Cogeneration user group energy efficiency optimization method and system based on bidirectional collaboration

Abstract

The application discloses a heat and power cogeneration user group energy efficiency optimization method and system based on bidirectional cooperation, relates to the technical field of heat and power cogeneration, and proposes the following scheme, which comprises setting a historical period, collecting historical power and heat data of users in the period, constructing a strategy feature vector, and the strategy feature vector comprising power and heat behavior characteristics, demand response participation characteristics and economic benefit characteristics; a clustering algorithm is used to classify the user strategy feature vector, identify information island users and main group users; and historical power and heat data of the island users are used. Through the historical power and heat data, three-dimensional characteristics of power consumption behavior, demand response participation and economic benefit are extracted, a strategy feature vector is constructed, accurate description of the power and heat behavior of users is realized, the limitation of traditional single-dimensional analysis is broken through, accurate data basis is provided for user clustering and strategy optimization, the accuracy of feature recognition is improved, and the system can accurately analyze the power and heat rules.

Description

Technical Field

[0001] This invention relates to the field of cogeneration technology, specifically to a method and system for optimizing the energy efficiency of cogeneration user groups based on bidirectional synergy. Background Technology

[0002] In the field of combined heat and power (CHP) technology, especially in scenarios involving energy efficiency optimization for user groups, existing technologies mostly rely on historical electricity and heat consumption data of users to carry out strategy formulation. The core idea is to analyze basic data such as electricity load curves and heat demand trends, extract simple statistical characteristics such as average, peak, and valley values ​​of electricity load, and combine them with electricity pricing policies to formulate generalized energy allocation and dispatch strategies for the main user group.

[0003] However, existing technologies still have significant shortcomings: on the one hand, the characterization of user features is mostly limited to a single dimension, failing to comprehensively consider the deep-seated patterns and benefit relationships of user behavior, resulting in overly coarse user classification and difficulty in accurately matching optimization strategies; on the other hand, the segmentation of user groups mostly adopts clustering methods with fixed parameters, which cannot effectively identify isolated users with special behavioral patterns, and the strategy optimization is mostly unidirectional, ignoring the potential value of high-efficiency strategies among isolated users, which easily leads the main group strategy to get trapped in local optima. Therefore, we propose a bidirectional collaborative cogeneration user group energy efficiency optimization method and system to solve this problem. Summary of the Invention

[0004] To address the aforementioned technical issues, this paper provides a method and system for optimizing the energy efficiency of cogeneration user groups based on bidirectional collaboration. This solution overcomes the significant shortcomings of existing technologies mentioned in the background: firstly, the characterization of user features is often limited to a single dimension, failing to comprehensively consider the deep-seated patterns and benefit relationships of user behavior, resulting in overly coarse user classification and difficulty in accurately matching optimization strategies; secondly, the segmentation of user groups often employs fixed-parameter clustering methods, which cannot effectively identify isolated users with unique behavioral patterns, and strategy optimization is often unidirectional, neglecting the potential value of high-efficiency strategies among isolated users, easily leading the main group's strategy into local optima.

[0005] To achieve the above objectives, the technical solution adopted by this invention is as follows: This invention provides a method for optimizing the energy efficiency of a cogeneration user group based on bidirectional synergy, the method comprising:

[0006] Set a historical period, collect historical electricity and heat consumption data of users within the period, and construct a strategy feature vector. The strategy feature vector includes electricity and heat consumption behavior characteristics, demand response participation characteristics, and economic benefit characteristics.

[0007] Clustering algorithms are used to classify user strategy feature vectors to identify users in information silos and main user groups;

[0008] Based on the historical electricity and heat consumption data of isolated users, the monthly load variance and the frequency of abnormal electricity price response are calculated, and the historical strategy paths of similar main user groups are matched.

[0009] Guide isolated users to converge to the historical strategy paths of the main group of users, while using strategies of isolated users that are more effective than the preset potential value to reverse correct the historical strategy path library of the main group.

[0010] Establish a dynamic feedback mechanism to monitor the implementation effect of the historical strategy path correction for the main user group in real time, calculate the comprehensive efficiency index of the corrected path, and optimize the reference group path when the index is lower than the preset benchmark value for three consecutive periods.

