution branch and the historical data is calculated; and each cost evolution branch is weighted based on the degree of fit to obtain the operation and maintenance cost prediction result. The present application reduces the deviation between the prediction result and the actual operation and maintenance cost, thereby improving the accuracy of the operation and maintenance cost prediction.

Description

Technical Field

[0001] The present application belongs to the field of methods specifically applicable to management purposes, and in particular, relates to an operation and maintenance cost prediction method, system, storage medium and program product. Background Art

[0002] With the continuous growth of global energy demand and the increasing complexity of energy systems, energy operation and maintenance work faces greater and greater challenges. Energy operation and maintenance not only needs to ensure the stability and reliability of energy supply, but also needs to control operation and maintenance costs. However, due to the lack of effective operation and maintenance cost forecasting methods, energy companies often find it difficult to accurately estimate the operation and maintenance budget, resulting in problems such as overspending on operation and maintenance costs or improper resource allocation, which has an adverse impact on the operation management and economic benefits of the energy system.

[0003] In related technologies, the operation and maintenance costs in the future can be predicted using a time series model by analyzing historical energy consumption data and operation and maintenance records. First, collect historical operation data of the energy system, such as electricity, oil and gas energy consumption, and corresponding operation and maintenance work order information, such as equipment maintenance, fault handling, etc. Then, organize these data in time series, use time series prediction models such as ARIMA and Prophet, and establish a cost prediction model by fitting historical data. When making a prediction, input the energy demand estimate and operation and maintenance plan for the future period, and substitute them into the model to obtain the predicted value of the operation and maintenance costs for the corresponding period.

[0004] However, due to the complex and changeable operating environment of the energy system and the numerous factors that affect the operation and maintenance costs, it is difficult to fully consider the impact of various complex factors such as weather changes, equipment aging, energy price fluctuations, etc. by relying solely on a single time series model. These factors may cause a large deviation between the predicted results and the actual operation and maintenance costs, reducing the accuracy of the operation and maintenance cost forecast. Summary of the invention

[0005] The present application provides an operation and maintenance cost prediction method, system, storage medium and program product, which are used to reduce the deviation between the prediction results and the actual operation and maintenance costs, thereby improving the accuracy of the operation and maintenance cost prediction.

[0006] In a first aspect, the present application provides an operation and maintenance cost prediction method, which obtains the historical operation data of the energy system within a preset time period, and the historical data includes energy consumption data, equipment operation and maintenance records, environmental data and cost records;

[0007] Divide the historical data into a high-cost data group and a low-cost data group according to the numerical value of the cost record;

[0008] Based on the expense records of the high-expense data group, time series backtracking is performed to obtain the preceding symptom sequence of the high-expense event;

[0009] Retrieving feature segments whose similarity to the preceding symptom sequence is greater than a preset threshold in the low-cost data set;

[0010] Generate a bifurcation point data table of cost evolution based on the characteristic fragments, the bifurcation point data table includes critical conditions for cost change;

[0011] The characteristic distance calculation is performed between the real-time acquired operation data and the bifurcation point data table to obtain a characteristic distance matrix;

[0012] Calculate the probability weights of the real-time acquired operation data on different cost evolution paths in the bifurcation point data table based on the characteristic distance matrix;

[0013] Generate several cost evolution branches within the forecast period according to probability weights;

[0014] Calculate the degree of consistency between each cost evolution branch and historical data;

[0015] Based on the degree of fit, each cost evolution branch is weighted and calculated to obtain the operation and maintenance cost forecast result.

[0016] By adopting the above technical solution, by obtaining historical data containing energy consumption data, equipment operation and maintenance records, environmental data and cost records within a preset time period, and dividing it into high and low cost data groups according to the size of the cost records, the abnormal cost interval can be clearly identified. The high cost data group is traced back in time to obtain the pre-symptom sequence, and similar feature fragments are retrieved in the low cost data group and a bifurcation point data table is generated to establish a mapping relationship between the key nodes of cost evolution and the system operation status. The feature distance matrix of the real-time operation data and the bifurcation point data table is calculated and the probability weight is determined, so that the cost evolution trend of the system can be accurately predicted. By generating multiple cost evolution branches and performing weighted calculations based on the degree of fit with historical data, the possibility of a single evolution path is considered, and the uncertainty assessment of multi-scenario prediction is introduced, which improves the reliability of the prediction results, makes the prediction results closer to the actual operation situation, reduces the deviation between the prediction results and the actual operation and maintenance costs, and thus improves the accuracy of the operation and maintenance cost prediction.

[0017] In combination with some embodiments of the first aspect, in some embodiments, a time series backtracking is performed based on the cost records of the high-cost data group to obtain a pre-symptom sequence of a high-cost event, specifically including:

[0018] Calculate the cost change rate at adjacent moments in the high-cost data group and identify the cost mutation points;

[0019] A reverse tracing window is set based on the cost mutation point, and the time series characteristics of environmental data, energy consumption data and equipment operation and maintenance records are extracted within the reverse tracing window;

[0020] Count the fluctuation range of time series characteristics, and mark the characteristic parameters whose fluctuation range is greater than the preset fluctuation range;

[0021] A time series association graph is constructed based on characteristic parameters, and a characteristic sequence of cost mutation events in the time series association graph is extracted. A cost mutation event is a cost event that has a duration greater than a preset time and a cost peak value greater than a preset cost value;

[0022] Compare and analyze the characteristic sequences of different cost mutation events, and take the recurring characteristic sequences as the preceding sign sequences.

[0023] By adopting the above technical solution, the cost mutation point is identified by calculating the cost change rate at adjacent moments in the high-cost data group, and the time series characteristics of environmental data, energy consumption data and equipment operation and maintenance records are extracted in the reverse tracing window, so as to achieve accurate positioning of cost anomalies. The fluctuation range of the time series characteristics is counted and the characteristic parameters exceeding the preset fluctuation range are marked. The time series correlation map is constructed and the characteristic sequence of cost mutation events is extracted, and the correlation model between cost mutation and system operation status is established. The key features closely related to cost mutation can be extracted from massive operation data, which improves the accuracy and timeliness of cost anomaly warning.

