Big-data-driven intelligent energy-saving and carbon-reducing method capable of supplying according to demand

Through high-density regional algorithm, multiple interpolation and K-mean clustering technology, abnormal data and complex thermal network environment problems in user behavior analysis are solved, accurate energy use analysis and personalized adjustment are achieved, and energy waste and carbon are reduced emission.

CN119090210APending Publication Date: 2024-12-06DALIAN MARITIME UNIVERSITY
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411151785.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing user behavior analysis methods are subjective and difficult to cope with the randomness and uncertainty in complex thermal network environments. There are abnormal data in the historical sample data of thermal network users, which affects the accuracy of the analysis and leads to economic losses such as energy theft. .

Method used

The abnormal data is detected and deleted by high-density area algorithm, the missing values ​​are supplemented by multiple interpolation methods, the data is standardized by using the Min-max method, and the user's energy usage habits, working and rest time and temperature preferences are analyzed through the K-means clustering algorithm to achieve Classification of energy use.

Benefits of technology

Improve the accuracy and completeness of data, identify abnormal data and plagiarism, provide personalized energy regulation, reduce energy consumption and carbon emissions, and improve user comfort and satisfaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090210A_ABST
    Figure CN119090210A_ABST
Patent Text Reader

Abstract

The invention discloses a big-data-driven intelligent energy-saving and carbon-reducing method for supplying according to demand. The method comprises the following steps: acquiring historical sample data of a heat supply network user to be processed; detecting abnormal data in the historical sample data of the heat supply network user by adopting a high-density regional algorithm, marking the abnormal data, and deleting the abnormal data; adopting a multi-interpolation method to supplement missing values and abnormal values in the deleted historical sample data of the heat supply network users to obtain complemented historical data of the heat supply network users; performing standardization processing on the complemented historical sample data of the heat supply network users through a Min-max method, and converting the sample data into numerical values in the same interval; and performing clustering analysis on the sample data converted into the numerical values in the same interval through a K-means clustering algorithm according to the energy consumption habit, work and rest time and temperature preference of the user to obtain the energy consumption condition classification of the heat supply network. According to the method, the working state and setting of the energy equipment are automatically adjusted, so that energy services meeting individual requirements are provided.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of user behavior analysis, and in particular to a big data driven, demand-based, intelligent energy-saving and carbon-reduction method. Background Art

[0002] With the continuous development of technologies such as artificial intelligence, big data, and cloud computing, the methods and technical means of intelligent energy conservation and carbon reduction are also constantly updated and optimized. Therefore, intelligent energy conservation and carbon reduction have become the focus and hot issues of current social attention. Provide personalized services for users. Through user behavior analysis and intelligent algorithms, personalized adjustment and control can be achieved according to user needs and habits, resource collaborative optimization can be achieved, unnecessary energy waste can be reduced, and energy conservation and carbon reduction can be further strengthened. However, when faced with massive data, the current system often has the following problems:

[0003] 1. The existing user behavior analysis methods are highly subjective and the physical model is difficult to cope with the randomness and uncertainty of user behavior in complex heating network environments;

[0004] 2. There are abnormal data in the historical sample data of heating network users, which will affect the accurate analysis of user behavior.

[0005] 3. For existing users who steal energy, it will cause huge economic losses every year, hindering the normal operation and security of the energy system. Summary of the invention

[0006] In order to solve the above problems, the technical solution adopted by the present invention is: a big data driven intelligent energy saving and carbon reduction method based on demand, comprising the following steps:

[0007] S1: Obtain historical sample data of heating network users to be processed;

[0008] S2: Use a high-density area algorithm to detect abnormal data in the historical sample data of heating network users, mark the abnormal data, and delete the abnormal data;

[0009] S3: Use multiple interpolation methods to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of heating network users to obtain the completed historical data of heating network users;

[0010] S4: standardize the completed historical sample data of heating network users by using the Min-max method, and convert the sample data into values ​​in the same interval;

[0011] S5: The K-means clustering algorithm is used to perform cluster analysis on the sample data converted into numerical values ​​in the same interval, based on the user's energy usage habits, work and rest time, and temperature preference, to obtain the energy usage classification of the heating network.

