An ai-based heat station difference analysis method and system

Abstract

The present application relates to the technical field of data analysis, in particular to a heat station difference analysis method and system based on AI, the present application obtains a large amount of historical operation data of heat stations, sampling room temperature and the like, and then cleans the data to remove missing or invalid data. The difference between the value of the heat supply parameter and the reference value when the heat station is running in various situations is utilized, and by analyzing these differences, the heat supply parameter related to the health degree, i.e. the evaluation parameter, is found. Then, based on the historical samples of the evaluation parameter, a health degree influence model is constructed, so as to perform health degree evaluation for the training data, introduce the influence of the health degree dimension on the value of the heat supply parameter, and train the model. According to the trained model, the value of the heat supply parameter is automatically inferred, and the predicted value is used to compensate different heat supply stations, that is, the operation parameter compensation can be performed according to the actual situation of different heat supply stations. The problem that the heat supply parameter of the heat station is filled in manually, which is unreliable and ignores the influence of the health degree, is avoided.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of data analysis, and particularly relates to an AI-based heat station difference analysis method and system. BACKGROUND

[0002] With the acceleration of urbanization, the centralized heating scale in cities is continuously expanded. Many large and medium-sized cities have built a relatively large heating pipe network system, which covers a large number of residential areas, commercial areas and public buildings. The popularization rate of centralized heating is generally high in northern cities, which effectively improves the comfort of residents' winter life and also contributes to urban energy saving and emission reduction. For example, in some cities in Northeast China and North China, the popularization rate of centralized heating can reach more than 80%, which basically guarantees the heating demand of most users in urban areas.

[0003] As a key link connecting heat sources and users, the construction and operation of the heating pipe network are crucial. After years of development, the heating pipe network in most cities has formed a relatively complex ring or branch layout. At the same time, the hydraulic balance regulation of the pipe network system is difficult, and there may be uneven heating between different regions and different users, such as some users with indoor temperature being too high and some users with indoor temperature not meeting the standard.

[0004] Because the communities supplied by each heat station are built in different years, have different thermal insulation structures and different user heat dissipation forms, different heating parameters are needed for the operation of different heat station systems. In the past, the deviation of the heating parameters of the heat station system was mainly filled in by the operation personnel of the heat company according to experience, which was time-consuming and laborious, and the accuracy needed to be considered. In addition, the health degree of the heat station equipment also affects the heating parameters of the heat station system, and the health degree is implicit, so the parameters filled in by the operation personnel usually ignore the influence of the health degree. SUMMARY

[0005] Therefore, the purpose of the present application is to provide an AI-based heat station difference analysis method and system to solve the problems in the background art.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme:

[0007] The AI-based heat station difference analysis method of the present application comprises the following steps:

[0008] obtaining basic information of a plurality of heat stations, historical operation data of the plurality of heat stations at a plurality of historical time points, and sampling room temperatures of covered users at the plurality of historical time points; and obtaining reference values and maintenance records of a plurality of heating parameters of the plurality of heat stations, wherein the historical operation data comprises values of a plurality of heating parameters and values of a plurality of environmental parameters, and the basic information comprises unit number, construction year, construction year of a heating coverage area and heating area;

[0009] filter and clean historical operation data of the plurality of heat supply stations at a plurality of historical time points based on the sampling room temperature, to obtain target operation data of the heat supply stations in a normal working state;

[0010] For each heat supply station, a difference value between a value of a plurality of heating parameters in the target operation data and a reference value is calculated, and a time series statistical analysis is performed on the difference values of the plurality of heating parameters corresponding to the plurality of historical time points and the maintenance records, to obtain an evaluation parameter related to the health degree; and a health degree influence model is constructed based on values of the evaluation parameters of the plurality of heat supply stations at the plurality of historical time points, wherein the health degree influence model represents a corresponding relationship between a health degree value and the value of the evaluation parameter;

[0011] Preliminary training data are constructed based on basic information of the plurality of heat supply stations and the target operation data at the plurality of historical time points, and a health degree evaluation is performed on the training data based on the health degree influence model, to obtain training data fused with the health degree dimension; and an artificial neural network is trained based on the training data, to obtain a heating parameter prediction model;

[0012] A heating parameter of a heat supply station is predicted based on the health degree influence model and the heating parameter prediction model, to obtain a predicted difference value, and a reference value of a corresponding heating parameter is compensated based on the predicted difference value.

[0013] In an embodiment of the present application, the target operation data of the plurality of heat supply stations in a normal working state are obtained by filtering and cleaning historical operation data of the plurality of heat supply stations at a plurality of historical time points based on the sampling room temperature, including:

[0014] The sampling room temperature in the historical operation data is compared with a preset normal operation temperature range, and the historical operation data in which the sampling room temperature is within the preset normal operation temperature range is retained, to obtain intermediate data;

[0015] For the intermediate data with missing values, linear interpolation is performed to fill in the missing values or the missing values are deleted, to obtain the target operation data.

[0016] In an embodiment of the present application, the difference value between the value of the plurality of heating parameters in the target operation data and the reference value includes:

[0017] For any one heat supply station, values of a plurality of environmental parameters and values of a plurality of heating parameters corresponding to a plurality of historical time points are obtained, and values of the plurality of heating parameters corresponding to the plurality of historical time points are corrected based on a pre-constructed environmental correction model, to obtain corrected values of the plurality of heating parameters corresponding to the plurality of historical time points under standard environmental conditions, wherein the environmental correction model represents an influence proportion of the heating parameter value under different environmental conditions;

[0018] The difference value is obtained by subtracting the corresponding reference value from the corrected value of each heating parameter.

[0019] In an embodiment of the present application, the method for constructing the environment correction model comprises:

[0020] The multiple heat supply stations are clustered based on the basic information to obtain multiple heat supply station sets with similar hardware facility conditions;

[0021] The maintenance time points of the multiple heat supply stations are extracted from the maintenance record, and the multiple operation cycles of the multiple heat supply stations are extracted based on the maintenance time points, and the center time point of each operation cycle is taken as a reference time point;

[0022] The values of the multiple environment parameters at the multiple reference time points of each heat supply station set are normalized to obtain environment parameter sequences of the multiple environment parameters, and the values of the multiple heating parameters at the multiple reference time points of each heat supply station set are normalized to obtain multiple heating parameter sequences;

[0023] For any combination of environment parameters and heating parameters, the time sequence correlation coefficients of the corresponding environment parameter sequences and heating parameter sequences are calculated after binarization based on the sequence average, and the combination with a time sequence correlation coefficient greater than a set correlation coefficient is taken as a correlation combination, wherein the mathematical expression of the time sequence correlation coefficient is:

[0024]

[0025] In the formula, the binarized environment parameter sequence is represented by The binarized heating parameter sequence is represented by

[0026] All associated environment parameters of each heating parameter are extracted from the correlation combination, and a relationship between the heating parameter and all associated environment parameters of the heating parameter is constructed; and the relationship is subjected to multiple linear fitting based on the value of the heating parameter corresponding to the reference time point to obtain a relationship between the heating parameter and all associated environment parameters of the heating parameter.

