User behavior accurate portraying method based on typical adjustable equipment on user side

Abstract

The invention provides a user behavior accurate portraying method based on user-side typical adjustable equipment, and belongs to the technical field of power grid regulation and control measurement and calculation assistance. The method comprises the following steps: acquiring and preprocessing multi-source data of a user side, wherein the multi-source data at least comprises load data of typical adjustable equipment, time sequence data of a user power utilization gateway and user basic information; based on the preprocessed multi-source data, constructing a multi-dimensional attribute user behavior portrait of typical adjustable equipment on the user side; in combination with the multi-dimensional attribute user behavior portrait, evaluating the real-time regulation potential of typical adjustable equipment on the user side; and based on the real-time regulation potential, constructing a user side load regulation cost calculation model, and outputting a regulation excitation scheme. According to the method, a closed-loop technical link can be formed through full-process design of data preprocessing, multi-dimensional portraits, real-time potential and cost excitation, so that the precision of load-side resource coordinated regulation and control is effectively improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power grid regulation and control calculation assistance, and particularly relates to a user behavior accurate portrait method based on user side typical adjustable equipment. BACKGROUND

[0002] With the promotion of new energy vehicles, the emerging load of regional power grids grows rapidly, the source-load time sequence matching difficulty increases, the time period supply-demand contradiction intensifies, and the power grid peak shaving pressure significantly increases.

[0003] The current power supply and grid regulation capacity is limited, and the traditional source-load dynamic scheduling mode is difficult to cope with complex supply and demand changes, so it is urgent to tap the load side resources to participate in collaborative regulation and control to fill the regulation capacity gap and improve the flexibility of power grid operation.

[0004] However, in the prior art, it is difficult to effectively integrate user side multi-source data, and it is difficult to realize accurate portrayal of typical adjustable equipment behavior, quantification of regulation potential and scientific calculation of regulation cost, resulting in low precision of load side resource collaborative regulation and control, and difficulty in meeting the demand of safe and efficient operation of power grid. SUMMARY

[0005] The technical problem to be solved by the application is how to improve the precision of load side resource collaborative regulation and control.

[0006] To solve the above technical problems, the technical scheme adopted by the application is: The application provides a user behavior accurate portrait method based on user side typical adjustable equipment, comprising: S1, obtaining and preprocessing multi-source data of the user side, wherein the multi-source data at least includes load data of typical adjustable equipment, user power consumption port time sequence data and user basic information; S2, based on the preprocessed multi-source data, constructing a multi-dimensional attribute user behavior portrait of the user side typical adjustable equipment; S3, combining the multi-dimensional attribute user behavior portrait, evaluating the real-time regulation potential of the user side typical adjustable equipment; S4, based on the real-time regulation potential, constructing a user side load regulation cost calculation model, and outputting a regulation incentive scheme.

[0007] Compared with the prior art, the beneficial effects of the present application include: first, obtaining multi-source data on the user side and preprocessing. Among them, the multi-source data is a core data source set to comprehensively cover the user-side power consumption characteristics and equipment attributes, specifically including three types of key data: first, the load data of typical adjustable equipment, which refers to the power consumption characteristic data of equipment that can participate in grid regulation, such as electric vehicle charging piles / stations, light storage charging microgrids, etc., in the operation process, which is collected in real time through power sensors, current sensors, etc., and is used to reflect the basic adjustment capacity of the equipment itself; second, the time sequence data of the user power consumption gateway, which refers to the time sequence power consumption data at the user total power consumption gateway such as the household electric meter, which is recorded by the smart meter at 15 minutes / time or higher frequency, and contains information such as daily / weekly / monthly power consumption peak period, average power, etc., which is used to reflect the overall power consumption law of the user and the influence of equipment operation on the total load; third, "user basic information", which covers user nature, consumption level, adjustable equipment type and quantity, and equipment intelligence level, etc., which is obtained through user profile retrieval, field research or power consumption business data integration, and is used to establish the association between the user and the equipment; preprocessing is a key link set for the problems of abnormal values such as power mutation to 0 caused by sensor failure, missing values such as time sequence data blank caused by short-term disconnection of the electric meter, and magnitude differences such as power unit kW and current unit A, etc. Through data cleaning, completion, standardization, the data quality is ensured to meet the subsequent analysis requirements, and a reliable data foundation is laid for accurate portrait construction. On this basis, a multi-dimensional attribute user behavior portrait is constructed based on the preprocessed multi-source data. Among them, the multi-dimensional attribute system is the core design that breaks through the traditional single power consumption portrait, including basic attributes, behavior attributes and equipment attributes. In this way, the association is established from three dimensions of user, behavior and equipment, avoiding the defects of traditional portraits that only focus on power consumption and ignore the device adjustment potential and user response willingness. In the construction process, the preprocessed data is associated with the multi-dimensional attribute system one by one, such as associating the power consumption peak period in the gateway time sequence data with the load characteristics in the behavior attribute, then extracting the user behavior law through association rule mining technology, such as the charging pile of high response initiative user is mostly operated in the valley section of the power grid, and finally using the improved K-Means++ algorithm to group users, and combining the labeling technology to generate the exclusive portrait of each user, such as residential user, high response, 2 charging piles and high intelligence, realizing the accurate binding of user behavior and adjustable equipment. Then, the real-time regulation potential is evaluated combined with the multi-dimensional attribute user behavior portrait. The cooperation here is reflected in: the response initiative attribute in the portrait can correct the basic regulation potential of the equipment, for example, the regulation potential of high response initiative user is calculated as 100%, and the regulation potential of low response initiative user is reduced by 60%, avoiding the error caused by the traditional potential evaluation which only based on equipment parameters and ignored user willingness; real-time is realized through real-time collection of equipment operation state such as charging pile remaining power and energy storage capacity, ensuring that the potential evaluation is synchronized with the actual working condition of the equipment.Thus, the setting can solve the problem of static and non-combined user behavior in the prior art to make the potential result more in line with the actual regulation scene. Finally, a cost calculation model is constructed based on the real-time regulation potential and an incentive scheme is output. The user-side load regulation cost calculation model is designed for the economic loss of the user and the incentive investment of the power grid in the regulation process, is disassembled into power loss cost, regulation incentive cost and equipment investment conversion cost, and ensures that the cost calculation covers all links. The regulation incentive scheme is generated based on the balance of potential and cost, for example, the user with high potential and low cost is given priority to develop an incentive strategy. The above-mentioned step solves the problem that the existing regulation scheme lacks economic consideration and the incentive and potential are not matched, and realizes effective and cost-controllable regulation. In summary, through the whole process design of data preprocessing, multi-dimensional portrait, real-time potential and cost incentive, the problems of the existing technology, such as the inability to integrate multi-source data, inaccurate portrait, poor potential quantification and unscientific cost, are effectively solved, a closed-loop technical link is formed, the precision of the load-side resource collaborative regulation is effectively improved, and the regulation cost is reduced.