[0012] Further: the process of using the high-density area method to detect abnormal data in the historical sample data of the heating network users, marking the abnormal data, and deleting the abnormal data is as follows:

[0013] S21: First, define two important parameters for the high-density region algorithm: ε and MinPts; ε is used to define the neighborhood range of the data point, and MinPts defines the minimum number of neighbors required for a core point;

[0014] S22: Finding core points: A given data set is established based on the cumulative heat consumption of heat users every hour. The number of data points in the ε neighborhood of each data point is calculated, and it is determined whether the MinPts condition is met. If the ε neighborhood of a data point contains at least MinPts neighbors, the data point is considered to be a core point.

[0015] S23: Marking boundary points: For a data point in the ε neighborhood of a core point, if the number of data points in its neighborhood is less than MinPts but contains a core point, the data point is marked as a boundary point;

[0016] S24: Create clusters: Create one or more clusters by connecting the core points and the data points in their ε neighborhood. If the ε neighborhoods of two core points overlap, they will be classified into the same cluster. The remaining data points that are not classified into any cluster are classified as outliers;

[0017] S25: Eliminate data marked as outliers.

[0018] Further: the process of using the multiple interpolation method to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of the heating network users to obtain the completed historical data of the heating network users is as follows:

[0019] S31: establishing an interpolation function through known values, estimating the value to be interpolated, and then adding different deviations to the value to form multiple sets of optional interpolation values, thereby forming multiple sets of complete data sets to be evaluated;

[0020] S32: performing statistical analysis on multiple sets of data sets to be evaluated;

[0021] S33: Evaluate the results for each data set and select the best interpolation value based on the score of the statistical model.

[0022] Further: the completed historical sample data of the heating network users is standardized by the Min-max method, and the process of converting the sample data into values ​​in the same interval is as follows: the following method is adopted:

[0023] S41: The data was standardized by using the Min-max method, and the calculation formula is as follows:

[0024]

[0025] In addition, the output value needs to be denormalized, and the calculation formula is as follows:

[0026]

[0027] In the above formula, x is the sample data, x max 、x min are the maximum and minimum values ​​of the sample data, respectively, nor The sample data is standardized.

[0028] Further: the process of converting the sample data into numerical values ​​of the same interval by the K-means clustering algorithm is as follows:

[0029] S51: Randomly select k samples as the initial cluster centers C = {c1, c2 ... c k};

[0030] S52: For each sample x in the data set i , calculate the distance from the sample to the K cluster centers respectively, and

[0031] Divide it into the class corresponding to the cluster center with the smallest distance;

[0032] S53: Recalculate each category c j , the cluster center of:

[0033]

[0034] Where n is the category c j The number of data points contained in ;

[0035] S54: Repeat the above two steps S52 and S53 until the cluster center no longer changes, and output the clustering result.

[0036] Further: the process of clustering analysis based on the user's energy usage habits, work and rest time and temperature preference to obtain the energy usage classification of the power grid is as follows:

[0037] S61: Determine clustering variables; the clustering variables include:

[0038] Energy usage habits: including daily energy consumption, peak and valley hours, and energy types;

[0039] Work and rest time: including the user's living time and active time;

[0040] Temperature preference: user's preference for indoor temperature;

[0041] S62: performing clustering result analysis;

[0042] Determine cluster centers: Check the characteristic values ​​of each cluster center and analyze the typical user profile of each category;

[0043] S63: Output classification results;

[0044] Label users with category labels: label each user with a corresponding category label;

[0045] Conduct category description: Describe the user characteristics of each category in detail;

[0046] S64: Analysis of energy usage;

[0047] Energy usage pattern: Analyze the energy usage pattern of each category by category

[0048] Load forecasting: Use clustering results to predict load changes in different time periods to help optimize heating network scheduling and energy distribution.

[0049] The data-driven, demand-based intelligent energy-saving and carbon-reduction method provided by the present invention has the following advantages:

[0050] Through the collection and analysis of user data in the thermal power grid, we can understand each user's energy usage habits, work and rest time, temperature preferences and other information. Based on the user's heat demand data, the intelligent algorithm can automatically adjust the working status and settings of energy equipment according to the user's needs and habits to provide energy services that meet individual needs;

[0051] Through personalized energy regulation and control, user comfort and satisfaction can be improved, while energy consumption and carbon emissions can be reduced.