[0011] Furthermore, a bidirectional collaborative cogeneration user group energy efficiency optimization system is proposed to implement the bidirectional collaborative cogeneration user group energy efficiency optimization method described above, including:

[0012] The strategy feature vector construction module is used to set a historical period, collect historical electricity and heat consumption data of users within that period, and construct a strategy feature vector. The strategy feature vector includes electricity and heat consumption behavior characteristics, demand response participation characteristics, and economic benefit characteristics.

[0013] The user group classification module is used to classify user strategy feature vectors using clustering algorithms to identify users in information silos and main user groups.

[0014] The strategy path matching module is used to calculate the monthly load variance and the frequency of abnormal electricity price response based on the historical electricity and heat consumption data of isolated users, and to match the historical strategy paths of similar main groups of users.

[0015] The bidirectional collaborative optimization module is used to guide isolated users to converge to the historical strategy paths of the main group of users, while using strategies of isolated users with higher efficiency than the preset potential value to reverse correct the historical strategy path library of the main group.

[0016] The dynamic feedback and iteration module is used to establish a dynamic feedback mechanism to monitor the implementation effect of the historical strategy path correction for the main user group in real time, calculate the comprehensive efficiency index of the corrected path, and optimize the reference group path when the index is lower than the preset benchmark value for three consecutive periods.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] 1. By extracting three-dimensional features of electricity consumption behavior, demand response participation, and economic benefits from historical electricity and heat consumption data, a strategy feature vector is constructed to accurately depict users' electricity and heat consumption behavior. This breaks through the limitations of traditional single-dimensional analysis, provides a precise data foundation for user clustering and strategy optimization, improves the accuracy of feature recognition, and helps the system accurately analyze electricity and heat consumption patterns.

[0019] 2. By matching similar main group paths, isolated users are guided to converge towards efficient strategies. At the same time, strategies with efficiency exceeding the preset potential value among isolated users are used to correct the main group path library. This two-way collaboration not only eliminates inefficient behaviors of isolated users, but also discovers niche high-quality strategies to feed back to the main group, avoiding local optima in the main group path and improving the comprehensiveness and efficiency of the system strategy.

[0020] 3. Based on the reference group path, the load adjustment range and electricity price trigger threshold are extracted. When the real-time electricity price exceeds the threshold, adjustment suggestions are pushed. Combined with time-series electricity cost calculation, the economic benefits are optimized, the timeliness and accuracy of users' response to electricity price signals are improved, the system operating pressure is reduced and users' electricity costs are reduced. Attached Figure Description

[0021] Figure 1 This is a flowchart of an energy efficiency optimization method for cogeneration user groups based on bidirectional collaboration proposed in this invention;

[0022] Figure 2 This is a flowchart of the method for constructing strategy feature vectors in this invention;

[0023] Figure 3 This is a flowchart of the method for matching historical strategy paths in this invention;

[0024] Figure 4 This is a structural block diagram of a bidirectional collaborative cogeneration user group energy efficiency optimization system proposed in this invention. Detailed Implementation

[0025] The following description is intended to disclose the invention so that those skilled in the art can implement it. The preferred embodiments described below are merely examples, and other obvious modifications will occur to those skilled in the art.

[0026] Reference Figure 1-3 As shown, a method for optimizing the energy efficiency of cogeneration user groups based on bidirectional collaboration is described, the method comprising:

[0027] Set a historical period, collect historical electricity and heat consumption data of users within the period, and construct a strategy feature vector. The strategy feature vector includes electricity and heat consumption behavior characteristics, demand response participation characteristics, and economic benefit characteristics.

[0028] Clustering algorithms are used to classify user strategy feature vectors to identify users in information silos and main user groups;

[0029] Based on the historical electricity and heat consumption data of isolated users, the monthly load variance and the frequency of abnormal electricity price response are calculated, and the historical strategy paths of similar main user groups are matched.

[0030] Guide isolated users to converge to the historical strategy paths of the main group of users, while using strategies of isolated users that are more effective than the preset potential value to reverse correct the historical strategy path library of the main group.