[0024] In combination with some embodiments of the first aspect, in some embodiments, weighted calculation is performed on each cost evolution branch based on the degree of fit to obtain an operation and maintenance cost prediction result, which specifically includes:

[0025] Calculate the Euclidean distance and dynamic time warping distance between each cost evolution branch and historical data;

[0026] Construct a comprehensive similarity evaluation index based on Euclidean distance and dynamic time warping distance;

[0027] The comprehensive similarity evaluation index is normalized to obtain the weight coefficient of each cost evolution branch;

[0028] The predicted values ​​of each cost evolution branch are weighted averaged based on the weight coefficient to obtain the operation and maintenance cost prediction result.

[0029] By adopting the above technical solution, by calculating the Euclidean distance and dynamic time warping distance between the cost evolution branch and the historical data, a comprehensive similarity evaluation index is constructed and normalized to obtain the weight coefficient, and a quantitative evaluation mechanism for the credibility of the prediction results is established. The Euclidean distance reflects the degree of closeness between the predicted value and the historical data in terms of numerical value, and the dynamic time warping distance reflects the similarity of the morphological characteristics of the time series. The weight coefficient is used to perform a weighted average on the predicted values ​​of each cost evolution branch, so that the final prediction result not only maintains consistency with the historical data, but also enhances the scientificity and reliability of the prediction result.

[0030] In conjunction with some embodiments of the first aspect, in some embodiments, after performing weighted calculation on each cost evolution branch based on the degree of agreement to obtain an operation and maintenance cost prediction result, the method further includes:

[0031] The time series decomposition method is used to extract trend items, period items and random items from the operation and maintenance cost forecast results;

[0032] Construct the cost growth acceleration indicator based on the changing slope of the trend item;

[0033] Calculate the expense abnormality score based on the fluctuation characteristics of the periodic items;

[0034] The entropy method is used to evaluate the uncertainty level of random items;

[0035] Fusion of acceleration index, anomaly score and uncertainty level to generate warning threshold curve;

[0036] Provide graded warning based on the warning threshold curve.

[0037] By adopting the above technical solution, the operation and maintenance cost forecast results are decomposed into trend items, cycle items and random items through time series decomposition, realizing a refined analysis of the cost change characteristics. Based on the change slope of the trend item, the cost growth acceleration index is constructed, and the cost anomaly degree score is calculated in combination with the fluctuation characteristics of the cycle item. The entropy method is used to evaluate the uncertainty level of the random item, and a multi-dimensional cost change feature characterization system is established. By integrating the acceleration index, the anomaly degree score and the uncertainty level to generate the early warning threshold curve, a dynamic early warning of cost anomalies is realized. This method constructs an early warning mechanism from the three dimensions of the speed, volatility and uncertainty of cost changes, improves the sensitivity and reliability of the early warning, and reduces the false alarms and missed alarms of the early warning.

[0038] In conjunction with some embodiments of the first aspect, in some embodiments, performing graded warning based on the warning threshold curve specifically includes:

[0039] Construct a multi-dimensional early warning feature space that includes cost mutation risk, equipment status, and environmental impact;

[0040] The kernel principal component analysis method is used to reduce the dimension of the multidimensional warning feature space and obtain the reduced-dimensional multidimensional warning features;

[0041] Construct a state transfer matrix based on the reduced-dimensional multi-dimensional warning features;

[0042] A dynamic warning response plan containing multi-level warning information is generated based on the state transition matrix.

[0043] By adopting the above technical solutions, the system's warning status characteristics are fully characterized by constructing a multidimensional warning feature space that includes cost mutation risks, equipment status, and environmental impacts. The kernel principal component analysis method is used to reduce the dimensionality of the multidimensional warning feature space, which reduces data redundancy while retaining the main information and improves the computational efficiency of warning analysis. The stability of the warning status is evaluated by combining the Bayesian network model, which improves the reliability of the warning results. Finally, a dynamic warning response plan containing multi-level warning information is generated, which enhances the accuracy, timeliness, and effectiveness of the warning.

[0044] In combination with some embodiments of the first aspect, in some embodiments, after performing graded warning based on the warning threshold curve, the method further includes:

[0045] Construct high-dimensional heterogeneous complex networks and identify key influencing paths within them;

[0046] Based on the dynamic feature extraction and spatiotemporal evolution pattern of the core node group, a deep representation learning framework with multiple attention mechanisms is constructed;

[0047] Use the deep representation learning framework to dynamically simulate the response characteristics under different operation and maintenance strategies;

[0048] Based on the response characteristics, a multi-objective optimization problem with nonlinear coupling constraints is constructed, and an improved hybrid intelligent algorithm is used to solve it and obtain the solution;

[0049] Generate the optimal operation and maintenance strategy set based on the solution results;

[0050] Convert the optimal operation and maintenance strategy set into a phased implementation plan.

[0051] By adopting the above technical solutions, the spatiotemporal evolution pattern is extracted based on the dynamic features of the core node group, and a deep representation learning framework containing multiple attention mechanisms is constructed to accurately characterize the dynamic characteristics of the system. The response characteristics under different operation and maintenance strategies are dynamically simulated using the deep representation learning framework, and the effects of various operation and maintenance strategies can be pre-evaluated. By constructing a multi-objective optimization problem with nonlinear coupling constraints and solving it with an improved hybrid intelligent algorithm, the optimal operation and maintenance strategy set under multiple objectives is obtained, which comprehensively considers the complexity, dynamics and multi-objective nature of the system. Through the combination of deep learning and intelligent optimization algorithms, the scientific nature and feasibility of the operation and maintenance strategy are improved.

[0052] In conjunction with some embodiments of the first aspect, in some embodiments, converting the optimal operation and maintenance strategy set into a stage implementation plan specifically includes:

[0053] The optimal operation and maintenance strategy set is divided into short-term maintenance tasks, mid-term transformation tasks and long-term update tasks according to the execution sequence;

[0054] Prioritize tasks at each stage based on equipment importance, cost-effectiveness, and implementation difficulty;

[0055] Develop task execution plans that take into account resource constraints and time windows;

[0056] Set evaluation indicators for the implementation effect of tasks at each stage;

[0057] Establish a mechanism to link tasks between tasks at each stage;

[0058] Based on the task execution plan, implementation effect evaluation indicators and linkage mechanism, a phased implementation plan including specific implementation steps is generated.

[0059] By adopting the above technical solutions, a task execution plan that takes into account resource constraints and time windows is formulated to make the operation and maintenance work operational. By setting the implementation effect evaluation indicators of each stage of the task, the completion of the task can be objectively evaluated. The connection mechanism between the tasks of each stage is established to improve the continuity and coordination of the operation and maintenance work. Based on the task execution plan, the implementation effect evaluation indicators and the connection mechanism, a stage implementation plan containing specific implementation steps is generated, making the operation and maintenance work more standardized and standardized, and improving the planning and execution efficiency of the operation and maintenance work.