[0052] Automated data cleaning and repair: It can automatically identify and clean errors and inconsistencies in data, and provide intelligent data repair functions to automatically fill in missing values, remove duplicate data, etc., thereby improving data accuracy and completeness;

[0053] Identify abnormal data: Use the high-density area method to detect abnormal data. When abnormal data is generated, the staff can be alerted in time and the energy theft behavior of energy users can be reported in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0055] Figure 1 is a flow chart of the present invention;

[0056] Figure 2 To invent the algorithm flow chart for medium and high density areas;

[0057] Figure 3 Flowchart of the k-means clustering algorithm in the present invention. DETAILED DESCRIPTION

[0058] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Figure 1 is a flow chart of the present invention;

[0061] A big data driven intelligent energy saving and carbon reduction method based on demand and supply, comprising the following steps:

[0062] S1: Obtain historical sample data of heating network users to be processed;

[0063] S2: Use a high-density area algorithm to detect abnormal data in the historical sample data of heating network users, mark the abnormal data, and delete the abnormal data;

[0064] S3: Use multiple interpolation methods to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of heating network users to obtain the completed historical data of heating network users;

[0065] S4: standardize the completed historical sample data of heating network users by using the Min-max method, and convert the sample data into values ​​in the same interval;

[0066] S5: The K-means clustering algorithm is used to perform cluster analysis on the sample data converted into numerical values ​​in the same interval, based on the user's energy usage habits, work and rest time, and temperature preference, to obtain the energy usage classification of the heating network.

[0067] Steps S1, S2, S3, S4 and S5 are performed sequentially;

[0068] The process of obtaining the historical sample data of the heating network users to be processed is as follows:

[0069] S11: Obtain the database type of the historical sample data of the heating network users to be processed;

[0070] S12: Create a connection object according to the type of database used, and enter the database host address, port, user name, and password information;

[0071] S13: Create a cursor object: used to execute SQL queries and retrieve results;

[0072] S14: Use the execute() method of the cursor object to execute the SQL query statement;

[0073] S15: Use the fetchall(), fetchone() or fetchmany() method of the cursor object to obtain the query result;

[0074] S16: Close the cursor and connection to release resources and obtain historical sample data of hot network users.

[0075] Figure 2 To invent the algorithm flow chart for medium and high density areas;

[0076] The process of using the high-density area method to detect abnormal data in the historical sample data of heating network users, marking the abnormal data, and deleting the abnormal data is as follows:

[0077] S21: First, define two important parameters for the high-density region algorithm: ε and MinPts; ε is used to define the neighborhood range of the data point, and MinPts defines the minimum number of neighbors required for a core point;

[0078] S22: Finding core points: For the cumulative heat consumption of heat users every hour, a given data set is established. The number of data points in the ε neighborhood of each data point is calculated, and it is determined whether the MinPts condition is met. If the ε neighborhood of a data point contains at least MinPts neighbors, the data point is considered to be a core point.

[0079] In density clustering algorithms (such as DBSCAN), judging whether the `MinPts` condition is met refers to determining whether there are enough points in the neighborhood of a point to make it considered as a core point.

[0080] The epsilon neighborhood: For a given data point P, its epsilon neighborhood refers to the set of all points whose Euclidean distance from P is less than or equal to epsilon;

[0081] MinPts: This refers to the minimum number of points (including point P itself) that must be included in the epsilon neighborhood of a point so that the point is considered a core point.

[0082] The core point: If the number of points in the epsilon neighborhood of a point is greater than or equal to `MinPts`, then the point is a core point (Core Point).

[0083] The border point: If a point itself is not a core point, but it is located in the epsilon neighborhood of a core point, then it is a border point.

[0084] The noise point: a point that is neither a core point nor a boundary point is regarded as a noise point.

[0085] How to determine whether the `MinPts` condition is met is as follows:

[0086] S221: Calculate the number of points in the epsilon neighborhood: For each point P, calculate the number of points in its epsilon neighborhood.

[0087] S222: Determine whether the condition of `MinPts` is met:

[0088] If the number of points in the epsilon neighborhood (including P itself) is greater than or equal to `MinPts`, then P is a core point;

[0089] If the number of points in the epsilon neighborhood is less than `MinPts`, then P is not a core point;

[0090] S23: Marking boundary points: For a data point in the ε neighborhood of a core point, if the number of data points in its neighborhood is less than MinPts but contains a core point, the data point is marked as a boundary point;

[0091] S24: Create clusters: Create one or more clusters by connecting the core points and the data points in their ε neighborhood. If the ε neighborhoods of two core points overlap, they will be classified into the same cluster. The remaining data points that are not classified into any cluster are classified as outliers;

[0092] S25: Eliminate data marked as outliers.