[0027] The environment correction model is constructed based on the relationships of the multiple heating parameters of the multiple heat supply station sets.

[0028] In an embodiment of the present application, the evaluation parameters related to the health degree are obtained by performing time sequence statistical analysis on the difference values of the multiple heating parameters corresponding to the multiple historical time points and the maintenance record, and the evaluation parameters include:

[0029] The maintenance time points of the multiple heat supply stations are extracted from the maintenance record;

[0030] ​​Based on the maintenance time points, one or more operating cycles for each heating parameter and the difference values ​​of multiple historical time points within the operating cycle are extracted. The two ends of the operating cycle are the maintenance time points, with the first maintenance time point being zero. ;

[0031] For each historical time point within the aforementioned operating cycle Calculate historical time points With the zero point Interval duration The duration sequence was obtained; and based on each heating parameter The difference value is constructed from the difference values ​​at each historical time point within the said operating cycle. Sequence, where, Represent natural numbers;

[0032] The duration sequence and the difference value sequence are normalized respectively to obtain the normalized duration sequence and the normalized difference value sequence;

[0033] Calculate the correlation between the normalized duration sequence and the normalized difference sequence. Heating parameters with a correlation greater than a preset correlation threshold are used as evaluation parameters related to health, wherein the correlation is calculated as follows:

[0034]

[0035] In the formula, Represents the normalized i-th Duration, This represents the average normalized duration in the normalized duration sequence. Represents the normalized i-th One difference value, These are the normalized differences in the normalized difference value sequence.

[0036] In one embodiment of this application, a health impact model is constructed based on the values ​​of evaluation parameters from multiple thermal stations at multiple historical time points, including:

[0037] Based on the aforementioned basic information, multiple heating stations are clustered to obtain a set of multiple heating stations with similar hardware facilities.

[0038] Extract maintenance time points for multiple heating stations from the maintenance records;

[0039] Based on the maintenance time points, multiple operating cycles C are extracted from the maintenance records, and the initial reference values ​​of the evaluation parameters at the first time point of each operating cycle C are extracted. The final reference value of the evaluation parameters at the last time point. ; and correcting the initial reference value and the end reference value based on the environment correction model to obtain an initial corrected reference value and an end corrected reference value , wherein, denotes the serial number of the heat supply parameter;

[0040] calculating the first-end difference of the evaluation parameter in each operation cycle , ; screening out the maximum difference , and setting the health degree of the operation cycle corresponding to the maximum difference to 0, and setting the health degrees of other operation cycles to

[0041]

[0042] In the formula, is the first-end difference of the evaluation parameter in the operation cycle C;

[0043] calculating, for any heat station set, a relationship formula based on multiple evaluation parameters:

[0044]

[0045] In the formula, denotes the number of evaluation parameters, denotes the item coefficient, denotes the offset, denotes the health degree;

[0046] based on the initial corrected reference value , the end corrected reference value and the health degree of multiple operation cycles in the heat station set, performing multiple linear fitting on the relationship formula to obtain a health degree influence relationship formula of the current heat station set; and based on the health degree influence relationship formulas of multiple heat station sets, constructing a health degree influence model.

[0047] In an embodiment of the present application, based on the health degree influence model, the health degree of the training data is evaluated to obtain training data fused with the health degree dimension, including:

[0048] obtaining a source heat station of the training data, and based on the source heat station, determining a target health degree influence relationship formula from the health degree influence model;

[0049] substituting the value of the evaluation parameter in the training data into the target health degree influence relationship formula to obtain the health degree of the source heat station when the training data is generated, and obtaining preliminary training data fused with the health degree dimension; ​

[0050] calculating a difference value between the preliminary training data and pre-constructed reference data to obtain training data, wherein the reference data is experience data.

[0051] In an embodiment of the present application, the artificial neural network is trained based on the training data to obtain a heating parameter prediction model, including:

[0052] S1, inputting the training data into the artificial neural network to obtain a prediction value;

[0053] S2, calculating a loss between the prediction vector and the heating parameter value in the training data based on a loss function, and adjusting the parameters of the artificial neural network based on the loss, and returning to step S1 until fitting, to obtain a heating parameter prediction model.

[0054] In an embodiment of the present application, the heating parameter of the heat station is predicted based on the health degree influence model and the heating parameter prediction model to obtain a prediction difference value, including:

[0055] obtaining values of multiple heating parameters and multiple environmental parameters of the target heat station at a latest collection time point, and obtaining environmental prediction data of a region where the target heat station is located;

[0056] correcting the values of the multiple heating parameters based on the environmental correction model to obtain corrected values of the multiple heating parameters;

[0057] inputting the value of the evaluation parameter in the multiple heating parameters into the health degree influence model to obtain a health degree;

[0058] vectorizing the basic information of the target heat station, the health degree, the values of the multiple heating parameters and the multiple environmental parameters obtained at the latest collection time point, and inputting into the heating parameter prediction model to obtain a heating parameter prediction vector.

[0059] The present application also provides an AI-based heat station difference analysis system, including:

[0060] The acquisition module is used to acquire basic information of multiple heat stations, historical operation data of the multiple heat stations at multiple historical time points, and sampling room temperatures of covered users at the multiple historical time points, and to acquire reference values and maintenance records of multiple heating parameters of the multiple heat stations, wherein the historical operation data includes values of multiple heating parameters and values of multiple environmental parameters, and the basic information includes unit number, construction year, construction year of heating coverage area, and heating area.

[0061] The data cleaning module is used to filter and clean the historical operation data of the multiple heat stations at the multiple historical time points based on the sampling room temperatures to obtain target operation data of the heat stations under normal working conditions.