[0060] In the second aspect, an embodiment of the present application provides an operation and maintenance cost prediction system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to enable the system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0061] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a system, causes the system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0062] In a fourth aspect, an embodiment of the present application provides a computer program product, characterized in that when the computer program product runs on a system, the system executes the method described in any possible implementation manner in the first aspect.

[0063] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0064] 1. The present application provides a method for predicting operation and maintenance costs. By obtaining historical data containing energy consumption data, equipment operation and maintenance records, environmental data, and cost records within a preset time period, and dividing them into high and low cost data groups according to the size of the cost records, the abnormal cost interval can be clearly identified. The high-cost data group is traced back in time to obtain the preceding symptom sequence, similar feature fragments are retrieved in the low-cost data group and a bifurcation point data table is generated to establish a mapping relationship between the key nodes of cost evolution and the system operation status. The feature distance matrix is ​​calculated from the real-time operation data and the bifurcation point data table and the probability weight is determined, so that the cost evolution trend of the system can be accurately predicted. By generating multiple cost evolution branches and performing weighted calculations based on the degree of fit with historical data, the possibility of a single evolution path is considered, and the uncertainty assessment of multi-scenario prediction is introduced, which improves the reliability of the prediction results, makes the prediction results closer to the actual operation conditions, reduces the deviation between the prediction results and the actual operation and maintenance costs, and thus improves the accuracy of the operation and maintenance cost prediction.

[0065] 2. The present application provides an operation and maintenance cost forecasting method, which decomposes the operation and maintenance cost forecasting results into trend items, cycle items and random items through time series decomposition, thereby realizing a refined analysis of the cost change characteristics. The cost growth acceleration index is constructed based on the change slope of the trend item, and the cost anomaly degree score is calculated in combination with the fluctuation characteristics of the cycle item. The entropy method is used to evaluate the uncertainty level of the random item, and a multi-dimensional cost change feature characterization system is established. By integrating the acceleration index, the anomaly degree score and the uncertainty level to generate a warning threshold curve, a dynamic warning of cost anomalies is realized. This method constructs a warning mechanism from the three dimensions of the speed, volatility and uncertainty of cost changes, improves the sensitivity and reliability of the warning, and reduces the false alarms and omissions of the warning.

[0066] 3. The present application provides a method for predicting operation and maintenance costs, which extracts spatiotemporal evolution patterns based on the dynamic features of the core node group, and constructs a deep representation learning framework containing multiple attention mechanisms, thereby achieving an accurate characterization of the dynamic characteristics of the system. The deep representation learning framework is used to dynamically simulate the response characteristics under different operation and maintenance strategies, and the effects of various operation and maintenance strategies can be evaluated in advance. By constructing a multi-objective optimization problem with nonlinear coupling constraints and solving it with an improved hybrid intelligent algorithm, the optimal set of operation and maintenance strategies under multiple objectives is obtained, which comprehensively considers the complexity, dynamics and multi-objective nature of the system. Through the combination of deep learning and intelligent optimization algorithms, the scientificity and feasibility of the operation and maintenance strategy are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flow chart of a method for predicting operation and maintenance costs in an embodiment of the present application.

[0068] Figure 2It is another flow chart of an operation and maintenance cost prediction method in an embodiment of the present application.

[0069] Figure 3 It is a schematic diagram of the physical device structure of an operation and maintenance cost prediction system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0070] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more listed items.

[0071] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0072] The following uses an embodiment and combines Figure 1 , a method for predicting operation and maintenance costs in an embodiment of the present application is described:

[0073] See also Figure 1 , which is a flow chart of an operation and maintenance cost prediction method in an embodiment of the present application.

[0074] S101, obtaining historical data of energy system operation within a preset time period, and dividing the historical data into a high-cost data group and a low-cost data group according to the value of the cost record;

[0075] The system obtains the historical data of energy system operation within a preset time period, and divides the historical data into high-cost data group and low-cost data group according to the numerical value of the cost record. The historical data includes energy consumption data, equipment operation and maintenance records, environmental data and cost records.

[0076] In this step, the system first needs to obtain the historical operation data of the energy system within the preset time period, which includes energy consumption data, equipment operation and maintenance records, environmental data, and cost records. Among them, energy consumption data can be the consumption of electricity, gas, water and other energy at various time points; equipment operation and maintenance records can be the operating status of the equipment, maintenance records, etc.; environmental data can be external environmental parameters such as temperature, humidity, and air pressure; and cost records are various cost expenditures related to energy consumption, equipment operation and maintenance, etc.

[0077] After acquiring the historical data, the system needs to divide the historical data into high-cost data groups and low-cost data groups according to the numerical values ​​of the expense records. In specific implementation, the system can set a cost threshold, and divide the data with expense records above the threshold into the high-cost data group, and divide the data with expense records below the threshold into the low-cost data group. The setting of the threshold can be obtained by statistical analysis based on the cost distribution of historical data. In addition, the system can also consider using a clustering algorithm to automatically divide the historical data, and automatically classify data with similar expense records into the same data group.

[0078] S102, performing time series backtracking based on the expense records of the high-expense data group to obtain a sequence of premonitions of high-expense events;

[0079] The system performs time series backtracking based on the cost records of the high-cost data group to obtain the preceding symptom sequence of high-cost events. Specifically: the cost change rate at adjacent moments in the high-cost data group is calculated to identify the cost mutation point; a reverse tracing window is set based on the cost mutation point, and the time series characteristics of environmental data, energy consumption data and equipment operation and maintenance records are extracted in the reverse tracing window; the fluctuation amplitude of the time series characteristics is counted, and the characteristic parameters with fluctuation amplitude greater than the preset fluctuation amplitude are marked; a time series association map is constructed based on the characteristic parameters, and the characteristic sequence of the cost mutation event in the time series association map is extracted. The cost mutation event is a cost event that meets the requirement of a duration greater than the preset time and a cost peak value exceeding the preset cost value; the characteristic sequences of different cost mutation events are compared and analyzed, and the repeated characteristic sequences are used as the preceding symptom sequence.