[0093] Furthermore, the process of using the multiple interpolation method to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of the heating network users to obtain the completed historical data of the heating network users is as follows:

[0094] S31: establishing an interpolation function through known values, estimating the value to be interpolated, and then adding different deviations to the value to form multiple sets of optional interpolation values, thereby forming multiple sets of complete data sets to be evaluated;

[0095] Wherein: the known value is the accumulated calories per hour of the user in the normal data with normal data and abnormal data;

[0096] Common interpolation functions include:

[0097] Mean interpolation: fill missing values ​​with the mean of the variable;

[0098] Regression interpolation: predicting missing values ​​using regression models;

[0099] Multiple Imputation Chained Equations (MICE): Iteratively fills in missing values ​​using a regression model;

[0100] Bayesian imputation: Generate interpolation values ​​for missing data based on Bayesian methods;

[0101] k-nearest neighbor interpolation (k-NN): fill missing values ​​with the mean of the k samples closest to the missing value;

[0102] Random Forest Imputation: Predict missing values ​​using a random forest model;

[0103] In multiple imputation, multiple (e.g., m) imputations are usually performed, and each imputation produces a set of estimated parameters. For the estimated parameters obtained from each imputation, its variance can be calculated. The variance estimate of each imputed sample is calculated based on the imputed data set, so it may be different;

[0104] Parameter Estimation is a core concept in statistics, which refers to the process of using sample data to infer the value of a population parameter. A population parameter is a constant that describes the characteristics of a population (such as the population mean, population variance, etc.), while an estimated parameter is a statistic used to approximate these population parameters. Since it is usually impossible to obtain data on the entire population in practice, sample data is used to estimate population parameters;

[0105] The forming of multiple sets of optional interpolation values ​​to form multiple sets of complete data sets to be evaluated includes:

[0106] Generate multiple (usually 5 to 10) imputed datasets using the chosen imputation algorithm, each of which is imputed using the same method but with different randomness introduced to reflect uncertainty;

[0107] In the multiple imputation process, different randomness is introduced to ensure that each generated imputed data set reflects the uncertainty in the data. This randomness allows a certain variability between multiple imputed data sets, rather than simply replicating the same imputation results.

[0108] Here are some common ways to introduce different kinds of randomness:

[0109] 1. Introducing random errors in regression interpolation

[0110] When using a regression model to predict missing values, you can add a random error term to the regression model. Specifically, when calculating the interpolated value, not only the predicted value of the regression equation is used, but also an error term randomly drawn from a normal distribution (or other appropriate distribution) is added to the predicted value.

[0111] 2. Random Sampling in Bayesian Methods

[0112] When using Bayesian imputation, randomness can be introduced by randomly sampling from the posterior distribution. This is done by first computing the posterior distribution based on the known data and the prior distribution, then randomly sampling parameters from the posterior distribution and generating imputed values ​​based on these parameters.

[0113] This method ensures that the model parameters used for each imputation are different, resulting in variation in the imputation results.

[0114] 3. Random Start in Multiple Imputation Chained Equations (MICE)

[0115] In the MICE method, the initial imputation is usually performed by simple methods such as mean imputation. In order to introduce randomness, a random start can be used in this initial stage, that is, the initial imputation values ​​are determined by random sampling.

[0116] In addition, in each iteration of MICE, model parameters can also be selected by random sampling to ensure that the interpolation results of each iteration are different.

[0117] 4. Introducing randomness in nonparametric interpolation methods

[0118] For non-parametric imputation methods such as k-nearest neighbor imputation or random forest imputation, randomness can be introduced when choosing the imputation value. For example, in k-nearest neighbor imputation, a value can be randomly drawn from the nearest few neighbors instead of simply taking their average.

[0119] The random forest imputation method can ensure differences between different imputed datasets by generating different samples for each tree (by introducing randomness or bootstrapping).

[0120] 5. Random Group Imputation

[0121] Another way to introduce randomness is to divide the dataset into multiple random subsets and perform independent imputation on each subset. Then, the complete imputed dataset is generated by combining these subsets. Due to the randomness of the subsets, the results of different imputed datasets will be different.

[0122] 6. Uncertainty in parameter estimates

[0123] During the imputation process, randomness can be introduced by accounting for uncertainty in the model parameters. For example, in regression imputation, the regression coefficients can be randomly sampled from their estimated distribution (such as the normal distribution) rather than using fixed estimates.