[0008] Optionally, in the S1, the load data includes steady-state characteristic data of an electric vehicle charging pile / station and a light storage and charging micro-grid and transient-state characteristic data of the light storage and charging micro-grid; the steady-state characteristic data at least includes device rated power, power factor, voltage stability value and current stability value; the transient-state characteristic data at least includes power fluctuation curve when the device starts and stops, current mutation amplitude and voltage sag duration; the user power consumption gate timing data at least includes daily / weekly / monthly power consumption peak period, average power, device start and stop time and power mutation point; and the user basic information at least includes user nature, consumption level, adjustable device type and quantity and device intelligence level.

[0009] Optionally, the preprocessing in the S1 includes the following steps: S11, adopting a 3σ criterion to eliminate abnormal values in the multi-source data, the abnormal values including data with power mutation of 0 and voltage exceeding the rated range by ±10%; S12, adopting a linear interpolation method to complete missing timing data in the multi-source data, and the time resolution of the completed data is not less than 15 minutes / time; S13, performing standardization processing on the preprocessed multi-source data by the following formula: , wherein, the is the multi-source data after standardization, the is original data of the preprocessed multi-source data, the is the minimum value in the original data, and the is the maximum value in the original data.

[0010] Optionally, the step of constructing the multi-dimensional attribute user behavior portrait of the typical adjustable device in S2 specifically comprises the following steps: S21, associating the pre-processed multi-source data with a preset multi-dimensional attribute system, the multi-dimensional attribute system comprising basic attributes, behavior attributes and device attributes; S22, based on the multi-source data and the preset multi-dimensional attribute system, extracting user behavior rules using association rules, the behavior rules at least comprising coincidence degree of user peak electricity consumption period and adjustable device operation period and device start-stop frequency in different electricity price intervals; S23, grouping users, and generating an exclusive behavior portrait for each user based on the grouping result combined with a labeling technology.

[0011] Optionally, in S21, the basic attributes comprise user properties, consumption levels and energy composition; the behavior attributes comprise response initiative and load characteristics; and the device attributes comprise adjustable device types and device intelligence levels.

[0012] Optionally, S3 specifically comprises: S31, constructing exclusive control models for multiple types of typical adjustable devices, the typical adjustable devices comprising electric vehicle charging piles / stations and light storage charging microgrids; S32, based on the multi-dimensional attribute user behavior portrait, extracting attributes of the response initiative, and reducing the basic control potential output by the exclusive control model by the following formula: , wherein, the is the reduced actual control potential, the is the basic control potential, and the is a response initiative reduction coefficient; S33, collecting the operation state of the typical adjustable device, updating the actual control potential, forming a real-time control potential, and outputting quantitative results of adjustable capacity and response duration.

[0013] Optionally, in S31, the exclusive control model of the electric vehicle charging pile / station is an association model of charging power, charging duration and residual power, in the process of outputting the basic control potential by the association model, when the residual power of the charging pile is > 80%, the charging is suspended for 2-4 hours, the basic control potential is equal to the rated power of the charging pile, when the residual power of the charging pile is 50%-80%, the charging power is reduced by 30%-50%, the basic control potential is equal to the product of the rated power and the proportion of the reduced charging power, and when the residual power of the charging pile is < 50%, the basic control potential is 0.

[0014] Optionally, in the S31, the dedicated regulation model of the light storage and charging micro-grid is a correlation model of photovoltaic output, energy storage capacity and composite demand, in the process of outputting the basic regulation potential in the correlation model, when the photovoltaic output > the load demand and the energy storage capacity is not full, the basic regulation potential is equal to the difference between the photovoltaic output and the load demand, when the photovoltaic output < the load demand and the energy storage capacity has a surplus, the basic regulation potential is equal to the energy storage releasable power, when the energy storage capacity is full and the photovoltaic output ≤ the load demand, the basic regulation potential is 0.

[0015] Optionally, the construction of the user-side load adjustment cost measurement model in the S4 specifically includes: S41, disassembling the composition of the user-side load adjustment cost, the composition including power loss cost, regulation incentive cost and equipment investment conversion cost; S42, respectively measuring each cost item of the composition, and obtaining the total adjustment cost based on the measurement results, and constructing the user-side load adjustment cost measurement model.

[0016] Optionally, the output of the regulation incentive scheme in the S4 specifically includes: S43, determining the regulation scenarios of the regional power grid, the regulation scenarios including at least peak regulation scenarios, transformer power reverse sending / overload scenarios, photovoltaic consumption scenarios and user energy consumption cost saving scenarios; S44, based on the quantification results of the adjustable capacity and the response duration, the total adjustment cost is calculated, and a multi-objective optimization model is constructed; S45, solving the multi-objective optimization model, outputting the load-side resource collaborative regulation incentive scheme for multiple regulation scenarios, the regulation incentive scheme including at least preferentially calling the adjustable equipment of high response initiative users and preferentially calling the light storage and charging micro-grid reverse power in the photovoltaic consumption scenario. BRIEF DESCRIPTION OF DRAWINGS

[0017] The application will be further described in detail below with reference to the accompanying drawings.