[0062] a health degree model module, configured to calculate, for each heat station, a difference value between a value of a plurality of heat supply parameters in the target operation data and a reference value, and perform time series statistical analysis on the difference values of the plurality of heat supply parameters corresponding to a plurality of historical time points to obtain an evaluation parameter related to the health degree; and construct a health degree influence model based on values of the evaluation parameter of the plurality of heat stations at the plurality of historical time points, wherein the health degree influence model represents a corresponding relationship between a health degree value and the value of the evaluation parameter;

[0063] a prediction model construction module, configured to construct preliminary training data based on basic information of the plurality of heat stations and the target operation data at the plurality of historical time points, perform health degree evaluation on the training data based on the health degree influence model to obtain training data fused with the health degree dimension, and train an artificial neural network based on the training data to obtain a heat supply parameter prediction model;

[0064] a prediction analysis module, configured to perform prediction on the heat supply parameters of the heat station based on the health degree influence model and the heat supply parameter prediction model to obtain a predicted difference value, and perform compensation on the reference value of the heat supply parameter of the corresponding heat station based on the predicted difference value.

[0065] The application has the beneficial effects that the AI-based heat station difference analysis method and system can obtain a large amount of historical operation data of heat stations and sampling room temperature, clean the data by removing missing or invalid data, find heat supply parameters related to the health degree, i.e., evaluation parameters, by analyzing the differences between the values of the heat supply parameters and the reference values when the heat stations operate in various situations, construct a health degree influence model based on historical samples of the evaluation parameters, perform health degree evaluation on the training data, introduce the influence of the health degree dimension on the values of the heat supply parameters, train the model, automatically infer the difference values of the heat supply parameters according to the trained model, and compensate different heat supply stations by using the predicted difference values, that is, the running parameter compensation can be performed according to the actual situation of different heat supply stations. The problems of unreliability and neglect of the influence of the health degree when manually filling in the heat supply parameters of the heat stations are avoided. BRIEF DESCRIPTION OF DRAWINGS

[0066] The application will be further described below in combination with the drawings and embodiments:

[0067] Figure 1 is a flowchart of an AI-based heat station difference analysis method according to an embodiment of the application;

[0068] Figure 2 is a data cleaning flowchart according to an embodiment of the application;

[0069] Figure 3A flowchart of a difference value calculation process in an embodiment of the present application;

[0070] Figure 4 A flowchart of an evaluation parameter analysis process in the present application;

[0071] Figure 5 A flowchart of a health degree influence model construction method in an embodiment of the present application;

[0072] Figure 6 A flowchart of a heat supply parameter prediction process in an embodiment of the present application;

[0073] Figure 7 A structural diagram of an AI-based heat station difference analysis system in an embodiment of the present application. DETAILED DESCRIPTION

[0074] The present application can be implemented or applied in other different specific embodiments, and each detail in the present specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0075] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and only the layers related to the present application are shown in the diagrams, not the number of layers, shapes and sizes when actually implemented. The actual implementation of each layer can be randomly changed in type, number and proportion, and the layer layout pattern can also be more complex.

[0076] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application, however, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details.

[0077] Figure 1 A flowchart of an AI-based heat station difference analysis method in an embodiment of the present application, as shown in Figure 1 , the specific process includes:

[0078] S110, obtaining basic information of a plurality of heat stations, historical operation data of the plurality of heat stations at a plurality of historical time points, and sampling room temperatures of covered users at the plurality of historical time points; and obtaining reference values of a plurality of heat supply parameters and maintenance records of the plurality of heat stations, wherein the historical operation data includes values of a plurality of heat supply parameters and values of a plurality of environmental parameters;

[0079] Specifically, the basic information of the heat station includes: unit number, construction year, construction year of heat supply coverage area, and heat supply area, etc.

[0080] The heat supply parameters include: primary supply and return temperature, pressure; secondary supply and return temperature, pressure; unit heat supply area; radiator type; unit primary instantaneous flow, instantaneous heat, cumulative flow, cumulative heat; feedback opening of electric regulating valve, etc. If the data is incomplete, the minimum data category includes: unit number, date and time point, secondary supply and return temperature, unit heat supply area;

[0081] The environmental parameters include: temperature, humidity, and wind speed.

[0082] The sampling room temperature refers to the temperature of the sampling point in the heat supply residential area, which is collected regularly.

[0083] Exemplarily, the simplest running data sample includes {unit number, collection time point, secondary supply and return temperature, unit heat supply area, temperature, humidity, wind speed, sampling room temperature}.

[0084] S120, based on the sampling room temperature, the historical running data of the plurality of heat stations at a plurality of historical time points are screened and cleaned to obtain target running data of the heat station in a normal working state;

[0085] Figure 2 For an embodiment of the present application, the data cleaning flowchart is shown in Figure 2 For the running data samples collected at different time points, there may be missing or distorted data, so before formally using the running data samples to perform analysis, the data needs to be cleaned, including:

[0086] S121, comparing the sampling room temperature in the historical running data with a preset normal running temperature range, and retaining the historical running data with the sampling room temperature in the preset normal running temperature range to obtain intermediate data;

[0087] According to the device design parameters or industry standards (for example: 20℃≤sampling room temperature≤30℃), for each record in the historical data, it is checked whether the sampling room temperature is in the preset range. The data meeting the range is retained, and the data not meeting the range is rejected. Thus, the data in the warming-up stage or abnormal data is screened out, and the data with normal heat supply parameters is retained.

[0088] S122, for the intermediate data with missing items, linear interpolation is performed to fill in or delete the missing items to obtain target running data.

[0089] For the data with less than 3 missing items, linear interpolation can be performed according to the data of the two time points before and after to fill in the missing items, but the data with more than 3 missing items is directly discarded.

[0090] S130, for each heat station, calculate the difference value of the value of the plurality of heat supply parameters in the target operation data and the reference value, and perform time series statistical analysis on the plurality of heat supply parameter difference values corresponding to a plurality of historical time points and the maintenance record to obtain an evaluation parameter related to the health degree; and construct a health degree influence model based on the values of the evaluation parameters of a plurality of heat stations at a plurality of historical time points, wherein the health degree influence model represents the correspondence between the health degree value and the value of the evaluation parameter;

[0091] In this application, the decline of the health degree of heat station equipment (such as heat exchanger, circulating water pump, valve, etc.) (such as scaling, wear, blockage, aging, etc.) will directly lead to the deviation of the heat supply parameter from the design reference value. For example: if the heat exchange surface is scaled, the secondary network water supply temperature will be lower than the design value, resulting in an increase in the supply and return water temperature difference. Insufficient flow will cause the supply and return water pressure of the primary network or the secondary network to drop, and the temperature difference will be abnormal. The pressure or flow parameter will deviate from the normal range, and even cause system imbalance. These deviations (difference values) can be regarded as signals of the health status of the equipment, and through long-term monitoring and data analysis, a quantitative relationship between the difference value and the health degree can be established.