[0124] 7. Use of random seeds

[0125] The simplest method is to use different random seeds in the interpolation algorithm. By changing the random seed, the random number sequence generated each time the interpolation algorithm is run will be different, which will affect the interpolation results.

[0126] S32: performing statistical analysis on the generated data set;

[0127] The purpose of statistical analysis of the data sets generated after multiple imputation is to comprehensively estimate the results of each imputed data set and evaluate the robustness of statistical inference. This process can be carried out through the following steps:

[0128] 1. Perform single data set analysis

[0129] First, perform the required statistical analysis on each imputed data set independently. This may include regression analysis, analysis of variance, correlation test, etc. Each analysis should be completed independently to obtain the estimated results (such as regression coefficients, means, proportions, etc.) for each data set.

[0130] 2. Summarize the analysis results

[0131] Summarize the analysis results obtained for each imputed data set. A common method is to calculate the mean of these estimates as the final estimate. This step is called summary analysis and involves the following key statistics:

[0132] Within-imputation variance (\(\overline{W}\)): Calculate the variance of the estimated results for each data set and find the mean of these variances.

[0133] Between imputation variance (B): Calculates the variance between the means of the estimates of different imputed data sets.

[0134] Total variance (\(T\)): The total variance within the interpolation and the variance between the interpolations.

[0135] 3. Calculate comprehensive statistics

[0136] Point estimate: The mean of all estimated values ​​of the imputed data set is used as the final statistical estimate.

[0137] Confidence intervals: Use the total variance \(T\) to calculate the confidence interval of the estimate to ensure the probability of containing the true value.

[0138] Hypothesis testing: Use the combined estimate and total variance to make statistical inferences, for example, using a t-test to assess whether a regression coefficient is significantly different from zero.

[0139] 4. Graphical display

[0140] In order to compare and display the data more intuitively, the following graphs can be drawn:

[0141] Forest plots: Displays the estimated value and its confidence interval for each imputed dataset, as well as the overall estimate.

[0142] Histograms and boxplots: Display the distribution of estimates for different imputed datasets and check for consistency and outliers.

[0143] 5. Sensitivity analysis

[0144] Sensitivity analysis was performed to evaluate the impact of the choice of imputation method on the final results. The results produced by different imputation techniques (such as MICE, Bayesian imputation, etc.) were compared, and how changing the number of imputation datasets affected the final estimates.

[0145] 6. Model Diagnosis

[0146] Diagnostics were performed on the statistical model used for each imputed data set to check whether the model assumptions were met, for example, whether the residuals of the linear regression model were normally distributed and homoscedastic.

[0147] S33: Evaluate the results for each data set and select the best interpolation value based on the model's score.

[0148] The results of evaluating each data set are described, and the best interpolation value is selected according to the model's score.

[0149] Integrate the analysis results of multiple imputed data sets. Specifically, it includes the following steps:

[0150] Calculate the mean of the estimates for each imputed data set: get the overall estimate.

[0151] Calculate the within-imputation variance ($\overline{W}$): the average of the variances of the estimates for each imputed dataset.

[0152] Calculate the between-imputation variance ($B$): that is, the variance between the estimates of different imputed data sets.

[0153] Calculate the total variance ($T$): Combine the intra-interpolation variance and the inter-interpolation variance to get the total variance.

[0154] Estimated degrees of freedom: The degrees of freedom are calculated based on the within-imputation variance and the between-imputation variance for statistical inference.

[0155] Further: the completed historical sample data of the heating network users is standardized by the Min-max method, and the process of converting the sample data into values ​​in the same interval is as follows: the following method is adopted:

[0156] S41: The data was standardized by using the Min-max method, and the calculation formula is as follows:

[0157]

[0158] In addition, the output value needs to be denormalized, and the calculation formula is as follows:

[0159] x=x nor (x max -x min )+x min (2)

[0160] In the above formula, x is the sample data, x max 、x min are the maximum and minimum values ​​of the sample data, respectively, nor The sample data is standardized.

[0161] Figure 3 Flow chart of the k-means clustering algorithm in the present invention;

[0162] The process of using the K-means clustering algorithm to perform cluster analysis on sample data converted into numerical values ​​in the same interval based on the user's energy usage habits, work and rest time, and temperature preference to obtain the energy usage classification of the heat network is as follows:

[0163] S51: Randomly select k samples as the initial cluster centers C = {c1, c2 ... c k};

[0164] S52: For each sample x in the data set i , calculate the distances between the sample and the K cluster centers respectively, and divide it into the class corresponding to the cluster center with the smallest distance;

[0165] The K-means clustering algorithm takes the number of clusters k as an input parameter, divides n sample data objects into k categories, and makes the similarity of data points in each category as large as possible, while the similarity of data points between different categories as small as possible.