[0018] Figure 1 The flowchart of the user behavior accurate portrait method based on typical adjustable equipment on the user side in the embodiment of the application. DETAILED DESCRIPTION

[0019] For better understanding of the present application, the present application is further clarified below in conjunction with examples, but the protection scope of the present application is not limited to the following examples. In the following description, a large number of specific details are given in order to provide a more thorough understanding of the present application. However, it is obvious to those skilled in the art that the present application can be implemented without one or more of these details.

[0020] The term "comprising" and variations thereof as used herein are open-ended, that is "including, but not limited to"; the term "based on" is, at least in part based on; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optional" means "optional in at least some embodiments". Related definitions are given throughout the description. It is noted that the concepts of "first", "second", etc. mentioned in the present application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0021] It should be noted that the modification of "one" or "multiple" mentioned in the present application is illustrative rather than limiting, and those skilled in the art should understand that unless otherwise explicitly indicated in the context, it should be understood as "one or more".

[0022] An embodiment of the present application provides a user behavior accurate portrait method based on user side typical adjustable equipment, comprising: S1, acquiring multi-source data of the user side and pre-processing, the multi-source data at least including load data of the typical adjustable equipment, user power consumption port timing data and user basic information; S2, based on the pre-processed multi-source data, constructing a multi-dimensional attribute user behavior portrait of the user side typical adjustable equipment; S3, combining the multi-dimensional attribute user behavior portrait, evaluating the real-time regulation potential of the user side typical adjustable equipment; S4, based on the real-time regulation potential, constructing a user side load adjustment cost calculation model, and outputting a regulation incentive scheme.

[0023] In the present embodiment, first, as shown in FIG. 1, the user side typical adjustable equipment is divided into three categories: 1) the user side typical adjustable equipment, such as air conditioners, water heaters, washing machines, etc., which can be adjusted by the user; 2) the user side typical adjustable equipment, such as refrigerators, televisions, etc., which cannot be adjusted by the user; 3) the user side typical adjustable equipment, such as air conditioners, water heaters, washing machines, etc., which can be adjusted by the user. Figure 1S1, acquires multi-source data of the user side and performs preprocessing. The multi-source data is a core data source set to comprehensively cover the user-side power consumption characteristics and device attributes, and specifically includes three types of key data: 1) load data of typical adjustable devices, which refers to the power consumption characteristic data of devices such as electric vehicle charging piles / stations and light storage charging microgrids that can participate in grid regulation during operation, and is acquired in real time through power sensors, current sensors and the like of the devices, and is used to reflect the adjustment capability basis of the devices themselves; 2) user power consumption gateway time series data, which refers to the time series power consumption data at the user total power consumption gateway such as the household electric meter, and is recorded by the smart meter at a frequency of 15 minutes / time or higher, and contains information such as daily / weekly / monthly power consumption peak period and average power, and is used to reflect the overall power consumption law of the user and the influence of device operation on the total load; 3) user basic information, which covers user nature, consumption level, adjustable device type and quantity, and device intelligence level, and is acquired through user profile retrieval, field investigation or power consumption service data integration, and is used to establish the association between the user and the device; the preprocessing is a key link set for problems such as abnormal values such as power mutation to 0 caused by sensor failure, missing values such as time series data blank caused by short-term disconnection of the electric meter, and magnitude differences such as power unit kW and current unit A, and through data cleaning, completion and standardization, the data quality is ensured to meet the subsequent analysis requirements, and a reliable data foundation is laid for accurate portrait construction. On this basis, as shown in S2, Figure 1 a multi-dimensional attribute user behavior portrait is constructed based on the preprocessed multi-source data. The multi-dimensional attribute system is the core design that breaks through the traditional single power consumption portrait, and includes basic attributes, behavior attributes and device attributes, so that the association is established from the three dimensions of user, behavior and device, and the defects of the traditional portrait that only focuses on power consumption and ignores the device adjustment potential and user response willingness are avoided, and in the construction process, the preprocessed data is associated with the multi-dimensional attribute system one by one, such as associating the power consumption peak period in the gateway time series data with the load characteristics in the behavior attribute, then extracting the user behavior law through association rule mining technology, such as the charging pile of the high response active user being operated in the valley period of the grid, and finally using the improved K-Means++ algorithm to group the users, and generating the exclusive portrait of each user such as resident user, high response, 2 charging piles and high intelligence through the labeling technology, to realize the accurate binding of user behavior and adjustable devices. Then, as shown in Figure 1The real-time regulation potential is evaluated in combination with the multi-dimensional attribute user behavior portrait, as shown in S3. The coordination here is reflected in that the response initiative attribute in the portrait can correct the basic regulation potential of the device, for example, the regulation potential of a high response initiative user is counted as 100%, and that of a low response initiative user is reduced by 60%, to avoid the error caused by ignoring the user's willingness in the traditional potential evaluation based on device parameters; the real-time is realized by real-time collection of device operating states such as the remaining power of the charging pile and the energy storage capacity, to ensure that the potential evaluation is synchronized with the actual working condition of the device. This setting can solve the problem of static and non-combination with user behavior in the prior art, and make the potential result more in line with the actual regulation scene. Finally, as shown in S4 in the Figure 1 Based on the real-time regulation potential, a cost calculation model is constructed and an incentive scheme is output, as shown in S4. The user-side load regulation cost calculation model is designed for the economic loss of the user and the incentive investment of the power grid in the regulation process, and is decomposed into power loss cost, regulation incentive cost and device investment conversion cost, to ensure that the cost calculation covers all links; the regulation incentive scheme is generated based on the balance between potential and cost, for example, high-potential and low-cost users are given priority in formulating incentive strategies. This step solves the problem of lack of economic consideration in the existing regulation scheme and the mismatch between incentives and potential, and realizes effective regulation and controllable cost. In summary, through the whole process design of data preprocessing, multi-dimensional portrait, real-time potential and cost incentive, the problems of inability to integrate multi-source data, inaccurate portrait, poor potential quantification and unscientific cost in the prior art are effectively solved, forming a closed-loop technical link, which effectively improves the precision of load-side resource collaborative regulation and reduces the regulation cost.