[0092] However, due to the influence of environmental conditions, but there are many condition interferences in this application, therefore, before calculating the difference value, an environmental correction model needs to be constructed to supplement the influence of different environmental conditions on the heat supply parameter, and the construction method of the environmental correction model includes:

[0093] (1) Clustering a plurality of heat stations based on the basic information to obtain a plurality of heat station sets similar in hardware facility conditions;

[0094] The plurality of heat stations are clustered first, and the goal is to group heat stations similar in hardware facility conditions and covering similar residential areas to unify the basic information conditions.

[0095] In this embodiment, K-means clustering or hierarchical clustering (Hierarchical Clustering) is used to group features. Before clustering, standardization processing (such as Z-score standardization) needs to be performed first. The optimal number of clusters is determined by the silhouette coefficient (Silhouette Score) or the elbow method (Elbow Method).

[0096] (2) Extracting maintenance time points of a plurality of heat stations from the maintenance record, and extracting a plurality of operation cycles of a plurality of heat stations based on the maintenance time points, and taking the center time point of each operation cycle as a reference time point;

[0097] The maintenance time point refers to the time point at which the maintenance personnel record the maintenance of the main equipment of the heat station. Generally, the equipment in the heat station needs to be maintained regularly. The following table is a reference table of the maintenance cycle of the main equipment of the heat station:

[0098] Table 1. Reference data table of maintenance cycle of main equipment of heat station

[0099]

[0100] Generally, the maintenance personnel will perform maintenance according to the predetermined maintenance cycle, but since the heat station monitoring system also monitors some heating parameters of the main equipment, maintenance will also be performed if the heating parameter value is abnormal, so the maintenance time point is not completely fixed.

[0101] The present application extracts the maintenance cycle of any two adjacent maintenance time points, and each maintenance cycle represents the process of the health degree of the equipment from 100% to a certain percentage. In order to control the influence of the health degree on the heating parameters, the present application extracts the values of multiple heating parameters at the center time point of each maintenance cycle as reference values.

[0102] In the present application, for the data of each heat station set, the parameter The value at the reference time point is represented as .

[0103] (3) The values of multiple environmental parameters at multiple reference time points of each heat station set are normalized to obtain an environmental parameter sequence of the multiple environmental parameters; and the values of multiple heating parameters at multiple reference time points of each heat station set are normalized to obtain a heating parameter sequence of the multiple heating parameters;

[0104] The present application first uses normalization processing to eliminate the dimensions of the multiple heating parameters and environmental parameters. The normalization conversion method can be maximum-minimum normalization. Then, the reference time points are used to align the sequences to calculate the correlation between the two sequences.

[0105] In addition, since the data of different heat stations in the same heat station set may overlap in time, for example, there are two reference values of parameters at the reference time point For this case, the average value is combined.

[0106] For example, for the parameter , after normalization of the reference values at multiple reference time points, the sequence { , ,... } is obtained;

[0107] (4) For any combination of environmental parameters and heating parameters, the time-series correlation coefficient of the corresponding environmental parameter sequence and heating parameter sequence is calculated after binarization based on the average value within the sequence. Combinations with a time-series correlation coefficient greater than a set correlation coefficient are considered as related combinations. The time-series correlation coefficient... The mathematical expression is:

[0108]

[0109] In the formula, This represents the binarized sequence of environmental parameters. This represents the binarized sequence of heating parameters;

[0110] Because environmental parameter sequences or heating parameter sequences may exhibit significant fluctuations, the Pearson correlation coefficient is not suitable for calculating correlation. Therefore, in this embodiment, time-series correlation analysis is used to determine the correlation. First, the average value of all reference values ​​within the sequence is used to binarize all reference values, for example:

[0111] The original environmental sequence is {0.31, 0.52, ... 0.74}, with an average value of 0.45. The binarized environmental parameter sequence is {0, 0, ... 1}.

[0112] Binarizing the sequence simplifies the value distribution of the sequence. Then, the difference between the binarized heating parameter sequence and the binarized environmental parameter sequence is calculated. The difference sequence reflecting the difference between the two was calculated. Then, the difference sequence was analyzed. The average value of each element in the algorithm is calculated. If the difference is large or small, it indicates a positive or negative correlation. For ease of representation, the absolute value of the average is taken and compared with a threshold (which can be 0.7) to filter out the environmental parameters related to each heating parameter.

[0113] For example, the secondary supply and return temperature is related to temperature, humidity and wind speed.

[0114] (5) Extract all associated environmental parameters of each heating parameter from the relevant combination, and construct the relationship between the heating parameter and all associated environmental parameters of the heating parameter; and perform multivariate linear fitting on the relationship based on the value of the heating parameter corresponding to the reference time point to obtain the relationship between the heating parameter and all associated environmental parameters of the heating parameter.

[0115] After obtaining the correlation judgment results, the correlation needs to be quantified. This application obtains the relationship between heating parameters and related environmental parameters based on multiple linear fitting. The mathematical expression of the relationship is as follows:

[0116]

[0117] wherein, represents the coefficient of the term of the term, represents the coefficient of the term of the environmental parameter, is an offset;

[0118] Finally, the coefficients of the relationship are estimated using the least squares method or linear regression, the specific process being prior art, which will not be repeated here.

[0119] (6) Based on the relationship of multiple heat supply parameters of multiple heat stations, an environmental correction model is constructed.

[0120] Through the environmental correction model, the parameters under different environmental conditions can be corrected to unify the fluctuations of the heat supply parameters caused by different environmental conditions.