[0031] Establish a dynamic feedback mechanism to monitor the implementation effect of the historical strategy path correction for the main user group in real time, calculate the comprehensive efficiency index of the corrected path, and optimize the reference group path when the index is lower than the preset benchmark value for three consecutive periods.

[0032] Reference Figure 2 As shown, the construction strategy feature vector specifically includes:

[0033] Based on historical electricity and heat consumption data, electricity load curves and heat demand data curves are proposed, and corresponding statistical indicators and time periodic characteristics are extracted.

[0034] The statistical indicators include mean, variance, peak load and valley ratio; the time periodicity features include the frequency components and phase of the main period amplitude; and the electricity and heat consumption behavior features are jointly characterized by statistical indicators and time series periodicity features.

[0035] For example, the historical period is set to one year. The power load curve is plotted using a sampling frequency of 15 minutes / point. The daily average load is calculated as the mean, the daily load variance is statistically analyzed, and the monthly highest load and the percentage of periods with load below the mean within the year are recorded. Fourier transform is performed on the load curve to extract the daily dominant frequency component, and the peak occurrence time is determined through phase analysis. For example, 14:00 every day is the starting phase of the peak electricity consumption. The characteristics of electricity and heat consumption behavior can be represented as [mean = 12.5, variance = 8.2, peak value = 28, valley percentage = 0.35, dominant frequency amplitude = 0.6, phase = 14];

[0036] The participation rate, average response time, and average response volume of users in the electricity price response project were statistically analyzed within the historical period. The participation rate is the ratio of the number of responses to the number of electricity price adjustments.

[0037] The demand response participation characteristic is jointly characterized by participation rate, average response time, and average response volume.

[0038] For example, the historical period is one year, that is, the electricity price is adjusted 30 times in a year, and the number of user responses is 15 times, so the participation rate is 50%. At the same time, the average delay time from the electricity price announcement to the load adjustment and the average load change for each response are recorded. The demand response participation characteristics can be expressed as [participation rate = 0.5, average response time = 2.3, average response amount = 4.5];

[0039] Based on real-time electricity prices and subsidies for response projects, the actual electricity prices for different time periods are obtained. Combined with historical electricity and heat consumption data, the time-series electricity cost of users in that historical period is obtained and used as an economic benefit characteristic.

[0040] For example, real-time electricity prices are obtained, such as RMB 1.2 / kWh during peak hours and RMB 0.5 / kWh during off-peak hours. The subsidy for the responding project is RMB 0.05 / kWh. The proportion of each time period in the historical electricity and heat consumption data within a day is obtained, thereby calculating the electricity cost for each time period within that day, accumulating the monthly electricity cost, and arranging it into time-series electricity cost in monthly order.

[0041] Specifically, the method of using clustering algorithms to classify user strategy feature vectors and identify users in information silos and main user groups includes:

[0042] Based on the characteristics of the user group of the cogeneration system, the range of values ​​for the number of clusters is set, and the K-means++ algorithm is used to select the initial cluster center for each cluster number value.

[0043] For example, in real-world scenarios, users can typically be divided into 3-5 typical groups, such as high-response users, low-response users, seasonally fluctuating users, and information silo users. The number of clusters can be appropriately expanded to 2-6 on this basis.

[0044] Calculate the Euclidean distance from each user's policy feature vector to each cluster center, and assign it to the nearest cluster. Update the cluster centers until convergence.

[0045] Calculate the profile coefficient of each user in the clustering results, plot the average profile coefficient curve corresponding to different cluster number values, and select the cluster number value that maximizes the profile coefficient as the final cluster number.

[0046] If the average profile coefficient of a user group is lower than a set threshold, or the number of users in the group is less than a set proportion, then the group is determined to be an information island user group; if neither is true, then the group is determined to be a main user group.

[0047] For example, the threshold can be set to 0.2, and the percentage can be set to 5%.

[0048] The formula for calculating the contour coefficient is as follows:

[0049]

[0050] In the formula, s(i) is the silhouette coefficient, b(i) is the average Euclidean distance between user i and other users in the same cluster, and a(i) is the minimum average Euclidean distance between user i and other clusters.