[0024] Optionally, in S1, the load data includes steady-state characteristic data of the electric vehicle charging pile / station and the light storage and charging micro-grid, and transient characteristic data of the light storage and charging micro-grid; the steady-state characteristic data at least includes device rated power, power factor, voltage stability value and current stability value; the transient characteristic data at least includes power fluctuation curve, current mutation amplitude and voltage sag duration when the device starts and stops; the user electricity consumption gateway timing data at least includes daily / weekly / monthly electricity consumption peak period, average power, device start and stop time and power mutation point; the user basic information at least includes user nature, consumption level, adjustable device type and quantity, and device intelligence level.

[0025] In this optional embodiment, first, the load data of a typical adjustable device is divided into steady-state characteristic data and transient-state characteristic data: the steady-state characteristic data refers to the core electrical parameters when the device is stably running, including the rated power, power factor, voltage stable value and current stable value. In this way, the basic adjustment capacity boundary of the device in the normal running state can be reflected, for example, the rated power directly determines the maximum adjustable capacity of the device; the transient-state characteristic data refers to the dynamic electrical parameters when the device starts or stops or the working condition is switched, including the power fluctuation curve when the device starts or stops, the current mutation amplitude and the voltage sag duration. In this way, the dynamic influence in the adjustment process of the device can be captured, and the defect of ignoring transient impact in the traditional steady-state can be avoided, for example, the current mutation amplitude can be used to judge the impact risk of the device on the power grid when it is regulated. Secondly, the specific indicators of the user power outlet timing data are: including daily / weekly / monthly power consumption peak period, average power, device start and stop time and power mutation point. In this way, the relationship between device operation and total load can be associated from the overall perspective of the user, for example, whether the start and stop of the device cause the total load of the user to enter the peak segment of the power grid. Finally, the specific items of the user basic information are: the user nature is determined by the power consumption business classification, the consumption level is graded according to the annual power consumption, the type and number of adjustable devices are confirmed through on-site verification or user declaration data, and the intelligent level of the device is judged according to whether the device supports remote control and the data upload frequency. In this way, the association label of the user and the device can be established, and direct basis can be provided for the construction of the basic attributes and device attributes in the subsequent multi-dimensional portrait. In summary, by refining the specific content and acquisition method of multi-source data, the problem of unclear data source and unclear collection direction in the prior art can be effectively solved, the coverage of multi-source data can be effectively improved, the targeted data support in the subsequent preprocessing and portrait construction links can be ensured, and the operability and accuracy of the overall technical scheme can be further improved.

[0026] Optionally, the preprocessing in S1 includes the following steps: S11, adopting 3σ criterion to eliminate outliers in the multi-source data, the outliers including data with power mutation of 0 and voltage exceeding the rated range by ±10%; S12, adopting linear interpolation method to complete the missing time series data in the multi-source data, and the time resolution of the completed data is not less than 15 minutes / time; S13, performing standardization processing on the preprocessed multi-source data by the following formula: (1.1), wherein, is the standardized multi-source data, is the original data of the preprocessed multi-source data, is the minimum value in the original data, is the maximum value in the original data.

[0027] Specifically In this optional embodiment, first, the 3σ criterion is used to eliminate outliers. The 3σ criterion is an outlier identification method based on the normal distribution characteristics of data. Its core concept is that if the data follows a normal distribution, about 99.7% of the data falls within ±3σ of the mean, and data outside this range is considered an outlier. The outliers to be eliminated are clearly two types: one is that the power suddenly drops to 0 and the duration is less than 1 minute, which is caused by sensor failure; the other is that the voltage exceeds the rated range of ±10%, such as 220V domestic voltage exceeding the range of 198V-242V, which is caused by power grid fluctuation or meter failure. The purpose of outlier elimination is to avoid false data interference in subsequent profiling and potential assessment. By using the 3σ criterion instead of a simple threshold method, outliers and normal fluctuation data can be more accurately distinguished to ensure data authenticity. Second, linear interpolation is used to complete the missing time series data. Linear interpolation is a method based on the linear relationship between the two adjacent valid data points to estimate the missing point data. The formula is where x is the missing point time, x1 and x2 are the adjacent valid data times, and y1 and y2 are the power / current values corresponding to the times. The time resolution of the completed data is set to no less than 15 minutes / time, because the conventional time granularity of power grid dispatching is 15 minutes, and a resolution lower than this will result in inaccurate user electricity consumption patterns. The purpose of completing missing values is to solve the problem of discontinuous time series data and ensure the integrity of the gateway time series data, providing continuous time dimension support for subsequent behavior pattern extraction. Finally, the preprocessed data is standardized by formula (1.1). The purpose of standardization is to solve the analysis deviation caused by the order of magnitude difference of multi-source data, so that different types of data can be included in the same model analysis, and the rationality of data fusion is ensured. In this way, through the three-step preprocessing process of outlier elimination, missing value completion and standardization, the problem of poor quality of multi-source data is effectively solved, the data integrity is improved, the data consistency is improved, and a high-quality data foundation is provided for subsequent multi-dimensional portrait construction and control potential assessment, avoiding analysis errors caused by data problems.