[0121] Figure 3 The difference value calculation flowchart in an embodiment of the present application is shown in FIG. 6. Through the above environmental correction model, the difference value between the value of the multiple heat supply parameters in the target operation data and the reference value is calculated, including: Figure 3

[0122] S1301, for any one heat station, the values of multiple environmental parameters and multiple heat supply parameters corresponding to multiple historical time points are obtained, and the values of multiple heat supply parameters corresponding to multiple historical time points are corrected based on the pre-constructed environmental correction model to obtain the corrected values of multiple heat supply parameters corresponding to multiple historical time points under standard environmental conditions, wherein the environmental correction model represents the influence proportion of different environmental conditions on the heat supply parameter value;

[0123] When correcting the heat supply parameters of each heat station, the corresponding relationship is found from the environmental correction model, the standard environmental conditions (for example: temperature 0℃, humidity 40%, wind speed 5m / s) are substituted into the relationship, and the corrected value is obtained.

[0124] S1302, the corrected value of each heat supply parameter is subtracted from the corresponding reference value to obtain the difference value.

[0125] After correction, the difference value is obtained by subtracting the corrected value of each heat supply parameter from the reference value. The reference value is also the value collected after maintenance under standard environmental conditions.

[0126] Figure 4 The evaluation parameter analysis flowchart in the present application is shown in FIG. 7. After calculating the difference value, the difference value can be used for health degree correlation analysis, which specifically includes: Figure 4

[0127] ​​S1311, Extract maintenance time points of multiple heating stations from the maintenance records;

[0128] S1312, based on the maintenance time point, extract one or more operating cycles for each heating parameter and the difference values ​​of multiple historical time points within the operating cycle, wherein the two ends of the operating cycle are maintenance time points, and the maintenance time point at the beginning is zero. ;

[0129] The health status in this application will be reset during each maintenance. For example, when a water pump is running continuously, scale buildup can cause the pump speed to decrease, which in turn can lead to a decrease in water flow. The health status will be restored after maintenance.

[0130] Therefore, it is necessary to break down and analyze the data based on the operating cycle.

[0131] S1313, for each historical time point within the aforementioned operating cycle Calculate historical time points With the zero point Interval duration The duration sequence was obtained; and based on each heating parameter The difference value is constructed from the difference values ​​at each historical time point within the said operating cycle. Sequence, where, Represent natural numbers;

[0132] First, for each heating station, evaluation parameters related to its health status are identified. These parameters change as health status decreases. In this application, health status is assumed to decrease linearly over the operating cycle; therefore, Pearson correlation coefficient analysis is employed. Thus, a sequence of differences in operating time and heating parameters is first constructed.

[0133] S1314, Normalize the duration sequence and the difference value sequence respectively to obtain a normalized duration sequence and a normalized difference value sequence;

[0134] Normalization eliminates dimensions, facilitating subsequent processing. The Pearson correlation coefficient is then used to calculate the correlation.

[0135] S1315, Calculate the correlation between the normalized duration sequence and the normalized difference value sequence. Heating parameters with a correlation greater than a preset correlation threshold are used as evaluation parameters related to health, wherein the correlation is calculated as follows:

[0136]

[0137] In the formula, Represents the normalized i-th Duration, This represents the average normalized duration in the normalized duration sequence. Represents the normalized i-th One difference value, These are the normalized differences in the normalized difference value sequence.

[0138] Figure 5 This is a flowchart of a method for constructing a health impact model according to an embodiment of this application, as shown below. Figure 5 As shown, the methods for constructing the health impact model include:

[0139] S1321, Based on the basic information, multiple heating stations are clustered to obtain a set of multiple heating stations with similar hardware facilities.

[0140] First, multiple heating stations are clustered. The goal is to group heating stations with similar hardware facilities and coverage of residential areas to unify basic information conditions.

[0141] In this embodiment, K-means clustering or hierarchical clustering is used to group the features. Before clustering, standardization (such as Z-score standardization) is required. The optimal number of clusters is determined using the silhouette score or the elbow method.

[0142] S1322, Extract maintenance time points of multiple heating stations from the maintenance records;

[0143] S1323, Based on the maintenance time point, extract multiple operating cycles C from the maintenance record, and extract the initial reference value of the evaluation parameter at the first time point of each operating cycle C. The final reference value of the evaluation parameters at the last time point. ; and the initial reference value based on the environmental correction model, the values ​​of the environmental parameters at the first time point, and the values ​​of the environmental parameters at the last time point. and the aforementioned end reference value Make corrections to obtain the initial correction reference value. and end of correction reference value ,in, Indicates the sequence number of the heating parameter;

[0144] An operating cycle C is defined as the period between two consecutive maintenance points. Evaluation parameters include supply and return water temperature, pressure, flow rate, and power consumption. The heating parameter values ​​at the beginning and end of each operating cycle are corrected to avoid interference from different meteorological conditions.

[0145] S1324, Calculate the difference between the beginning and end values ​​of the evaluation parameters in each running cycle. , ; the maximum difference value is screened out , and the maximum difference value is set to 0, and the health degrees of other operation cycles are :

[0146]

[0147] In the formula, is the first-end difference value of the evaluation parameter of the operation cycle C;

[0148] For any heating parameter, the difference value between the initial correction value and the end correction value of the evaluation parameter in each operation cycle C (such as the supply and return water temperature difference ) is calculated.

[0149] The health degree is set to 0: if the difference value of a certain cycle is the largest (such as the maximum ΔT), it indicates that the equipment performance degradation is significant, and the health degree is set to 0. The health degrees of other cycles are distributed according to the difference value. That is .

[0150] S1325, for any heat station set, a relationship is constructed based on multiple evaluation parameters:

[0151]

[0152] In the formula, represents the number of evaluation parameters, represents the term coefficient, represents the offset, is the health degree.

[0153] The application calculates multiple term coefficients and the value of the offset based on multiple linear fitting, so as to obtain the value change relationship between multiple evaluation parameters and the health degree.

[0154] S1326, based on the initial correction reference value , the end correction reference value and the health degree of multiple operation cycles in the heat station set, multiple linear fitting is performed on the relationship to obtain the health degree influence relationship of the current heat station set; and based on the health degree influence relationship of multiple heat station sets, a health degree influence model is constructed.

[0155] Finally, the complete health degree influence model is constructed based on the relationship of multiple heat station sets.