[0166] Suppose there are n sample data X = {x1, x2, ..., x n}∈R d is the data set to be clustered, where X j =(x j1 , x j2 , …, x jd ) T is a d-dimensional vector. The K-means algorithm is to find a set C containing k cluster centers = (c1, c2, ..., c k ) T , minimize the objective function:

[0167]

[0168] Where S i represents the set of samples in the i-th category, c i YesS i All samples x j The cluster center point, d(x j , c i ) represents the sample data x j With cluster center c i The Euclidean distance between is defined as follows:

[0169]

[0170] represents the center position of the i-th category, i = 1, 2, ..., k, n i It is c i The number of sample data in x j Represents class c i Sample data in .

[0171] The flowchart of the k-means clustering algorithm is as follows Figure 1 As shown in the figure, the specific steps of the algorithm are as follows:

[0172] Step 1: Randomly select k sample data, each sample data represents an initial cluster center, the set of initial cluster centers is C = (c1, c2, ..., c k ) T ;

[0173] Step 2: According to the Euclidean distance formula, for each remaining sample data, calculate the distance from each data sample to each cluster center, and then divide each data sample into the category with the minimum distance;

[0174] Step 3: Recalculate the values ​​of k cluster centers according to formula (5);

[0175] Step 4: Check whether the clustering operation needs to be ended. If the objective function formula (1) is minimized or remains unchanged, the iteration ends.

[0176] S53: For each category c j , recalculate its cluster centers:

[0177]

[0178] Where: n is category c j The number of data points contained in ;

[0179] S54: Repeat the above two steps S52 and S53 until the cluster center no longer changes, and output the clustering result.

[0180] Based on the user's energy usage habits, work and rest time, and temperature preference, cluster analysis is performed to obtain the energy usage classification process of the power grid as follows:

[0181] S61: Determine clustering variables; the clustering variables include:

[0182] Energy usage habits: may include daily energy consumption, peak and valley hours, and the usage ratio of energy types (such as electricity, hot water, etc.).

[0183] Work and rest time: includes the user's living time, active time (such as working time, rest time), etc.

[0184] Temperature preference: User preference for indoor temperature (such as heating temperature in winter and air conditioning temperature setting in summer).

[0185] S62: performing clustering result analysis;

[0186] Determine cluster centers: Check the characteristic values ​​of each cluster center and analyze the typical user profile of each category. For example, users of a category may have peak energy consumption at night and have a higher temperature preference.

[0187] S63: Output classification results

[0188] User category label: assign a corresponding category label to each user.

[0189] Category description: Provide a detailed description of the user characteristics of each category, such as:

[0190] Category 1: Peak energy consumption at night, higher temperature preference, probably mainly household users.

[0191] Category 2: High energy usage during the day, medium temperature preference, probably office buildings or commercial users.

[0192] S64: Energy Usage Analysis

[0193] Energy usage patterns: Analyze the energy usage patterns of each category according to the category. For example, users in a certain category may rely more on nighttime heating, which may affect the nighttime load of the heating network.

[0194] Load forecasting: Use clustering results to predict load changes in different time periods to help optimize heating network scheduling and energy distribution.

[0195] S65: Visualization Output

[0196] Cluster diagram: shows the distribution of users of different categories in the feature space (such as 2D / 3D scatter plot after PCA dimensionality reduction).

[0197] Energy usage curve: Displays the energy usage curves of different categories of users on typical days (such as weekdays and weekends).

[0198] Load distribution diagram: shows the load contribution of different types of users in the heating network, helping to identify high-load periods and key user categories.

[0199] Application example: Analysis of heating habits of users of a city heating company

[0200] Embodiment 1:

[0201] Background: A city heating company hopes to optimize heating services by analyzing users' heating habits to develop differentiated methods, improve heating efficiency and increase user satisfaction.

[0202] A big data driven intelligent energy saving and carbon reduction method based on demand and supply includes the following steps:

[0203] S1. Data collection

[0204] Data sources: smart heat meter data, hourly heat consumption, user’s geographic location, weather data (temperature, humidity, etc.).