[0028] Optionally, the step S2 of constructing a multi-dimensional attribute user behavior portrait of a typical adjustable device on the user side specifically includes the following steps: S21, associating the preprocessed multi-source data with a preset multi-dimensional attribute system, the multi-dimensional attribute system including basic attributes, behavior attributes and device attributes; S22, based on the multi-source data and the preset multi-dimensional attribute system, extracting user behavior patterns using association rules, the behavior patterns including at least the coincidence degree of user electricity consumption peak period and adjustable device operation period, and the device start-stop frequency in different price intervals; S23, grouping users, and generating an exclusive behavior portrait for each user based on the grouping results and labelization technology.

[0029] In the optional embodiment, first, the preprocessed multi-source data is associated with the preset multi-dimensional attribute system. The preset multi-dimensional attribute system is an image framework designed in advance based on the demand of power grid regulation and control, which includes three categories of basic attributes, behavior attributes and device attributes, and the purpose of setting is to cover the core factors affecting the regulation and control of the user side. In the association process, the preprocessed data and the attributes need to be matched one by one: for example, the user property in the user basic information is matched to the basic attribute, the coincidence degree of the power consumption peak and the device operation period in the gateway time series data is matched to the behavior attribute, and the device type in the load data is matched to the device attribute. In this way, the correspondence between data and attributes is established, the data source for subsequent behavior rule extraction is clear, and the deviation of the image caused by the disconnection between data and attributes is avoided. Secondly, based on the multi-source data and the multi-dimensional attribute system, the user behavior rules are extracted by using the association rule. The association rule mining technology is an algorithm for discovering the implicit association between variables from data, which can be used to extract two types of key rules: one is the coincidence degree of the user power consumption peak period and the adjustable device operation period, which is calculated by counting the overlap length ratio of the user total load peak period and the device operation period in the gateway time series data, reflecting the influence of device operation on power peak and valley; the second is the device start-stop frequency in different price intervals, which is divided into three grades of peak / flat / valley according to the price, and the start-stop times of the device in each grade are counted, reflecting the sensitivity of the user to the price. The purpose of behavior rule extraction is to convert scattered data into information with regulation value, to provide basis for user clustering, and to avoid the image staying in static attributes and lacking dynamic behavior description. Finally, the users are clustered, and the exclusive behavior image is generated based on the clustering results combined with the labeling technology. The user clustering adopts the improved K-Means++ algorithm, which solves the problem of easy falling into local optimum of traditional K-Means by optimizing the selection of initial clustering center, and the clustering basis is the extracted behavior rules and multi-dimensional attributes; the labeling technology adds feature labels to each cluster and individual user, making the image more intuitive and easy to understand. The purpose of clustering and labeling is to realize the fine classification of users, for example, high response and low price sensitivity users can be formulated with low incentive regulation scheme, and low response and high price sensitivity users can be promoted to participate through price incentive. In this way, through the image construction process of data, attribute association, rule extraction and clustering labeling, the problem of single dimension of traditional image and disconnection with device adjustment demand is effectively solved, the accuracy of user clustering is improved, the matching degree of image and regulation demand is improved, and accurate user behavior basis is provided for subsequent real-time regulation potential evaluation.

[0030] Optionally, in S21, the basic attributes include user property, consumption level and energy composition; the behavior attributes include response initiative and load characteristics; and the device attributes include adjustable device type and device intelligence level.

[0031] Specifically, the basic attributes include three types of core information: user nature: divided into residents, industries, and businesses according to the electricity consumption type, obtained through the business classification code corresponding to the user electricity meter number, and the purpose of the setting is to distinguish the electricity consumption characteristics of different users; consumption level: divided into high (residents > 4800 kWh / year), medium (2400-4800 kWh / year), and low (< 2400 kWh / year) three levels according to the annual electricity consumption of users, obtained through the annual electricity consumption statistics of smart meters, and used to reflect the sensitivity of users to electricity costs; energy composition refers to whether the user is connected to distributed photovoltaic, energy storage, etc., obtained through distributed power grid connection records or on-site verification, and used to determine whether the user has the adjustment ability of self-generation and self-use and surplus power on the network. The behavior attribute includes two types of key information: response enthusiasm, calculated through the user's historical participation in demand response records, specifically response rate = actual participation times / invitation times, combined with invitation response speed comprehensive evaluation, divided into high, medium and low three levels, extracted from the historical response records of the power grid dispatching system, and the core role is to correct the control potential; load characteristics include the proportion of adjustable equipment electricity consumption (adjustable equipment electricity consumption / total user electricity consumption) and the coincidence degree of electricity consumption peak and power grid peak valley (the overlap ratio of user electricity consumption peak time and power grid peak time), calculated through the gateway time series data and equipment load data, and used to determine the influence degree of user equipment on the power grid load. The equipment attribute includes two types of core information: the type of adjustable equipment is clear, which is electric vehicle charging piles / stations and photovoltaic-storage-charging microgrids, determined through equipment account or on-site verification, and the control model of different types of equipment is different (such as charging piles adjust according to the remaining power, and microgrids adjust according to photovoltaic-storage-load coordination); the intelligent level of equipment: divided into high (data upload frequency ≤ 5 minutes / time, support remote start-stop), medium (15 minutes / time, support remote power regulation), and low (30 minutes / time, only local control) three levels according to the data upload frequency and remote control ability, obtained through equipment communication protocol detection and remote control test, and used to determine the real-time control feasibility of the equipment (highly intelligent equipment can be directly included in the remote dispatching of the power grid).

[0032] In the optional embodiment, by specifying the specific content, acquisition method and role of multi-dimensional attributes, the problem of existing portrait attribute ambiguity and weak correlation with control demand is solved, so that each attribute can directly serve subsequent potential evaluation (such as response enthusiasm for potential reduction) and incentive design (such as users with low intelligent level need additional incentives for equipment modification), further improving the practicality and control pertinence of the portrait.