[0156] S140, constructing preliminary training data based on the basic information of multiple heat stations and the target operation data at multiple historical time points, evaluating the health degree of the training data based on the health degree influence model, and obtaining training data fused with the health degree dimension; and training an artificial neural network based on the training data to obtain a heating parameter prediction model;

[0157] After obtaining the health degree influence model, the health degree evaluation dimension can be introduced in the subsequent training process, which includes:

[0158] S1401, obtaining a source heat station of the training data, and determining a target health degree influence relationship from the health degree influence model based on the source heat station;

[0159] S1402, substituting the value of the evaluation parameter in the training data into the target health degree influence relationship to obtain the health degree of the source heat station when the training data is generated, and obtaining preliminary training data fused with the health degree dimension;

[0160] S1403, calculating the difference value of the preliminary training data and the pre-constructed reference data to obtain training data, wherein the reference data is experience data.

[0161] In constructing a single training data, according to selecting a historical time point, the basic information of any one heat station corresponding to the historical time point, the values of multiple meteorological parameters, and the health degree (obtained based on the health degree influence model) are taken as training samples, wherein the training samples are essentially difference values for subsequent data reference of the staff. The values of multiple heating parameters are taken as labels. The above data are all subjected to vectorization processing.

[0162] Then the training can be performed, including:

[0163] S1, inputting the training data into an artificial neural network to obtain a prediction value;

[0164] The training data is transmitted through the input layer, the hidden layer (which can be multiple layers) and the output layer. Each layer of neurons performs weighted summation on the input and applies an activation function (such as ReLU, Sigmoid, etc.) to generate a prediction value.

[0165] S2, calculating the loss of the prediction value and the heating parameter value in the training data based on a loss function, adjusting the parameters of the artificial neural network based on the loss, and returning to step S1 until fitting, to obtain a heating parameter prediction model.

[0166] The loss function uses mean square error (MSE) to measure the gap between the prediction value and the true value, and calculates the gradient of the loss function to each parameter through the back propagation algorithm.

[0167] S150, predicting the heat supply parameter of the heat station based on the health degree influence model and the heat supply parameter prediction model, obtaining a prediction difference value, and compensating the heat supply parameter reference value of the corresponding heat station based on the prediction difference value.

[0168] The compensation value corresponding to each station is determined according to the energy-saving building, non-energy-saving building, hanging warm, floor heating, building age and other attributes, in addition, the heat supply administrator can also adjust according to the experience value and the complaint rate in the running process. After multiple adjustments, the compensation value is closer and closer to the actual deviation of the heat dissipation of the building of each station.

[0169] Figure 6 The heat supply parameter prediction process in an embodiment of the present application is shown in the schematic diagram as shown in Figure 6 , which includes:

[0170] (1) First, the values of the multiple heat supply parameters and the multiple environment parameters obtained by the target heat station at the latest collection time point are obtained, and the environment prediction data of the region where the target heat station is located is obtained;

[0171] (2) Then, the values of the multiple heat supply parameters are corrected using the environment correction model and the values of the multiple environment parameters, to obtain the corrected values of the multiple heat supply parameters; the correction process is described above and will not be repeated here;

[0172] (3) The value of the evaluation parameter in the multiple heat supply parameters is input into the health degree influence model to obtain the health degree;

[0173] (4) The basic information of the target heat station, the health degree, the values of the multiple heat supply parameters and the multiple environment parameters obtained at the latest collection time point are vectorized and input into the heat supply parameter prediction model to obtain a heat supply parameter prediction vector.

[0174] The basic information of the target heat station, the health degree, the values of the multiple heat supply parameters and the multiple environment parameters obtained at the latest collection time point form a vector, which is input into the prediction model, and the output vector is the predicted value of the heat supply parameter. This value can be directly filled into the statistical file to provide relevant personnel for execution. The problem of easy error in manual filling is solved.

[0175] The application provides an AI-based heat station difference analysis method, which comprises the following steps: obtaining historical operation data and sampling room temperature of a large number of heat stations, and then cleaning the data to remove missing or invalid data. The difference between the value of a heat supply parameter and a reference value when the heat station is running under various conditions is utilized, and the heat supply parameter related to the health degree, i.e., the evaluation parameter, is found by analyzing the difference. Then, a health degree influence model is constructed based on the historical samples of the evaluation parameter, so as to perform health degree evaluation on the training data, introduce the influence of the health degree dimension on the value of the heat supply parameter, and train the model. The value of the heat supply parameter is automatically inferred according to the trained model. Thus, the problem that the value of the heat supply parameter of the heat station is unreliable and the influence of the health degree is neglected when the value is filled in manually is avoided.

[0176] As shown in Figure 7 The application further provides an AI-based heat station difference analysis system, which comprises the following steps:

[0177] An acquisition module is configured to acquire basic information of a plurality of heat stations, historical operation data of the plurality of heat stations at a plurality of historical time points, sampling room temperature of users covered at the plurality of historical time points, and reference values and maintenance records of a plurality of heat supply parameters of the plurality of heat stations, wherein the historical operation data comprises values of a plurality of heat supply parameters and values of a plurality of environmental parameters, and the basic information comprises unit number, construction year, construction year of a heat supply coverage area and heat supply area.

[0178] A data cleaning module is configured to filter and clean the historical operation data of the plurality of heat stations at the plurality of historical time points based on the sampling room temperature, so as to obtain target operation data of the heat stations under normal working conditions.

[0179] A health degree model module is configured to calculate, for each heat station, a difference value between the values of the plurality of heat supply parameters in the target operation data and reference values, and perform time sequence statistical analysis on the difference values of the plurality of heat supply parameters corresponding to the plurality of historical time points, so as to obtain an evaluation parameter related to the health degree. A health degree influence model is constructed based on the values of the evaluation parameters of the plurality of heat stations at the plurality of historical time points, wherein the health degree influence model represents the corresponding relationship between the health degree value and the value of the evaluation parameter.

[0180] A prediction model construction module is configured to construct preliminary training data based on the basic information of the plurality of heat stations and the target operation data at the plurality of historical time points, perform health degree evaluation on the training data based on the health degree influence model, so as to obtain training data fused with the health degree dimension, and train an artificial neural network based on the training data, so as to obtain a heat supply parameter prediction model.

[0181] The prediction analysis module is configured to predict the heating parameter of the heat supply station based on the health degree influence model and the heating parameter prediction model, to obtain a prediction difference value, and to compensate the reference value of the heating parameter of the corresponding heat supply station based on the prediction difference value.