[0080] In this step, the system needs to start with the high-cost data group, conduct a time-series retrospective analysis based on the cost records, and find out the pre-signs that lead to the high-cost event. First, the system needs to calculate the cost change rate at adjacent moments in the high-cost data group, and identify the cost mutation point by setting the change rate threshold. The mutation point is the moment when the cost changes sharply.

[0081] Then, the system will set a backward tracing time window based on the mutation point, and extract time series features such as environmental data, energy consumption data, and equipment operation and maintenance records within this window. Next, the system needs to count the fluctuation range of each time series feature and find out the characteristic parameters whose fluctuation range exceeds the preset threshold. These characteristic parameters may be associated with the cost mutation.

[0082] After obtaining the key characteristic parameters, the system will build a time series correlation map based on these parameters to examine the mutual influence and transmission relationship between different parameters in the time series. By analyzing the correlation map, the system can extract the typical characteristic sequence of high-cost events. The screening principle is that the duration of the event exceeds the preset time and the cost peak exceeds the preset value. The characteristic sequences of high-cost events are compared, and the repeated and highly consistent characteristic sequences are used as common pre-signs of high-cost events.

[0083] S103, searching the low-cost data group for feature segments whose similarity to the preceding symptom sequence is greater than a preset threshold;

[0084] In this step, the system needs to perform similarity matching between the pre-symptom sequence of high-cost events obtained in the previous step and the historical data of the low-cost data group, and retrieve feature fragments with high similarity to the pre-symptom sequence and exceeding the preset threshold. These feature fragments similar to high-cost events are likely to contain precursor information of the evolution from a low-cost state to a high-cost state.

[0085] In specific implementation, the system can use a sequence alignment algorithm to measure the similarity between the pre-significant sequence and the low-cost data. Common sequence similarity measurement methods include Euclidean distance, DTW dynamic time warping distance, longest common subsequence, etc. The system can calculate the similarity between the pre-significant sequence and each subsequence fragment in the low-cost data, and select fragments with a similarity higher than a preset threshold (such as 90%) as potential feature fragments.

[0086] At the same time, since the amount of data in the low-cost data group is usually large, in order to improve the retrieval efficiency, the system can first pre-process the low-cost data, such as using time series representation methods such as PAA segmented aggregation approximation and SAX symbolic approximation to convert the original time series data into a shorter high-level representation, and then perform similarity matching on the high-level representation to quickly lock suspicious data fragments, and then return to the original data for detailed comparison, thereby reducing the computing cost while ensuring the retrieval quality.

[0087] S104, generating a bifurcation point data table of cost evolution based on the characteristic fragments, wherein the bifurcation point data table includes critical conditions for cost changes;

[0088] After obtaining the characteristic fragments in the low-cost state in the previous step, in this step the system needs to analyze the subsequent evolution of these fragments, find out the key nodes and critical conditions that cause the cost state to change from low to high, and generate a bifurcation point data table of cost evolution.

[0089] The system can track each feature segment for a period of time (such as 1-2 cost cycles) to observe its subsequent cost change trend. For those segments where the cost increases significantly, it means that they are likely to be bifurcation points from low-cost states to high-cost states. The system should focus on and extract various parameter indicators of bifurcation points, such as ambient temperature and humidity, energy consumption, equipment status, etc., to find out their critical change ranges.

[0090] These critical change ranges can form a multi-dimensional parameter boundary surface. When the real-time operating parameters fall within the boundary surface, it indicates that the current state of the system has a high probability of changing from low cost to high cost. Therefore, the system summarizes the critical conditions of the parameters of each bifurcation point into a data table as a basis for judging the cost evolution trend. Each item in the bifurcation point data table corresponds to a fork in the road where the system enters a different cost path from the current state.

[0091] S105, performing characteristic distance calculation on the real-time acquired operation data and the bifurcation point data table to obtain a characteristic distance matrix;

[0092] In this step, the system needs to collect the operating parameter data of the energy system in real time, such as environmental sensor data, equipment operating status data, energy consumption metering data, etc., and compare the real-time data with the bifurcation point data table generated in the previous step, calculate the characteristic distance between the real-time operating status and each bifurcation point, and obtain a characteristic distance matrix.

[0093] In specific implementation, the system can use a variety of distance metrics to calculate feature distances, the most common of which are Euclidean distance, Manhattan distance, cosine similarity, etc. For different types of parameter indicators, the system can use different distance metric formulas. For example, Euclidean distance can be used for continuous numerical parameters, and Hamming distance can be used for discrete classification parameters. The system needs to calculate the distance values ​​of each operating parameter and the corresponding parameters in the bifurcation point data table, and then perform weighted summation of the distance values ​​of multiple parameters to obtain the comprehensive feature distance.

[0094] By calculating the characteristic distance between the real-time operating state and each bifurcation point one by one, the system can finally obtain an N×M-dimensional distance matrix, where N is the number of bifurcation points and M is the number of types of operating parameters. Each element d(i, j) in the distance matrix indicates the distance between the current operating state and the i-th bifurcation point on the j-th parameter. The closer the distance, the higher the similarity between the current state and the bifurcation point, and the future cost is likely to develop along the subsequent evolution path of the bifurcation point.

[0095] S106, calculating the probability weights of the real-time acquired operation data on different cost evolution paths in the bifurcation point data table based on the characteristic distance matrix;

[0096] In this step, the system needs to use the characteristic distance matrix obtained in the previous step to further calculate the probability of the real-time running state on the subsequent cost path at each bifurcation point, and obtain a probability weight distribution. This probability weight represents the relative probability of the current state of the system entering a high-cost or low-cost state in the future.

[0097] In specific implementation, the system can convert the distance value into a probability weight based on the distance value of each bifurcation point in the characteristic distance matrix, using probabilistic methods such as the soft-max function. After the soft-max transformation, the original distance value is mapped to the interval [0, 1] and converted into a set of normalized probability weight vectors, and the sum of the weights of all bifurcation points is equal to 1. The higher the probability weight, the closer the current state is to the bifurcation point, and the greater the possibility of evolving along the cost path in the future.

[0098] At the same time, due to the random fluctuations in system operation, the state at a single moment may be affected by accidental disturbances. In order to improve the robustness of the prediction, the system can continuously sample the operating data within a time window (such as 1 hour), calculate the probability weight vector at each moment in the time window, and then perform average weighted processing to obtain a comprehensive time period probability weight distribution, which can more stably reflect the overall operation trend of the system within that period of time.