[0205] S2: Data preprocessing: Use high-density area algorithm to detect abnormal data in historical sample data of heating network users, mark abnormal data, and delete abnormal data; Process missing values: Use multiple interpolation method to fill missing values;

[0206] S3: Use multiple interpolation methods to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of heating network users to obtain the completed historical data of heating network users;

[0207] Standardization: Standardize the heat consumption per hour to eliminate the magnitude differences between different users.

[0208] S4: standardize the completed historical sample data of heating network users by using the Min-max method, and convert the sample data into values ​​in the same interval;

[0209] S5: The K-means clustering algorithm is used to perform cluster analysis on the sample data converted into numerical values ​​in the same interval, based on the user's energy usage habits, work and rest time, and temperature preference, to obtain the energy usage classification of the heating network.

[0210] The feature engineering of this application is used to label the classified data;

[0211] Temporal feature extraction:

[0212] Total daily calorie intake.

[0213] Mean and variance of heat consumption per hour.

[0214] Peak heat usage hours and heat usage.

[0215] Difference in heat usage between weekdays and weekends.

[0216] User behavior characteristics:

[0217] Peak heat consumption in the morning and evening.

[0218] Changes in heat usage during holidays.

[0219] Seasonal changes (differences in heat consumption between winter, spring, and autumn).

[0220] Weather characteristics:

[0221] The correlation between temperature and heat consumption.

[0222] Correlation between humidity and heat usage.

[0223] The K-means clustering algorithm is used to perform cluster analysis on the sample data converted into numerical values ​​in the same interval based on the user's energy usage habits, work and rest time, and temperature preferences, and the energy usage of the heating network is classified as follows:

[0224] Determine the number of clusters K: K is determined to be 3 through the elbow rule, indicating that users are divided into 3 categories.

[0225] Execute the K-means algorithm: cluster the user's thermal behavior data into three groups.

[0226] Clustering result analysis

[0227] Group 1: High calorie users

[0228] Characteristics: High heat consumption throughout the day, with a significant peak in winter.

[0229] Typical users: large office buildings, commercial complexes.

[0230] Group 2: Users with high calorie consumption on weekdays

[0231] Characteristics: There are obvious peaks in the morning and evening on weekdays, and relatively low heat consumption on weekends.

[0232] Typical users: ordinary families and white-collar workers.

[0233] Group 3: Low calorie users

[0234] Features: The overall heat consumption is low, with no obvious peak period.

[0235] Typical users: single apartments, vacation homes, etc.

[0236] Result verification and application

[0237] Result verification: Use historical data to verify the clustering results, such as checking the actual heat usage data of each group to ensure clustering accuracy.

[0238] Application scenarios:

[0239] Differentiated pricing:

[0240] High heat users: Provide flexible pricing plans, such as time-of-day pricing, to encourage users to use heat during off-peak hours.

[0241] For users with high heat consumption on weekdays: formulate different pricing methods for weekdays and weekends to optimize the allocation of heating resources.

[0242] Low-heat-consumption users: provide preferential packages to encourage more users to use heating rationally.

[0243] Energy saving suggestions:

[0244] For users with high heat consumption: it is recommended to install intelligent temperature control equipment to optimize heat settings.

[0245] For users who use high amounts of heat on weekdays: it is recommended to adjust heating usage time to avoid peak hours.

[0246] Low heat users: Provide energy-saving awareness education and promote energy-saving equipment.

[0247] Equipment maintenance:

[0248] High heat users: Check heating equipment regularly to ensure it operates efficiently.

[0249] For users with high heat consumption on weekdays: Carry out equipment maintenance before peak hours to ensure heating quality.

[0250] Low-heat users: Perform regular inspections to prevent equipment failures caused by long-term idleness.

[0251] Personalized service:

[0252] For users with high heat consumption: a dedicated account manager is provided to monitor heat consumption in real time and provide optimization suggestions.

[0253] For users who use a lot of heat on weekdays: Provide personalized heat usage reports to help users understand their heat usage habits.

[0254] Low heat consumption users: heat consumption optimization suggestions are pushed via SMS or APP to improve user experience.

[0255] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 replace some or all of the technical features therein with equivalents. 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 invention.