[0033] Optionally, S3 specifically includes: S31, constructing a dedicated control model for a plurality of types of typical adjustable equipment, the typical adjustable equipment including electric vehicle charging piles / stations and photovoltaic-storage-charging microgrids; S32, extracting the attribute of response enthusiasm based on the multi-dimensional attribute user behavior portrait, and reducing the basic control potential output by the dedicated control model through the following formula: (2.1), wherein, is the reduced actual regulation potential, is the basic regulation potential, is the response enthusiasm reduction coefficient; S33, the operation state of the typical adjustable device is collected, the actual regulation potential is updated, the real-time regulation potential is formed, and the quantitative results of the adjustable capacity and the response time are output.

[0034] In the optional embodiment, first, a dedicated regulation model is constructed for various types of typical adjustable devices. The dedicated regulation model is a potential calculation model designed based on the operation characteristics of different devices, which avoids the errors caused by ignoring the device differences in the traditional unified model: for electric vehicle charging piles / stations, a charging power, charging time and remaining power correlation model is constructed; for light storage and charging microgrids, a photovoltaic output, energy storage capacity and load demand coordination model is constructed. In the model construction process, the device physical characteristics (such as the charging power curve of the charging pile) and the historical operation data (such as the photovoltaic output history of the microgrid) need to be combined, and the model is trained and generated through regression analysis or machine learning algorithm (such as linear regression, random forest), to ensure that the basic regulation potential output by the model is consistent with the actual regulation ability of the device. Secondly, the response enthusiasm attribute is extracted based on the multi-dimensional attribute portrait, and the basic regulation potential is reduced through formula (2.1). In this way, the defect that the traditional potential evaluation only considers the device parameters and ignores the user response willingness can be solved, and the potential calculation is more consistent with the actual regulation scene. Finally, the operation state of the typical adjustable device is collected in real time, and the actual regulation potential is updated. The operation state includes the remaining power of the charging pile, the energy storage capacity and the photovoltaic output of the light storage and charging microgrid, etc., which is collected through the device Internet of Things platform at a frequency of 5 minutes / time; the actual regulation potential is updated according to the change of the operation state , and the new is obtained in combination with α. Finally, the adjustable capacity and response time quantitative results are output. The purpose of real-time updating is to solve the problem that the traditional static potential evaluation cannot cope with the device working condition changes, and to ensure that the potential data is synchronized with the actual state of the device. In this way, through the potential evaluation process of the dedicated model, user reduction and real-time updating, the problems of large potential calculation error and insufficient real-time of the prior art are solved, the potential calculation error is reduced, the real-time is improved, and accurate dynamic potential data is provided for subsequent cost calculation and incentive scheme design.

[0035] Optionally, in S31, the dedicated regulation model of the electric vehicle charging pile / station is a correlation model of charging power, charging duration and residual power. In the process of outputting the basic regulation potential by the correlation model, when the residual power of the charging pile is > 80%, the charging is suspended for 2-4 hours, the basic regulation potential is equal to the rated power of the charging pile, when the residual power of the charging pile is between 50% and 80%, the charging power is reduced by 30%-50%, the basic regulation potential is equal to the product of the rated power and the proportion of the reduced charging power, and when the residual power of the charging pile is < 50%, the basic regulation potential is 0.

[0036] In the optional embodiment, the dedicated regulation model of the electric vehicle charging pile / station is a correlation model of charging power, charging duration and residual power. The regulation capacity of the charging pile directly depends on the charging progress, and the adjustable power and duration are significantly different according to the residual power. Therefore, the potential output rules need to be formulated according to the residual power to balance the regulation demand of the power grid and the user energy experience, avoid user resistance caused by forced regulation, and improve the executability of the regulation scheme. In this way, by designing the potential output rules of the charging pile according to the residual power, the problem that the existing model does not combine the charging working condition and the potential calculation does not meet the user demand is solved, the accuracy of the basic regulation potential calculation of the charging pile is improved, the effective regulation capacity is ensured for the power grid, and the user energy experience is also considered, which provides the potential data that can be implemented for subsequent actual regulation.

[0037] Optionally, in S31, the dedicated regulation model of the light storage and charging micro-grid is a correlation model of photovoltaic output, energy storage capacity and composite demand. In the process of outputting the basic regulation potential by the correlation model, when the photovoltaic output > the load demand and the energy storage capacity is not full, the basic regulation potential is equal to the difference between the photovoltaic output and the load demand, when the photovoltaic output < the load demand and the energy storage capacity has a surplus, the basic regulation potential is equal to the energy storage releasable power, and when the energy storage capacity is full and the photovoltaic output ≤ the load demand, the basic regulation potential is 0.

[0038] In the optional embodiment, the dedicated regulation model of the light storage and charging micro-grid is a "photovoltaic output, energy storage capacity and load demand coordination model. The light storage and charging micro-grid includes photovoltaic, energy storage and charging pile / load resources, and the regulation capacity depends on the real-time balance relationship of the three resources. The basic regulation potential needs to be determined through coordination analysis. In this way, the setting can avoid forcibly calculating the potential when the resources are insufficient, ensure the authenticity of the potential data, and avoid the misjudgment of the regulation capacity by the power grid dispatching. In this way, by designing the potential output rules through the coordination analysis of the relationship among photovoltaic, energy storage and load, the problem that the existing model considers a single resource in isolation and cannot reflect the coordinated regulation capacity of the micro-grid is solved, the matching degree of the basic regulation potential calculation of the light storage and charging micro-grid to the actual scene is improved, and accurate multi-resource coordination potential data is provided for the power grid dispatching.

[0039] Optionally, the constructing the user-side load adjustment cost measurement model in S4 specifically comprises: S41, disassembling the components of the user-side load adjustment cost, the components comprising power loss cost, regulation incentive cost, and device investment conversion cost; S42, respectively measuring each cost item of the components, and obtaining the total adjustment cost based on the measurement results, and constructing the user-side load adjustment cost measurement model.