[0099] In addition, since the evolution of the actual cost path has a cumulative effect, the system can also consider introducing time series probability models such as Markov chains, which not only consider the probability weight of the current state, but also combine the state transition sequence over a period of history to make time series correlation predictions for subsequent cost paths, capture the long-term trend and cumulative law of cost evolution, make full use of time series information in weight calculations, and improve the overall accuracy of predictions.

[0100] S107, generating several cost evolution branches within the forecast period according to the probability weight, and calculating the degree of consistency between each cost evolution branch and the historical data;

[0101] After obtaining the probability weight distribution of the real-time operating status on the cost path of each bifurcation point, in this step the system needs to generate multiple future cost evolution scenarios based on the probability weights, and then compare these scenarios with historical data to calculate their degree of consistency with the actual evolution laws.

[0102] In specific implementation, the system can use Monte Carlo simulation, random walk and other methods to perform multiple random samplings within the forecast time period (such as the next week) according to the probability weight distribution to generate several possible cost evolution paths. Each path is a Markov chain formed by jumping with a certain probability on the cost path at different bifurcation points with the current state as the starting point. Through multiple sampling simulations, the system can obtain a set of cost prediction branches covering the main evolution scenarios in the future.

[0103] Then, the system needs to match these predicted branches with the actual cost paths in the historical data, and calculate the degree of fit between each branch and the actual situation. An intuitive similarity measurement method is to calculate the correlation coefficient between the predicted path and the actual path. The higher the correlation coefficient, the closer the predicted branch is to the actual evolution law. At the same time, the system can also use the dynamic time warping (DTW) algorithm to calculate the best match between two time series, which can allow a certain degree of expansion and contraction of the path on the time axis to capture more flexible similarities.

[0104] By calculating the degree of fit with historical data, the system can evaluate the rationality and reliability of each prediction branch and find the evolution scenario that is closest to the actual situation. However, since historical data may also have special cases and abnormal situations, simply pursuing the degree of fit with historical data may bring the risk of overfitting. Therefore, the system can also introduce penalty items to appropriately suppress overly complex and extreme prediction branches, balance the fitting ability and generalization ability of the prediction model, and improve the robustness of prediction branch selection.

[0105] S108. Perform weighted calculation on each cost evolution branch based on the degree of fit to obtain an operation and maintenance cost prediction result.

[0106] The system performs weighted calculation on each cost evolution branch based on the degree of fit to obtain the operation and maintenance cost prediction result. Specifically: the Euclidean distance and dynamic time warping distance between each cost evolution branch and the historical data are calculated; a comprehensive similarity evaluation index is constructed based on the Euclidean distance and the dynamic time warping distance; the comprehensive similarity evaluation index is normalized to obtain the weight coefficient of each cost evolution branch; based on the weight coefficient, the predicted value of each cost evolution branch is weighted averaged to obtain the operation and maintenance cost prediction result.

[0107] After evaluating the consistency of each prediction branch with historical data in the previous step, the system needs to combine the consistency and occurrence probability of each branch in this step to give a comprehensive operation and maintenance cost prediction result. The prediction result should consider both the rationality of each branch and the possibility of its occurrence.

[0108] In specific implementation, the system can first normalize the degree of fit of each prediction branch and map the degree of fit to the interval [0, 1] as the weight coefficient of the branch reliability. Then, the system multiplies the reliability weight of each branch by its probability weight to obtain the comprehensive weight of the branch. This comprehensive weight takes into account both the possibility and rationality of the branch.

[0109] Next, the system needs to estimate the cost value of each branch within the forecast period. It can use historical data fitting, trend extrapolation and other methods to model and predict the subsequent cost trend of the branch to obtain the cost forecast value sequence of each branch. By multiplying the cost forecast value sequence of each branch by its comprehensive weight and then summing them up, a weighted average operation and maintenance cost forecast result can be obtained.

[0110] This comprehensive forecast result can more comprehensively reflect the overall cost level under various possible scenarios, and is neither too radical nor too conservative. However, in actual applications, the system can also make appropriate adjustments to the forecast results based on the decision maker's risk preferences and management needs. For example, if the decision maker pays more attention to cost control and hopes to take cost-saving measures as soon as possible, the weight of the high-cost branch can be appropriately increased to give a relatively conservative forecast result; on the contrary, if the decision maker pays more attention to operating income and hopes to seize possible profit opportunities, the weight of the low-cost branch can be appropriately increased to give a relatively optimistic forecast result.

[0111] In addition, since the operation and maintenance environment of the energy system may have external shocks and uncertain disturbances, causing the actual cost evolution to deviate from expectations, the system can provide a certain confidence interval and variability analysis while giving the cost forecast results to evaluate the uncertainty level of the forecast results. Decision makers can weigh the risks and benefits of operation and maintenance cost management based on the width of the confidence interval and the size of the variability, and formulate corresponding flexible response strategies.

[0112] In the above embodiment, by obtaining historical data containing energy consumption data, equipment operation and maintenance records, environmental data and cost records within a preset time period, and dividing it into high and low cost data groups according to the size of the cost records, the abnormal cost interval can be clearly identified. The high cost data group is traced back in time to obtain the pre-symptom sequence, and similar feature fragments are retrieved in the low cost data group and a bifurcation point data table is generated to establish a mapping relationship between the key nodes of cost evolution and the system operation status. The feature distance matrix is ​​calculated from the real-time operation data and the bifurcation point data table and the probability weight is determined, so that the cost evolution trend of the system can be accurately predicted. By generating multiple cost evolution branches and performing weighted calculations based on the degree of fit with historical data, the possibility of a single evolution path is considered, and the uncertainty assessment of multi-scenario prediction is introduced, which improves the reliability of the prediction results, makes the prediction results closer to the actual operation situation, reduces the deviation between the prediction results and the actual operation and maintenance costs, and thus improves the accuracy of the operation and maintenance cost prediction.

[0113] Combine the following Figure 2 , another operation and maintenance cost prediction method in the embodiment of the present application is described:

[0114] See also Figure 2 , is another flow chart of an operation and maintenance cost prediction method in an embodiment of the present application.

[0115] S201, extracting trend items, period items and random items from the operation and maintenance cost forecast results using a time series decomposition method;

[0116] In this step, the system uses the time series decomposition method to decompose the operation and maintenance cost forecast results into three parts: trend item, cycle item and random item. Through decomposition, we can better understand the changing rules and internal factors of operation and maintenance costs. The trend item reflects the overall change trend of operation and maintenance costs over time, the cycle item reflects the cyclical fluctuation pattern in operation and maintenance costs, and the random item reflects the uncertainty fluctuation caused by accidental factors.