[0182] The AI-based heat supply station difference analysis system provided in the present application obtains a large amount of historical operation data of heat supply stations and sampling room temperature, and then cleans the data to remove missing or invalid data. The difference between the value of the heating parameter and the reference value of the heating parameter of the heat supply station under various conditions is used to find the heating parameter related to the health degree, i.e., the evaluation parameter, by analyzing the difference. Then, the health degree influence model is constructed based on the historical samples of the evaluation parameter, so as to perform health degree evaluation on the training data, introduce the influence of the health degree dimension on the value of the heating parameter, and train the model. The value of the heating parameter is automatically inferred according to the trained model. Thus, the problem that the value of the heating parameter of the heat supply station is manually filled in, which is unreliable and ignores the influence of the health degree, is avoided.

[0183] The embodiment further provides an electronic terminal, including a processor and a memory.

[0184] The memory is configured to store a computer program, and the processor is configured to execute the computer program stored in the memory, so that the terminal executes any method in the embodiment.

[0185] The computer readable storage medium in the embodiment can be understood by those skilled in the art that all or part of the steps of the above-mentioned method embodiments can be completed by a computer program related hardware. The foregoing computer program can be stored in a computer readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the foregoing storage medium includes ROM, RAM, magnetic disc or optical disc and various storage medium that can store program codes.

[0186] The electronic terminal provided in the embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected with the processor and the transceiver and complete communication between each other. The memory is configured to store a computer program, the communication interface is configured to communicate, and the processor and the transceiver are configured to run the computer program, so that the electronic terminal executes each step of the method.

[0187] In the embodiment, the memory can include a random access memory (RAM) and can also include a non-volatile memory such as at least one disk memory.

[0188] The processor described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0189] In the above-described embodiments, although the present application has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art in light of the foregoing description. Embodiments of this application are intended to embrace all such alternatives, modifications and variations as can fall within the scope of the appended claims.

[0190] The above-described embodiments are merely illustrative for the principles and effects of the present application, but are not intended to limit the present application. Any modification or change made by any person skilled in the art without departing from the spirit and scope of the present application shall be covered by the claims of the present application.

Claims

1. A method for analyzing the differences between heat stations based on AI, characterized in that, include: Obtain basic information of multiple heating stations, historical operating data of multiple heating stations at multiple historical time points, and sampled room temperature of covered users at multiple historical time points; It also obtains reference values ​​and maintenance records of multiple heating parameters from multiple heating stations. The historical operation data includes values ​​of multiple heating parameters and multiple environmental parameters. The basic information includes unit number, year of construction, year of construction of the heating coverage area, and heating area. Based on the sampled room temperature, the historical operating data of multiple heating stations at multiple historical time points are filtered and cleaned to obtain the target operating data of the heating stations under normal operating conditions. For each heating station, the differences between the values ​​of various heating parameters and reference values ​​in the target operating data are calculated. Time-series statistical analysis is performed based on the differences in the values ​​of various heating parameters corresponding to multiple historical time points and the maintenance records to obtain evaluation parameters related to health. A health impact model is constructed based on the values ​​of the evaluation parameters at multiple historical time points for multiple heating stations. This health impact model characterizes the correspondence between health values ​​and the values ​​of the evaluation parameters. The calculation of the differences between the values ​​of various heating parameters and reference values ​​in the target operating data includes: for any heating station, obtaining the values ​​of various environmental parameters and various heating parameters corresponding to multiple historical time points; and correcting the values ​​of various heating parameters corresponding to multiple historical time points based on a pre-constructed environmental correction model to obtain corrected values ​​of various heating parameters corresponding to multiple historical time points under standard environmental conditions. The environmental correction model characterizes the proportion of influence of different environmental conditions on the values ​​of heating parameters. The difference between the corrected value of each heating parameter and the corresponding reference value is calculated to obtain the difference value. Preliminary training data is constructed based on basic information of multiple heating stations and target operation data at multiple historical time points. The health of the training data is evaluated based on the health impact model to obtain training data that integrates the health dimension. The artificial neural network is trained based on the training data to obtain a heating parameter prediction model. The heating parameters of the heating station are predicted based on the health status impact model and the heating parameter prediction model to obtain the predicted difference value, and the reference value of the heating parameter of the corresponding heating station is compensated based on the predicted difference value.

2. The AI-based thermal station difference analysis method according to claim 1, characterized in that, Based on the sampled room temperature, historical operating data from multiple heating stations at multiple historical time points are filtered and cleaned to obtain target operating data of the heating stations under normal operating conditions, including: The sampled room temperature in the historical operating data is compared with the preset normal operating temperature range, and the historical operating data in which the sampled room temperature is within the preset normal operating temperature range is retained to obtain intermediate data; For any missing intermediate data, perform linear interpolation to fill in the missing data or delete the missing data to obtain the target running data.

3. The AI-based thermal station difference analysis method according to claim 1, characterized in that, The method for constructing the environmental correction model includes: Based on the aforementioned basic information, multiple heating stations are clustered to obtain a set of multiple heating stations with similar hardware facilities. The maintenance time points of multiple heating stations are extracted from the maintenance records, and multiple operating cycles of multiple heating stations are extracted based on the maintenance time points, with the center time point of each operating cycle used as the reference time point; The values ​​of various environmental parameters at multiple reference time points for each heating station set are normalized to obtain a sequence of environmental parameters; and the values ​​of various heating parameters at multiple reference time points for each heating station set are normalized to obtain a sequence of heating parameters. For any combination of environmental and heating parameters, the time-series correlation coefficients of the corresponding environmental and heating parameter sequences are calculated after binarization based on the intra-series average values. Combinations with time-series correlation coefficients greater than a predetermined correlation coefficient are considered as correlated combinations. The time-series correlation coefficients... The mathematical expression is: In the formula, This represents the binarized sequence of environmental parameters. This represents the binarized sequence of heating parameters; Extract all associated environmental parameters for each heating parameter from the relevant combinations, and construct the relationship between the heating parameter and all associated environmental parameters; then perform multivariate linear fitting on the relationship based on the value of the heating parameter at the reference time point to obtain the relationship between the heating parameter and all associated environmental parameters. An environmental correction model is constructed based on the relationships between various heating parameters from multiple heating stations.