[0040] In this optional embodiment, first, the components of the user-side load adjustment cost are disassembled. The adjustment cost is the total cost paid by the power grid to promote users to participate in regulation, which is disassembled into three core costs: one is the power loss cost, which refers to the loss caused by the reduction of user power consumption due to regulation; the second is the regulation incentive cost, which refers to the compensation paid by the power grid to stimulate user participation; the third is the device investment conversion cost, which refers to the cost apportioned by the user for purchasing adjustable devices. In this way, disassembling can cover all links of user loss, grid incentive, and device depreciation, ensuring that there is no omission in cost measurement. Secondly, each cost item is measured and the total adjustment cost is obtained.

[0041] The first is the power loss cost measurement formula: , Wherein, is the power loss cost, i.e. the economic loss of the user due to the reduction of power consumption due to regulation; is the user power consumption before regulation, which is obtained by statistically analyzing the historical power consumption of the same period before regulation through the smart meter; is the user power consumption after regulation, which is obtained by statistically analyzing the actual power consumption of the regulation period in real time through the smart meter; is the user execution price, which is obtained by calling the user price file from the power grid marketing system. In this way, the direct power loss of the user due to regulation can be accurately quantified, providing a basis for the minimum compensation standard of the incentive scheme design.

[0042] The second is the device investment conversion cost measurement formula: , Wherein, is the annual device investment conversion cost, i.e. the device purchase cost that the user needs to apportion annually; is the device purchase cost, which is obtained by providing the device procurement invoice or market research by the user; is the device residual value, which is calculated according to the remaining value after the expiration of the device service life, usually taking 10% of the purchase cost; is the device service life, which is determined according to the device industry standard. In this way, the one-time device investment can be converted into an annual apportionment cost, avoiding the underestimation of the cost due to the neglect of the device depreciation, and ensuring the comprehensiveness of the cost measurement.

[0043] The third is the total adjustment cost measurement formula: , The total adjustment cost is the sum of the three types of costs, wherein, The total cost is used for judging the economy of the regulation scheme, and only when the grid benefit brought by the regulation is greater than the total adjustment cost, the scheme is feasible. In summary, the embodiment can solve the problem of unscientific cost calculation and inability to support economic decision-making in the prior art by explicitly defining the cost composition and quantitative calculation formula, reduce the total adjustment cost calculation error, provide accurate cost boundary for subsequent incentive scheme design, and ensure that the regulation scheme is effective and economical.

[0044] Optionally, the outputting the regulation incentive scheme in S4 specifically includes: S43, determining the regulation scenarios of the regional power grid, the regulation scenarios at least including a peak regulation scenario, a transformer power reverse sending / heavy overload scenario, a photovoltaic (PV) consumption scenario, and a user energy cost saving scenario; S44, calculating the total adjustment cost based on the quantitative results of the adjustable capacity and the response duration, and constructing a multi-objective optimization model; and S45, solving the multi-objective optimization model, and outputting a load side resource coordinated regulation incentive scheme for the multiple regulation scenarios, the regulation incentive scheme at least including preferentially calling the adjustable devices of the high response initiative users and preferentially calling the power reverse sending of the PV consumption scenario.

[0045] In the optional embodiment, first, the regulation scenarios of the regional power grid are determined. The regulation scenarios are target scenarios set based on the actual operation demand of the power grid, including four types of core scenarios: 1) the peak regulation scenario, which is for the power grid load peak, and the load needs to be reduced to avoid power supply gap; 2) the transformer power reverse sending / heavy overload scenario, which is for the case that the power of the distribution network transformer is reverse sent or the load rate is greater than 80%, and the load needs to be adjusted to protect the equipment; 3) the PV consumption scenario, which is for the case that the distributed PV output is excessive, leading to light abandonment, and the load or reverse sending power needs to be increased; and 4) the user energy cost saving scenario, which is for the case that the user electricity cost is high, and the user needs to be guided to use electricity in the valley section to reduce the electricity cost. The scenario determination is obtained through real-time monitoring data of the power grid dispatching system, to ensure that the scenario is consistent with the actual demand of the power grid. Second, based on the quantitative results of the adjustable capacity and the response duration and the total adjustment cost, a multi-objective optimization model is constructed. The multi-objective is to optimize the regulation effect and minimize the adjustment cost, and the two need to be optimized coordinately. The objective functions and constraint conditions of the model are designed as follows: The regulation effect of the objective function 1 is optimal, and is specifically as follows: , Wherein, is the actual regulation potential of the i-th user, is the response duration, and n is the number of participating users), to ensure that the total adjustment capacity meets the scenario demand; The adjustment cost of the objective function 2 is minimized, and is specifically as follows: , wherein, is the total adjustment cost of the i-th user, ensuring the economic regulation; The constraints include device power constraints (Pmax , is the device rated power) and response duration constraints (Tmax , is the maximum response duration of the device), ensuring that the regulation does not exceed the safe operating range of the device. The model is solved by multi-objective optimization algorithms such as NSGA-II, balancing effect and cost. Finally, the model is solved and the targeted regulation incentive scheme is output. The scheme design combines scenario demand and user portrait features, for example: in the "peak regulation scenario", the adjustable devices of high response initiative users are preferentially called, such users have low incentive cost and can participate without high incentives; in the photovoltaic consumption scenario, the anti-power of the micro-grid of photovoltaic storage and charging is preferentially called, is the difference between photovoltaic output and load demand, maximizing the consumption of excess photovoltaic power. The scheme output form is a user list, adjustment capacity, incentive amount and regulation period, directly supporting grid dispatching execution. In summary, this can effectively solve the problem of existing incentive schemes not adapting to the scene, poor economic efficiency, reduce the peak-valley difference of the grid in the peak regulation scenario, reduce the photovoltaic curtailment rate in the photovoltaic consumption scenario, and reduce the user's energy cost in the user's energy saving scenario.