[0051] It should be noted that user strategy characteristics in cogeneration systems are complex. If a fixed number of clusters is used directly, some isolated users may be misclassified into normal user clusters, or normal users may be misjudged as isolated users due to characteristic fluctuations. Using a number of clusters that maximizes the profile coefficient can accurately capture the real isolated user groups while avoiding over-segmentation.

[0052] Reference Figure 3 As shown, the calculation of monthly load variance and abnormal electricity price response frequency based on historical electricity and heat consumption data of isolated users, and matching historical strategy paths of similar main user groups, specifically includes:

[0053] Extract historical electricity and heat consumption data of isolated users, calculate their monthly load variance, set a delay threshold, and calculate the frequency of abnormal electricity price response by combining demand response participation characteristics.

[0054] The monthly load variance and abnormal frequency of electricity price response of the main user group are vectorized to form a historical strategy path library. Each path contains seasonal fluctuation patterns and electricity price response behavior.

[0055] It should be noted that the seasonal fluctuation pattern refers to the cyclical changes in user electricity demand due to seasonal, climatic, and holiday factors throughout historical periods. Its quantitative indicators include monthly load variance, intraday peak-to-valley difference rate, and seasonality index. The seasonality index is extracted by decomposing historical electricity and heat consumption data through time series analysis. Electricity price response behavior refers to the user's tendency to adjust energy consumption strategies based on electricity price signals. Its quantitative indicators are the electricity price elasticity coefficient, the frequency of abnormal electricity price response, and the delay time. The electricity price elasticity coefficient is the ratio of electricity consumption to electricity price, and the delay time is the difference between the user's strategy adjustment time and the electricity price announcement time.

[0056] The dynamic time warping algorithm is used to calculate the similarity of feature sequences between isolated users and the main group of users, and the main group of users with the highest similarity are selected as the reference group.

[0057] Based on the historical policy paths of the reference group, the load adjustment magnitude and electricity price adjustment trigger threshold are extracted to generate the policy adjustment baseline for isolated users.

[0058] It should be noted that the load adjustment range is the range of load changes for isolated users, and the electricity price adjustment trigger threshold is the critical value for users' response to changes in electricity prices. When real-time electricity price fluctuations exceed this threshold, users need to trigger policy adjustments.

[0059] It should be noted that the monthly load variance is used to lock in the long-term patterns of users, and the abnormal frequency of electricity price response is used to verify the consistency of user strategies. The combination of the two can efficiently screen out the main user group with similar behavior to isolated users and reliable strategies.

[0060] Specifically, the process of guiding isolated users to converge towards the historical strategy paths of the main group of users, while simultaneously using strategies among isolated users that have higher efficiency than a preset potential value to reverse-correct the historical strategy path library of the main group, includes:

[0061] Real-time collection of current electricity and heat consumption data and electricity price response project data from isolated users, and extraction of real-time electricity price from the electricity price response project data;

[0062] If the real-time electricity price exceeds the electricity price adjustment trigger threshold in the strategy adjustment baseline, the load adjustment magnitude will be pushed to the isolated user.

[0063] After users adjust their electricity consumption behavior according to the load adjustment range, the ratio of their monthly load data to their monthly electricity cost is used as the user efficiency.

[0064] The strategy adjustment baseline with user performance exceeding the preset potential value is weighted and merged with the corresponding main group's strategy path to obtain the corrected strategy path, which is then included in the main group's historical strategy path library; for example, the preset potential value can be 10%.

[0065] Specifically, the establishment of a dynamic feedback mechanism monitors the implementation effect of the historical strategy path correction for the main user group in real time, calculates the comprehensive efficiency index of the corrected path, and optimizes the reference group path when the index is lower than the preset benchmark value for three consecutive periods. This optimization includes:

[0066] Set a feedback cycle, collect actual electricity and heat consumption data and strategy execution records of the main user group in each feedback cycle, and calculate the average efficiency improvement rate of users;

[0067] Specifically, the average performance improvement rate = (user performance after strategy execution - user performance before strategy execution) / user performance before strategy execution;

[0068] The proportion of users who implemented the strategy to the total number of users in the main group is used as the percentage of users covered by the strategy.