[0117] The system can use classic time series decomposition methods such as moving average method and exponential smoothing method, or more advanced decomposition techniques such as STL decomposition (Seasonal-Trend decomposition using Loess) and X-11 decomposition. These methods extract the change characteristics on different time scales by smoothing the original time series data. For example, STL decomposition uses LOESS smoothing method to estimate trend term and cycle term respectively, and removes them from the original data to obtain random term.

[0118] In actual applications, due to the complexity of operation and maintenance scenarios, the original cost data may contain noise such as outliers and missing values. In order to improve the decomposition accuracy, the system can perform preprocessing operations such as anomaly detection and interpolation repair on the data before time series decomposition. In addition, for the periodicity of different time scales, such as monthly, quarterly, and annual, the system can adopt a multi-scale fusion strategy to integrate the decomposition results and enhance the expression of trend items, periodic items, and random items.

[0119] S202. constructing a cost growth acceleration indicator based on the change slope of the trend item;

[0120] In this step, the system uses the changing characteristics of the trend item to construct the cost growth acceleration index to quantify the dynamic changes in the growth of operation and maintenance costs. The cost growth acceleration reflects the rate of change of the cost growth rate and is an important basis for judging whether the cost growth trend is intensifying or slowing down. Through this indicator, the cost change trend can be captured more sensitively, providing a reference for subsequent early warning and decision-making.

[0121] The system can use the method of numerical differentiation to calculate the first-order and second-order derivatives of the trend term between adjacent time points, which respectively characterize the speed and acceleration of cost growth. Considering the discreteness of cost data, the differential quotient can be used instead of the differential quotient for calculation. In order to reduce the impact of local fluctuations, the system can also perform smoothing operations such as local polynomial regression or spline interpolation on the trend term before differentiation.

[0122] In practice, the construction of the cost growth acceleration indicator needs to take into account both timeliness and stability. On the one hand, the indicator needs to reflect the cost change trend in a timely manner, and on the other hand, it should avoid being too sensitive and causing frequent shocks. To this end, the system can introduce a time window and threshold mechanism to evaluate the acceleration level within a certain time range and set a reasonable threshold to filter noise. At the same time, comparing the acceleration indicator at the current moment with the historical level at the same period can also help determine whether the current change is abnormal.

[0123] S203. Calculate the cost abnormality score based on the fluctuation characteristics of the periodic items;

[0124] In this step, the system calculates the cost anomaly score by analyzing the fluctuation characteristics of the periodic items to measure the abnormal level of the periodic fluctuation of the operation and maintenance costs. The higher the anomaly score, the greater the deviation of the current cost change from the normal cycle law, and the higher the potential risk. By introducing the anomaly score, the cyclical anomaly of the cost can be quantitatively characterized to assist managers in quickly identifying risk points.

[0125] The system can use statistical methods, such as moving z-score, moving MAD (Median Absolute Deviation), etc., to measure the degree of deviation of the current period item value from the historical normal range. These methods compare the difference between the current value and the historical mean or median, and standardize it with the dispersion of historical fluctuations (such as standard deviation, MAD) to obtain an abnormal score. The larger the score, the more serious the deviation of the current value from the normal level.

[0126] S204. Use entropy method to evaluate the uncertainty level of random items;

[0127] This step uses the entropy method to quantitatively evaluate the uncertainty level of random items and characterize the volatility risk caused by random factors in operation and maintenance costs. Entropy is a physical quantity that describes the degree of disorder in a system. The larger the entropy value, the higher the uncertainty of the system. By calculating the entropy value of random items, the unpredictability of the random components in the cost can be measured, providing a decision-making basis for risk management.

[0128] The system can estimate the probability distribution of a random item sequence based on probability theory and information theory, and calculate information entropy from it. Common entropy estimation methods include histogram method, kernel density estimation method, maximum entropy method, etc. Among them, the histogram method divides the data into intervals of equal width, counts the frequency of each interval, and then calculates the entropy value from the frequency; the kernel density estimation method uses a kernel function to fit the probability density of the data, and then estimates the entropy value; the maximum entropy rule finds the probability distribution with the largest entropy value under the condition of satisfying known constraints.

[0129] S205, integrating the acceleration index, the abnormality score and the uncertainty level to generate a warning threshold curve;

[0130] In this step, the system comprehensively considers the information of three dimensions: the acceleration index of cost growth, the abnormality score, and the uncertainty level, and generates a dynamic warning threshold curve. The warning threshold represents the critical condition for triggering the warning, and the threshold curve depicts the continuous change of the threshold over time. By integrating multiple risk indicators to construct a threshold curve, risk signals in costs can be captured more comprehensively and accurately.

[0131] The system can use machine learning methods such as weighted average, logistic regression, and support vector machines to map the three indicators to the same scale, and combine business knowledge and historical data to learn indicator weights and threshold functions. In this process, the importance of different indicators can be adjusted according to the manager's risk preference. For example, risk-averse managers can increase the weight of abnormality scores to identify potential risks earlier. The form of the threshold function can be linear or nonlinear to adapt to complex patterns of cost changes.

[0132] S206. Perform graded warning based on the warning threshold curve.

[0133] The system performs graded warning based on the warning threshold curve. Specifically: it constructs a multidimensional warning feature space that includes cost mutation risk, equipment status and environmental impact; uses the kernel principal component analysis method to reduce the dimension of the multidimensional warning feature space to obtain reduced-dimensional multidimensional warning features; constructs a state transfer matrix based on the reduced-dimensional multidimensional warning features; and generates a dynamic warning response plan containing multi-level warning information based on the state transfer matrix.

[0134] In the last step, the system will issue a graded warning for the current expense risk status based on the generated warning threshold curve. By setting multiple threshold breakpoints, risks can be divided into multiple levels (such as low risk, medium risk, high risk, etc.) to characterize the severity of the risk. Graded warnings help managers quickly understand the current risk level and take corresponding control measures.

[0135] The system can monitor cost indicators in real time. When the indicator value exceeds a certain breakpoint of the threshold curve, it triggers the corresponding level of early warning signal and notifies the manager. Early warning information can be issued through visual dashboards, emails, text messages and other channels to ensure timely response. In addition to the alarm of the current risk level, the system can also display the historical change trend of risk indicators, year-on-year and month-on-month analysis, etc., to help managers comprehensively diagnose the causes of risks.