4. The AI-based heat station difference analysis method according to claim 1, characterized in that, Based on the differences in various heating parameters corresponding to multiple historical time points and the maintenance records, time-series statistical analysis was performed to obtain evaluation parameters related to health status, including: Extract maintenance time points for multiple heating stations from the maintenance records; Based on the maintenance time points, one or more operating cycles for each heating parameter and the difference values ​​of multiple historical time points within the operating cycle are extracted. The two ends of the operating cycle are the maintenance time points, with the first maintenance time point being zero. ; For each historical time point within the aforementioned operating cycle Calculate historical time points With the zero point Interval duration The duration sequence was obtained; and based on each heating parameter The difference value is constructed from the difference values ​​at each historical time point within the said operating cycle. Sequence, where, Represent natural numbers; The duration sequence and the difference value sequence are normalized respectively to obtain the normalized duration sequence and the normalized difference value sequence; Calculate the correlation between the normalized duration sequence and the normalized difference sequence. Heating parameters with a correlation greater than a preset correlation threshold are used as evaluation parameters related to health, wherein the correlation is calculated as follows: In the formula, Represents the normalized i-th Duration, This represents the average normalized duration in the normalized duration sequence. Represents the normalized i-th One difference value, These are the normalized differences in the normalized difference value sequence.

5. The AI-based thermal station difference analysis method according to claim 3, characterized in that, A health impact model is constructed based on the evaluation parameter values ​​of multiple heating stations at multiple historical time points, including: Based on the aforementioned basic information, multiple heating stations are clustered to obtain a set of multiple heating stations with similar hardware facilities. Extract maintenance time points for multiple heating stations from the maintenance records; Based on the maintenance time points, multiple operating cycles C are extracted from the maintenance records, and the initial reference values ​​of the evaluation parameters at the first time point of each operating cycle C are extracted. The final reference value of the evaluation parameters at the last time point. ; and the initial reference value based on the environmental correction model. and the aforementioned end reference value Make corrections to obtain the initial correction reference value. and end of correction reference value ,in, Indicates the sequence number of the heating parameter; Calculate the first and last differences of the evaluation parameters in each running cycle. , ; Filter out the largest difference and the maximum difference The health of the corresponding runtime cycle is set to 0, while the health of other runtime cycles is set to... for: In the formula, The difference between the first and last evaluation parameters for the operating cycle C; For any set of heating stations, the following relationship is constructed based on multiple evaluation parameters: In the formula, Indicates the number of evaluation parameters. Indicates the coefficient of the term. Indicates the offset. Indicates health status; Initial correction reference values ​​based on multiple operating cycles of the heating station cluster End of correction reference value and health The relationship is fitted with multiple linear equations to obtain the health influence relationship of the current set of heating stations; and a health influence model is constructed based on the health influence relationship of multiple sets of heating stations.

6. The AI-based thermal station difference analysis method according to claim 5, characterized in that, Based on the health impact model, the training data is evaluated to obtain training data that integrates health dimensions, including: The source heat station for obtaining training data is used to determine the target health influence relationship from the health influence model based on the source heat station. Substituting the values ​​of the evaluation parameters in the training data into the target health influence formula, we obtain the health of the source heat station when the training data was generated, and thus obtain preliminary training data that integrates the health dimension. The difference between the initial training data and the pre-constructed reference data is calculated to obtain the training data, wherein the reference data is empirical data.

7. The AI-based thermal station difference analysis method according to claim 1, characterized in that, The artificial neural network is trained based on the training data to obtain a heating parameter prediction model, including: S1, Input the training data into the artificial neural network to obtain the predicted value; S2, calculate the loss between the predicted value and the heating parameter value in the training data based on the loss function, adjust the parameters of the artificial neural network based on the loss, and return to step S1 until a fit is obtained to obtain the heating parameter prediction model.

8. The AI-based thermal station difference analysis method according to claim 3, characterized in that, Based on the health status impact model and the heating parameter prediction model, the heating parameters of the heating station are predicted to obtain the predicted difference values, including: Obtain the values ​​of various heating parameters and various environmental parameters acquired at the most recent data collection point of the target heating station, and obtain the environmental prediction data of the area where the target heating station is located; The values ​​of various heating parameters are corrected based on the environmental correction model to obtain corrected values ​​for various heating parameters; The values ​​of the evaluation parameters among various heating parameters are input into the health status impact model to obtain the health status; The basic information, health status, and values ​​of various heating parameters and environmental parameters obtained at the most recent collection time of the target heating station are vectorized and input into the heating parameter prediction model to obtain the heating parameter prediction vector.

9. A heat station difference analysis system based on AI, characterized in that, include: The acquisition module is used to acquire basic information of multiple heating stations, historical operating data of multiple heating stations at multiple historical time points, and sampled room temperature of covered users at multiple historical time points; and to acquire reference values ​​and maintenance records of multiple heating parameters of multiple heating stations. The historical operating data includes values ​​of multiple heating parameters and values ​​of multiple environmental parameters. The basic information includes unit number, year of construction, year of construction of the heating coverage area, and heating area. The data cleaning module is used to filter and clean the historical operating data of multiple heating stations at multiple historical time points based on the sampled room temperature, so as to obtain the target operating data of the heating stations under normal operating conditions. The health model module is used to calculate the differences between the values ​​of various heating parameters and reference values ​​in the target operating data for each heating station, and to perform time-series statistical analysis based on the differences of various heating parameters corresponding to multiple historical time points to obtain evaluation parameters related to health. It also constructs a health impact model based on the evaluation parameter values ​​of multiple heating stations at multiple historical time points, where the health impact model characterizes the correspondence between health values ​​and the values ​​of the evaluation parameters. The calculation of the differences between the values ​​of various heating parameters and reference values ​​in the target operating data includes: for any heating station, obtaining the values ​​of various environmental parameters and various heating parameters corresponding to multiple historical time points, and correcting the values ​​of various heating parameters corresponding to multiple historical time points based on a pre-constructed environmental correction model to obtain corrected values ​​of various heating parameters corresponding to multiple historical time points under standard environmental conditions, where the environmental correction model characterizes the proportion of influence of different environmental conditions on the values ​​of heating parameters; and subtracting the corrected value of each heating parameter from its corresponding reference value to obtain the difference value. The prediction model building module is used to construct preliminary training data based on the basic information of multiple heating stations and the target operation data at multiple historical time points, and to evaluate the health of the training data based on the health impact model to obtain training data that integrates the health dimension; and to train an artificial neural network based on the training data to obtain a heating parameter prediction model. The predictive analysis module is used to predict the heating parameters of the heating station based on the health impact model and the heating parameter prediction model, obtain the predicted difference value, and compensate the reference value of the heating parameter of the corresponding heating station based on the predicted difference value.

Images