[0069] The product of the average performance improvement rate and the proportion of users covered by the strategy is the comprehensive performance index. When the comprehensive performance index is lower than the preset benchmark value for three consecutive periods, the dynamic time warping algorithm is called again to match a better reference group path. For example, the preset benchmark value is set according to 80% of the historical best comprehensive performance index.

[0070] For example, the method for rematching a better reference group path is as follows: select the top three main group users with the highest similarity as reference groups, use the electricity consumption behavior of these three users in the most recent feedback cycle as the baseline vector, calculate the similarity of their feature sequences with each path in the three strategy path library through the dynamic time warping algorithm, select the top three paths with the highest similarity as candidate paths, and select the path with the highest comprehensive efficiency index as the better reference group path.

[0071] See Figure 4 As shown, this scheme proposes a bidirectional collaborative cogeneration user group energy efficiency optimization system to implement the aforementioned bidirectional collaborative cogeneration user group energy efficiency optimization method, including:

[0072] The strategy feature vector construction module is used to set a historical period, collect historical electricity and heat consumption data of users within that period, and construct a strategy feature vector. The strategy feature vector includes electricity and heat consumption behavior characteristics, demand response participation characteristics, and economic benefit characteristics.

[0073] The user group classification module is used to classify user strategy feature vectors using clustering algorithms to identify users in information silos and main user groups.

[0074] The strategy path matching module is used to calculate the monthly load variance and the frequency of abnormal electricity price response based on the historical electricity and heat consumption data of isolated users, and to match the historical strategy paths of similar main groups of users.

[0075] The bidirectional collaborative optimization module is used to guide isolated users to converge to the historical strategy paths of the main group of users, while using strategies of isolated users with higher efficiency than the preset potential value to reverse correct the historical strategy path library of the main group.

[0076] The dynamic feedback and iteration module is used to establish a dynamic feedback mechanism to monitor the implementation effect of the historical strategy path correction for the main user group in real time, calculate the comprehensive efficiency index of the corrected path, and optimize the reference group path when the index is lower than the preset benchmark value for three consecutive periods.

[0077] The strategy feature vector construction module specifically includes:

[0078] The multi-dimensional feature extraction unit is used to formulate power load curves and heat demand data curves based on historical electricity and heat consumption data, and extract corresponding statistical indicators and time periodic features.

[0079] The demand response characteristic statistics unit is used to calculate the participation rate, average response time, and average response amount of users participating in electricity price response projects within a historical period. The participation rate is the ratio of the number of responses to the number of electricity price adjustments.

[0080] The economic benefit calculation unit is used to obtain the actual electricity price for different time periods based on real-time electricity prices and subsidies for response projects, and combine it with historical electricity and heat consumption data to obtain the user's time-series electricity cost within the historical period.

[0081] The user group classification module specifically includes:

[0082] The clustering parameter adaptive unit is used to set the range of values ​​for the number of clusters according to the characteristics of the user group of the cogeneration system, and to select the initial cluster center for each cluster number value using the K-means++ algorithm;

[0083] The silhouette coefficient analysis unit is used to calculate the Euclidean distance from each user's policy feature vector to each cluster center, and assign it to the nearest cluster, updating the cluster center until convergence.

[0084] The dynamic feedback and iteration module specifically includes:

[0085] The efficiency index calculation subunit is used to set the feedback cycle. In each feedback cycle, the actual electricity and heat consumption data and strategy execution records of the main group of users are collected, and the average efficiency improvement rate of users is calculated. The proportion of users who execute the strategy to the total number of users in the main group is also counted as the proportion of users covered by the strategy.

[0086] The optimal path rematching unit is used to calculate the comprehensive performance index by multiplying the average performance improvement rate by the proportion of users covered by the strategy. When the comprehensive performance index is lower than the preset benchmark value for three consecutive periods, the dynamic time warping algorithm is called again to match a better reference group path.