[0046] The above is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for accurate user behavior profiling based on typical adjustable devices on the user side, characterized in that, include: S1. Acquire multi-source data from the user side and preprocess it. The multi-source data includes at least load data of typical adjustable equipment, time-series data of user power consumption, and basic user information. S2. Based on the preprocessed multi-source data, construct a multi-dimensional attribute user behavior profile of typical adjustable devices on the user side; S3. Based on the multi-dimensional attribute user behavior profile, evaluate the real-time control potential of the typical adjustable device on the user side; S4. Based on the real-time control potential, construct a user-side load adjustment cost calculation model and output a control incentive scheme.

2. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 1, characterized in that, In S1, the load data includes steady-state characteristic data of electric vehicle charging piles / stations, photovoltaic-storage-charging microgrids, and transient characteristic data of the photovoltaic-storage-charging microgrids; the steady-state characteristic data includes at least the rated power of the equipment, power factor, voltage stability value, and current stability value; the transient characteristic data includes at least the power fluctuation curve, current surge amplitude, and voltage sag duration during equipment start-up and shutdown; the user electricity consumption timing data includes at least the daily / weekly / monthly peak electricity consumption periods, average power, equipment start-up and shutdown times, and power surge points; the user basic information includes at least the user type, consumption level, type and quantity of adjustable equipment, and equipment intelligence level.

3. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 1, characterized in that, The preprocessing in S1 includes the following steps: S11. Use the 3σ criterion to remove outliers from the multi-source data. Outliers include data where the power suddenly drops to 0 and the voltage exceeds the rated range by ±10%. S12. Use linear interpolation to complete the missing time series data in the multi-source data, and the time resolution of the completed data is not less than 15 minutes / time. S13. The preprocessed multi-source data is standardized using the following formula: ， Among them, the For the standardized multi-source data, the The original data of the multi-source data after preprocessing, the The minimum value in the original data, This is the maximum value in the original data.

4. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 1, characterized in that, The construction of a multi-dimensional attribute user behavior profile for a typical adjustable device on the user side in S2 specifically includes the following steps: S21. Associate the preprocessed multi-source data with a preset multi-dimensional attribute system, which includes basic attributes, behavioral attributes, and device attributes. S22. Based on the multi-source data and the preset multi-dimensional attribute system, user behavior patterns are extracted using association rules. The behavior patterns include at least the overlap between peak electricity consumption periods and adjustable equipment operating periods, and the equipment start-up and shutdown frequency under different electricity price ranges. S23. Segment users and generate a unique behavioral profile for each user based on the segmentation results and tagging technology.

5. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 4, characterized in that, In step S21, the basic attributes include user characteristics, consumption level, and energy composition; the behavioral attributes include responsiveness and load characteristics; and the equipment attributes include adjustable equipment type and equipment intelligence level.

6. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 5, characterized in that, S3 specifically includes: S31. Construct dedicated control models for various types of typical adjustable devices, including electric vehicle charging piles / stations and photovoltaic-storage-charging microgrids; S32. Based on the multi-dimensional attribute user behavior profile, extract the attributes of the response positivity, and reduce the basic regulatory potential output by the dedicated regulatory model using the following formula: ， Among them, the The actual regulatory potential after reduction, the For the aforementioned basic regulatory potential, the The response incentive reduction factor; S33. Collect the operating status of the typical adjustable equipment, update the actual control potential, form the real-time control potential, and output the quantitative results of adjustable capacity and response time.

7. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 6, characterized in that, In step S31, the dedicated control model for the electric vehicle charging pile / station is a correlation model of charging power, charging time, and remaining power. During the output of the basic control potential by the correlation model, when the remaining power of the charging pile is >80%, charging is suspended for 2-4 hours. The basic control potential is equal to the rated power of the charging pile. When the remaining power of the charging pile is between 50% and 80%, the charging power is reduced by 30% to 50%. The basic control potential is equal to the product of the rated power and the ratio of the reduced charging power. When the remaining power of the charging pile is <50%, the basic control potential is 0.

8. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 6, characterized in that, In S31, the dedicated control model of the photovoltaic-storage-charging microgrid is a correlation model of photovoltaic output, energy storage capacity, and composite demand. During the process of outputting the basic control potential in the correlation model, when the photovoltaic output is greater than the load demand and the energy storage capacity is not full, the basic control potential is equal to the difference between the photovoltaic output and the load demand. When the photovoltaic output is less than the load demand and the energy storage capacity is surplus, the basic control potential is equal to the energy storage release power. When the energy storage capacity is full and the photovoltaic output is less than or equal to the load demand, the basic control potential is 0.

9. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 6, characterized in that, The specific components of constructing the user-side load adjustment cost calculation model in S4 include: S41. Deconstruct the composition of the user-side load regulation cost, the composition of which includes power loss cost, regulation incentive cost, and equipment investment conversion cost; S42. Calculate each cost item of the structure respectively, and obtain the total adjustment cost based on the calculation results, and construct the calculation model for the user-side load adjustment cost.

10. The method for accurate user behavior profiling based on typical adjustable devices on the user side as described in claim 9, characterized in that, The output control excitation scheme in S4 specifically includes: S43. Determine the control scenarios of the regional power grid, wherein the control scenarios include at least peak shaving scenarios, transformer power backflow / heavy overload scenarios, photovoltaic consumption sites, and user energy cost saving scenarios; S44. Based on the quantification results of the adjustable capacity and the response time, and considering the total adjustment cost, construct a multi-objective optimization model. S45. Solve the multi-objective optimization model and output load-side resource coordinated regulation incentive schemes for multiple regulation scenarios. The regulation incentive schemes include at least prioritizing the use of adjustable equipment by users with high responsiveness and prioritizing the use of power fed back from the photovoltaic-storage-charging microgrid in the photovoltaic consumption scenario.

Images