[0136] In operation and maintenance practice, it is crucial to adopt differentiated control strategies based on risk levels. The system can preset disposal plans corresponding to different risk levels, such as strengthening resource monitoring, controlling upstream traffic, dynamically adjusting capacity, etc., to achieve automated closed-loop disposal of risks. At the same time, the system can also continuously track the disposal effects of early warnings, optimize the graded early warning model and disposal strategies, and form a virtuous cycle of risk management.

[0137] The system performs graded warning based on the warning threshold curve. Specifically: it constructs a multidimensional warning feature space that includes cost mutation risk, equipment status and environmental impact; uses the kernel principal component analysis method to reduce the dimension of the multidimensional warning feature space to obtain reduced-dimensional multidimensional warning features; constructs a state transfer matrix based on the reduced-dimensional multidimensional warning features; and generates a dynamic warning response plan containing multi-level warning information based on the state transfer matrix.

[0138] On the basis of hierarchical warning, the system further constructs a multi-dimensional warning feature space, comprehensively considering factors in multiple dimensions such as cost mutation risk, equipment status and environmental impact. Through dimensionality reduction methods such as kernel principal component analysis, the system can extract the core pattern of the warning feature space, reveal the intrinsic relationship between different risk factors, and simplify the complexity of the warning model. On this basis, the system uses the warning features after dimensionality reduction to learn the state transition law, characterize the dynamic evolution relationship between risk levels, and generate a state transfer matrix. Based on the transfer matrix, the system can dynamically deduce the transmission path of risks between different levels and estimate the distribution of risk states in the future.

[0139] Finally, the system combines multi-level risk warning information and state transition prediction to generate a dynamic warning response plan. According to the severity of the risk, development trend, and real-time status of the business, the system dynamically adjusts the warning strategy and formulates differentiated disposal measures. For example, for initial-level risks, the system can focus on early prevention of problems, while for high-level risks, the system needs to initiate emergency plans to control further deterioration of risks. Based on real-time feedback data, the system can also continuously evaluate the effectiveness of warning responses, optimize disposal plans, and achieve precise risk management.

[0140] In the above embodiment, the operation and maintenance cost forecast results are decomposed into trend items, cycle items and random items through time series decomposition, thereby realizing a refined analysis of the cost change characteristics. Based on the change slope of the trend item, the cost growth acceleration index is constructed, and the cost anomaly degree score is calculated in combination with the fluctuation characteristics of the cycle item. The entropy method is used to evaluate the uncertainty level of the random item, and a multi-dimensional cost change feature characterization system is established. By integrating the acceleration index, the anomaly degree score and the uncertainty level to generate a warning threshold curve, a dynamic warning of cost anomalies is realized. This method constructs a warning mechanism from the three dimensions of the speed, volatility and uncertainty of cost changes, improves the sensitivity and reliability of the warning, and reduces the false alarms and omissions of the warning.

[0141] Further, in another embodiment, after performing graded warning based on the warning threshold curve, the system constructs a high-dimensional heterogeneous complex network and identifies key influencing paths within the high-dimensional heterogeneous complex network;

[0142] Based on the dynamic feature extraction and spatiotemporal evolution pattern of the core node group, a deep representation learning framework with multiple attention mechanisms is constructed;

[0143] Use the deep representation learning framework to dynamically simulate the response characteristics under different operation and maintenance strategies;

[0144] Based on the response characteristics, a multi-objective optimization problem with nonlinear coupling constraints is constructed, and an improved hybrid intelligent algorithm is used to solve it and obtain the solution;

[0145] Generate the optimal operation and maintenance strategy set based on the solution results;

[0146] Convert the optimal operation and maintenance strategy set into a phased implementation plan. Specifically, divide the optimal operation and maintenance strategy set into short-term maintenance tasks, mid-term transformation tasks, and long-term update tasks according to the execution sequence;

[0147] Prioritize tasks at each stage based on equipment importance, cost-effectiveness, and implementation difficulty;

[0148] Develop task execution plans that take into account resource constraints and time windows;

[0149] Set evaluation indicators for the implementation effect of tasks at each stage;

[0150] Establish a mechanism to link tasks between tasks at each stage;

[0151] Based on the task execution plan, implementation effect evaluation indicators and linkage mechanism, a phased implementation plan including specific implementation steps is generated.

[0152] The system abstracts the operation and maintenance scenario into a complex network composed of multi-level, multi-type nodes and connections, and depicts the interactive effects between factors such as equipment, environment, and cost. Based on the network structure and dynamic characteristics, the system uses algorithms such as community discovery and influence propagation to identify the key paths of risk transmission and reveal the global effect of risk. The system focuses on the core node group in the complex network and captures its dynamic change rules in time and space dimensions. By designing a multiple attention mechanism, the system can adaptively focus on key factors at different time and space scales and learn the deep representation of risk evolution. This deep learning framework can enhance the intelligence of risk analysis and prediction. For candidate operation and maintenance strategies, the system converts them into the evolution rules of complex networks and simulates the dynamic changes of network status after the implementation of the strategy. The system comprehensively considers multiple objectives such as risk control, cost-effectiveness, and resource constraints, and constructs an optimization model with nonlinear coupling constraints. For this complex combinatorial optimization problem, the system can use intelligent algorithms such as heuristic search and evolutionary computing, and introduce improvement mechanisms such as problem decomposition and multi-population collaboration to efficiently solve the optimal operation and maintenance strategy set. According to the time scale and implementation requirements of the strategy, the system divides it into phase tasks such as short-term maintenance, mid-term transformation, and long-term update. The system prioritizes tasks at each stage by taking into account factors such as equipment importance, cost-effectiveness, and implementation difficulty. At the same time, considering resource constraints and time windows, the system formulates a detailed task execution plan and sets phased effect evaluation indicators. In addition, the system also needs to build a linkage mechanism between tasks to ensure the continuity and consistency of strategy implementation.

[0153] In the above embodiment, the spatiotemporal evolution pattern is extracted based on the dynamic features of the core node group, and a deep representation learning framework including multiple attention mechanisms is constructed to achieve accurate characterization of the dynamic characteristics of the system. The response characteristics under different operation and maintenance strategies are dynamically simulated using the deep representation learning framework, and the effects of various operation and maintenance strategies can be pre-evaluated. By constructing a multi-objective optimization problem with nonlinear coupling constraints and solving it using an improved hybrid intelligent algorithm, the optimal operation and maintenance strategy set under multiple objectives is obtained, which comprehensively considers the complexity, dynamics and multi-objective nature of the system. Through the combination of deep learning and intelligent optimization algorithms, the scientificity and feasibility of the operation and maintenance strategy are improved.