[0087] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for optimizing the energy efficiency of cogeneration user groups based on bidirectional synergy, characterized in that, The method includes: Set a historical period, collect historical electricity and heat consumption data of users within the period, and construct a strategy feature vector. The strategy feature vector includes electricity and heat consumption behavior characteristics, demand response participation characteristics, and economic benefit characteristics. Clustering algorithms are used to classify user strategy feature vectors to identify users in information silos and main user groups; Based on the historical electricity and heat consumption data of isolated users, the monthly load variance and the frequency of abnormal electricity price response are calculated, and the historical strategy paths of similar main user groups are matched. Guide isolated users to converge to the historical strategy paths of the main group of users, while using strategies of isolated users that are more effective than the preset potential value to reverse correct the historical strategy path library of the main group. Establish a dynamic feedback mechanism to monitor the implementation effect of the historical strategy path correction for the main user group in real time, calculate the comprehensive efficiency index of the corrected path, and optimize the reference group path when the index is lower than the preset benchmark value for three consecutive periods. The calculation of monthly load variance and abnormal electricity price response frequency based on historical electricity and heat consumption data of isolated users, and matching historical strategy paths of similar main user groups, specifically includes: Extract historical electricity and heat consumption data of isolated users, calculate their monthly load variance, set a delay threshold, and calculate the frequency of abnormal electricity price response by combining demand response participation characteristics. The monthly load variance and abnormal frequency of electricity price response of the main user group are vectorized to form a historical strategy path library. Each path contains seasonal fluctuation patterns and electricity price response behavior. The dynamic time warping algorithm is used to calculate the similarity of feature sequences between isolated users and the main group of users, and the main group of users with the highest similarity are selected as the reference group. Based on the historical policy paths of the reference group, the load adjustment magnitude and electricity price adjustment trigger threshold are extracted to generate the policy adjustment baseline for isolated users. The process of guiding isolated users to converge towards the historical strategy paths of the main group of users, while simultaneously using strategies among isolated users that have higher efficiency than a preset potential value to reverse-correct the historical strategy path library of the main group, specifically includes: Real-time collection of current electricity and heat consumption data and electricity price response project data from isolated users, and extraction of real-time electricity price from the electricity price response project data; If the real-time electricity price exceeds the electricity price adjustment trigger threshold in the strategy adjustment baseline, the load adjustment magnitude will be pushed to the isolated user. After users adjust their electricity consumption behavior according to the load adjustment range, the ratio of their monthly load data to their monthly electricity cost is used as the user's efficiency. The strategy adjustment baseline for users whose performance exceeds the preset potential value is weighted and merged with the corresponding strategy path of the main group to obtain the corrected strategy path, and then incorporated into the historical strategy path library of the main group users. The establishment of a dynamic feedback mechanism monitors the implementation effect of historical strategy path corrections for the main user group in real time, calculates the comprehensive efficiency index of the corrected path, and optimizes the reference group path when the index is lower than the preset benchmark value for three consecutive periods. Specifically, this includes: Set a feedback cycle, and collect the actual electricity and heat consumption data and strategy execution records of the main user group in each feedback cycle, and calculate the average efficiency improvement rate of users; The proportion of users who implemented the strategy to the total number of users in the main group is used as the proportion of users covered by the strategy. The product of the average performance improvement rate and the proportion of users covered by the strategy is the comprehensive performance index. When the comprehensive performance index is lower than the preset benchmark value for three consecutive periods, the dynamic time warping algorithm is called again to match a better reference group path.

2. The method according to claim 1, characterized in that, The constructed strategy feature vector specifically includes: Based on historical electricity and heat consumption data, electricity load curves and heat demand data curves are proposed, and corresponding statistical indicators and time periodic characteristics are extracted. The statistical indicators include mean, variance, peak load and valley ratio; the time periodicity features include the frequency components and phase of the main period amplitude; and the electricity and heat consumption behavior features are jointly characterized by statistical indicators and time series periodicity features. The participation rate, average response time, and average response volume of users in the electricity price response project were statistically analyzed within the historical period. The participation rate is the ratio of the number of responses to the number of electricity price adjustments. The demand response participation characteristic is jointly characterized by participation rate, average response time, and average response volume. Based on real-time electricity prices and subsidies for response projects, the actual electricity prices for different time periods are obtained. Combined with historical electricity and heat consumption data, the time-series electricity cost of users within the historical period is obtained and used as an economic benefit characteristic.