Claims

1. A big data driven, demand-based intelligent energy-saving and carbon reduction method, characterized by: The following steps are involved: S1: Obtain historical sample data of heating network users to be processed; S2: Use a high-density area algorithm to detect abnormal data in the historical sample data of heating network users, mark the abnormal data, and delete the abnormal data; S3: Use multiple interpolation methods to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of heating network users to obtain the completed historical data of heating network users; S4: standardize the completed historical sample data of heating network users by using the Min-max method, and convert the sample data into values ​​in the same interval; S5: The K-means clustering algorithm is used to perform cluster analysis on the sample data converted into numerical values ​​in the same interval, based on the user's energy usage habits, work and rest time, and temperature preference, to obtain the energy usage classification of the heating network.

2. According to claim 1, a big data driven intelligent energy saving and carbon reduction method based on demand is characterized by: The process of using the high-density area method to detect abnormal data in the historical sample data of heating network users, marking the abnormal data, and deleting the abnormal data is as follows: S21: First, define two important parameters for the high-density region algorithm: ε and MinPts; ε is used to define the neighborhood range of the data point, and MinPts defines the minimum number of neighbors required for a core point; S22: Finding core points: A given data set is established based on the cumulative heat consumption of heat users every hour. The number of data points in the ε neighborhood of each data point is calculated, and it is determined whether the MinPts condition is met. If the ε neighborhood of a data point contains at least MinPts neighbors, the data point is considered to be a core point. S23: Marking boundary points: For a data point in the ε neighborhood of a core point, if the number of data points in its neighborhood is less than MinPts but contains a core point, the data point is marked as a boundary point; S24: Create clusters: Create one or more clusters by connecting the core points and the data points in their ε neighborhood. If the ε neighborhoods of two core points overlap, they will be classified into the same cluster. The remaining data points that are not classified into any cluster are classified as outliers; S25: Eliminate data marked as outliers.

3. According to claim 1, a big data driven intelligent energy saving and carbon reduction method based on demand is characterized by: The process of using the multiple interpolation method to supplement the missing values ​​and abnormal values ​​in the deleted historical sample data of the heating network users to obtain the completed historical data of the heating network users is as follows: S31: establishing an interpolation function through known values, estimating the value to be interpolated, and then adding different deviations to the value to form multiple sets of optional interpolation values, thereby forming multiple sets of complete data sets to be evaluated; S32: performing statistical analysis on multiple sets of data sets to be evaluated; S33: Evaluate the results for each data set and select the best interpolation value based on the score of the statistical model.

4. According to claim 1, the name of the intelligent energy saving and carbon reduction method driven by big data and based on demand is characterized by: The Min-max method is used to standardize the completed historical sample data of the heating network users, and the process of converting the sample data into values ​​in the same interval is as follows: S41: The data was standardized by using the Min-max method, and the calculation formula is as follows: In addition, the output value needs to be denormalized, and the calculation formula is as follows: x=x nor (x max -x min )+x min , (2) In the above formula, x is the sample data, x max 、x min are the maximum and minimum values ​​of the sample data, respectively, nor The sample data is standardized.

5. According to claim 1, the method for intelligent energy saving and carbon reduction based on demand and driven by big data is characterized by: The process of converting sample data into numerical values ​​of the same interval by the K-means clustering algorithm is as follows: S51: Randomly select k samples as the initial clustering center C = {c1, c2...c k }; S52: For each sample x in the data set i , calculate the distances between the sample and the K cluster centers respectively, and divide it into the class corresponding to the cluster center with the smallest distance; S53: Recalculate each category c j , the cluster center of: Where n is the category c j The number of data points contained in ; S54: Repeat the above two steps S52 and S53 until the cluster center no longer changes, and output the clustering result.

6. According to claim 1, the method for intelligent energy saving and carbon reduction based on demand and driven by big data is characterized by: The process of clustering analysis based on the user's energy usage habits, work and rest time and temperature preference to obtain the energy usage classification of the power grid is as follows: S61: Determine clustering variables; the clustering variables include: Energy usage habits: including daily energy consumption, peak and valley hours, and energy types; Work and rest time: including the user's living time and active time; Temperature preference: user's preference for indoor temperature; S62: performing clustering result analysis; Determine cluster centers: Check the characteristic values ​​of each cluster center and analyze the typical user profile of each category; S63: Output classification results; Label users with category labels: label each user with a corresponding category label; Conduct category description: Describe the user characteristics of each category in detail; S64: Analysis of energy usage; Energy usage pattern: Analyze the energy usage pattern of each category according to the category; Load forecasting: Use clustering results to predict load changes in different time periods to help optimize heating network scheduling and energy distribution.

Citation Information

Cited By

  • Health degree detection method and device of offline store and storage medium

    CN120746407A