[0154] The following describes the system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of the physical device structure of an operation and maintenance cost prediction system provided in an embodiment of the present application.

[0155] It should be noted that Figure 3 The structure of the system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0156] like Figure 3 As shown, the system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0157] The following components are connected to the I / O interface 305: an input section 306 including a camera, an infrared sensor, etc.; an output section 307 including a liquid crystal display (LCD) and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0158] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0159] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.

[0160] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0161] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the system described in the above embodiment; or may exist independently without being assembled into the system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of a system, the system implements the method provided in the above embodiment.

[0162] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0163] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0164] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0165] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A method for predicting operation and maintenance costs, characterized in that: include: Obtaining historical data of energy system operation within a preset time period, the historical data including energy consumption data, equipment operation and maintenance records, environmental data and cost records; Dividing the historical data into a high-cost data group and a low-cost data group according to the numerical value of the cost record; Based on the expense records of the high-expense data group, a time series backtracking is performed to obtain a sequence of premonitions of high-expense events, specifically including: Calculating the cost change rate at adjacent moments in the high-cost data group to identify cost mutation points; A reverse tracing window is set based on the cost mutation point, and time series characteristics of environmental data, energy consumption data and equipment operation and maintenance records are extracted within the reverse tracing window; Counting the fluctuation amplitude of the time series feature, and marking the feature parameters whose fluctuation amplitude is greater than the preset fluctuation amplitude; A time series association graph is constructed based on the characteristic parameters, and a characteristic sequence of a cost mutation event in the time series association graph is extracted, wherein the cost mutation event is a cost event that has a duration greater than a preset time and a cost peak value greater than a preset cost value; Compare and analyze the characteristic sequences of different cost mutation events, and use the repeated characteristic sequences as the preceding sign sequences; Retrieving a feature segment having a similarity with the preceding symptom sequence greater than a preset threshold value in the low-cost data set; Generate a bifurcation point data table of cost evolution based on the characteristic fragment, wherein the bifurcation point data table includes critical conditions for cost change; Performing characteristic distance calculation on the real-time acquired operation data and the bifurcation point data table to obtain a characteristic distance matrix; Calculate the probability weights of the real-time acquired operation data on different cost evolution paths in the bifurcation point data table based on the characteristic distance matrix; Generating a number of cost evolution branches within the forecast time period according to the probability weight; calculating the degree of consistency between each of the cost evolution branches and the historical data; Based on the degree of fit, weighted calculation is performed on each of the cost evolution branches to obtain an operation and maintenance cost prediction result.

2. The method according to claim 1, characterized in that The weighted calculation of each cost evolution branch based on the degree of agreement is performed to obtain an operation and maintenance cost prediction result, which specifically includes: Calculating the Euclidean distance and the dynamic time warping distance between each of the cost evolution branches and the historical data; constructing a comprehensive similarity evaluation index based on the Euclidean distance and the dynamic time warping distance; Normalizing the comprehensive similarity evaluation index to obtain a weight coefficient of each cost evolution branch; The predicted values ​​of each of the cost evolution branches are weighted averaged based on the weight coefficient to obtain an operation and maintenance cost prediction result.

3. The method according to claim 1, characterized in that After performing weighted calculation on each of the cost evolution branches based on the degree of agreement to obtain an operation and maintenance cost prediction result, the method further includes: Using a time series decomposition method to extract trend items, period items and random items from the operation and maintenance cost forecast results; Constructing a cost growth acceleration index based on the change slope of the trend item; calculating a cost anomaly degree score based on the fluctuation characteristics of the cycle item; An entropy method is used to evaluate the uncertainty level of the random term; Fusion of the acceleration index, the abnormality score and the uncertainty level to generate a warning threshold curve; A graded warning is performed based on the warning threshold curve.

4. The method according to claim 3, characterized in that The performing of graded warning based on the warning threshold curve specifically includes: Construct a multi-dimensional early warning feature space that includes cost mutation risk, equipment status, and environmental impact; The multidimensional warning feature space is processed by dimensionality reduction using a kernel principal component analysis method to obtain a reduced-dimensional multidimensional warning feature; a state transfer matrix is ​​constructed based on the reduced-dimensional multidimensional warning feature; A dynamic early warning response plan including multi-level early warning information is generated based on the state transition matrix.

5. The method according to claim 3, characterized in that: After performing a graded warning based on the warning threshold curve, the method further includes: Constructing a high-dimensional heterogeneous complex network and identifying key influencing paths within the high-dimensional heterogeneous complex network; Based on the dynamic feature extraction and spatiotemporal evolution pattern of the core node group, a deep representation learning framework with multiple attention mechanisms is constructed; Using the deep representation learning framework to dynamically simulate the response characteristics under different operation and maintenance strategies; Based on the response characteristics, a multi-objective optimization problem with nonlinear coupling constraints is constructed, and an improved hybrid intelligent algorithm is used to solve the problem to obtain a solution result; Generate an optimal operation and maintenance strategy set according to the solution result; The optimal operation and maintenance strategy set is converted into a phased implementation plan.

6. The method according to claim 5, characterized in that The converting the optimal operation and maintenance strategy set into a phase implementation plan specifically includes: Dividing the optimal operation and maintenance strategy set into short-term maintenance tasks, mid-term transformation tasks and long-term update tasks according to the execution sequence; Prioritize tasks at each stage based on equipment importance, cost-effectiveness, and implementation difficulty; develop a task execution plan that takes into account resource constraints and time windows; Setting evaluation indicators for the implementation effect of the tasks at each stage; building a linkage mechanism between the tasks at each stage; A phase implementation plan including specific implementation steps is generated based on the task execution plan, the implementation effect evaluation indicators and the connection mechanism.

7. An operation and maintenance cost prediction system, characterized in that: The system comprises: One or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the system to execute the method as described in any one of claims 1-6.

8. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a system, the system is caused to execute the method according to any one of claims 1 to 6.

9. A computer program product, characterized in that When the computer program product is executed on a system, the system is caused to execute the method according to any one of claims 1 to 6.

Images