3. The method according to claim 2, characterized in that, The method of using clustering algorithms to classify user strategy feature vectors and identify users in information silos and main user groups specifically includes: Based on the characteristics of the user group of the cogeneration system, the range of values ​​for the number of clusters is set, and the K-means++ algorithm is used to select the initial cluster center for each cluster number value. Calculate the Euclidean distance from each user's policy feature vector to each cluster center, and assign it to the nearest cluster. Update the cluster centers until convergence. Calculate the profile coefficient of each user in the clustering results, plot the average profile coefficient curve corresponding to different cluster number values, and select the cluster number value that maximizes the profile coefficient as the final cluster number. If the average profile coefficient of a user group is lower than a set threshold, or the number of users in the group is less than a set proportion, then the group is determined to be an information island user group; if neither is true, then the group is determined to be a main user group. The formula for calculating the contour coefficient is as follows: In the formula, s(i) is the silhouette coefficient, b(i) is the average Euclidean distance between user i and other users in the same cluster, and a(i) is the minimum average Euclidean distance between user i and other clusters.

4. An energy efficiency optimization system for a cogeneration user group based on bidirectional collaboration, characterized in that, The method for optimizing the energy efficiency of a cogeneration user group based on bidirectional collaboration as described in any one of claims 1-3 includes: The strategy feature vector construction module is used to set a historical period, collect historical electricity and heat consumption data of users within that period, and construct a strategy feature vector. The strategy feature vector includes electricity and heat consumption behavior characteristics, demand response participation characteristics, and economic benefit characteristics. The user group classification module is used to classify user strategy feature vectors using clustering algorithms to identify users in information silos and main user groups. The strategy path matching module is used to calculate the monthly load variance and the frequency of abnormal electricity price response based on the historical electricity and heat consumption data of isolated users, and to match the historical strategy paths of similar main groups of users. The bidirectional collaborative optimization module is used to guide isolated users to converge to the historical strategy paths of the main group of users, while using strategies of isolated users with higher efficiency than the preset potential value to reverse correct the historical strategy path library of the main group. The dynamic feedback and iteration module is used to establish a dynamic feedback mechanism to monitor the implementation effect of the historical strategy path correction for the main user group in real time, calculate the comprehensive efficiency index of the corrected path, and optimize the reference group path when the index is lower than the preset benchmark value for three consecutive periods.

5. The energy efficiency optimization system for cogeneration user groups based on bidirectional collaboration according to claim 4, characterized in that, The strategy feature vector construction module specifically includes: The multi-dimensional feature extraction unit is used to formulate power load curves and heat demand data curves based on historical electricity and heat consumption data, and extract corresponding statistical indicators and time periodic features. The demand response characteristic statistics unit is used to calculate the participation rate, average response time, and average response amount of users participating in electricity price response projects within a historical period. The participation rate is the ratio of the number of responses to the number of electricity price adjustments. The economic benefit calculation unit is used to obtain the actual electricity price for different time periods based on real-time electricity prices and subsidies for response projects, and combine it with historical electricity and heat consumption data to obtain the user's time-series electricity cost within the historical period.

6. The energy efficiency optimization system for cogeneration user groups based on bidirectional collaboration as described in claim 5, characterized in that, The user group classification module specifically includes: The clustering parameter adaptive unit is used to set the range of values ​​for the number of clusters according to the characteristics of the user group of the cogeneration system, and to select the initial cluster center for each cluster number value using the K-means++ algorithm; The silhouette coefficient analysis unit is used to calculate the Euclidean distance from each user's policy feature vector to each cluster center, and assign it to the nearest cluster, updating the cluster center until convergence.

7. The energy efficiency optimization system for cogeneration user groups based on bidirectional collaboration according to claim 6, characterized in that, The dynamic feedback and iteration module specifically includes: The efficiency index calculation subunit is used to set the feedback cycle. In each feedback cycle, the actual electricity and heat consumption data and strategy execution records of the main group of users are collected, and the average efficiency improvement rate of users is calculated. The proportion of users who execute the strategy to the total number of users in the main group is also counted as the proportion of users covered by the strategy. The optimal path rematching unit is used to calculate the comprehensive performance index by multiplying the average performance improvement rate by the proportion of users covered by the strategy. When the comprehensive performance index is lower than the preset benchmark value for three consecutive periods, the dynamic time warping algorithm is called again to match a better reference group path.

Images