Direct current stacking storage-based power grid connection management system

Abstract

The application discloses a kind of based on direct current stacking storage's electric energy grid-connected management system, comprising: user aggregation information construction module, for obtaining the user information of each charging station in target area, constructs user quantity knowledge graph;User habit construction module, for obtaining the charging habit of user, constructs charging habit knowledge graph;Fusion module is based on the user quantity knowledge graph and charging habit knowledge graph, generates fusion knowledge graph;Information output module, pre-set charging station energy charging habit, according to fusion knowledge graph generates the energy charging and discharging plan of all charging stations in target area;Energy charging and discharging plan includes the charging capacity of each time period and the discharge capacity of each time period.In the technical scheme provided in the application, the use frequency of battery in charging station can be greatly increased, the idle period of battery is reduced, and the benefit of charging station is increased.

Description

Technical Field

[0001] This application relates to the field of charging station grid connection technology, and more specifically, to a power grid connection management system based on DC superimposed energy storage. Background Technology

[0002] Urban charging stations typically include a power acquisition module, a power distribution module, and a DC-DC power storage unit. The power acquisition module collects electrical energy from the power grid. The DC-DC power storage unit stores electrical energy, and the power distribution module distributes the power to each user (charging gun). Due to cost constraints during construction, the total power collected by the power acquisition module is less than the total power required for all users to operate simultaneously. Therefore, in practice, the DC-DC power storage unit is needed to compensate for the power transmission capacity limitations of the power acquisition module. During operation, the DC-DC power storage unit can receive or release high currents. This allows for rapid response to dispatch demands and timely power delivery to the power distribution module.

[0003] In practice, urban DC-DC charging stations exhibit a clear cyclical pattern. During peak daytime electricity demand, most vehicles are used for production and daily life, resulting in fewer vehicles charging. However, a large number of vehicles come to charge at night. Consequently, the energy storage systems of urban charging stations experience extended periods of idle time, reducing equipment utilization. Summary of the Invention

[0004] The summary section of this application is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0005] To address the technical problems mentioned in the background section, some embodiments of this application provide a power grid-connected management system based on DC superimposed energy storage, including:

[0006] The user aggregation information construction module is used to obtain user information of each charging station in the target area and construct a user quantity knowledge graph.

[0007] The user habit building module is used to obtain users' charging habits and build a charging habit knowledge graph.

[0008] The fusion module generates a fused knowledge graph based on the user quantity knowledge graph and the charging habit knowledge graph.

[0009] The information output module generates charging and discharging plans for all charging stations within the target area based on pre-set charging station charging habits and a fused knowledge graph.

[0010] The charge / discharge plan includes the amount of charge and discharge for each time period.

[0011] This application constructs a knowledge graph that integrates regional vehicle aggregation and charging habits, enabling accurate prediction of the total future charging and discharging demand of all charging stations within the target area. Based on this prediction, the system can provide the power grid with a clear overall charging and discharging plan in advance, significantly reducing the pressure on power grid dispatch and reducing the resistance to large-scale energy storage grid connection. At the same time, this solution effectively increases the utilization frequency of energy storage batteries at charging stations within the area, significantly reduces equipment idle time, and thus significantly enhances the overall operational efficiency of charging stations.

[0012] When constructing a knowledge graph of user numbers, user clustering habits are highly influenced by multiple factors such as time, weather, holidays, and temperature. Static graphs built solely based on historical clustering samples (such as 24-hour changes in the number of users at each station plus environmental factors) are insufficient to effectively capture and express the deep relationships between these complex factors. This results in a lack of correlation between graph elements and inaccurate descriptions of user clustering relationships, ultimately limiting the predictive power and practicality of the graph.

[0013] Furthermore, the user aggregation information construction module includes:

[0014] The initial map construction unit is used to generate a preliminary distribution map based on the input historical user aggregation information.

[0015] The graph information update unit dynamically updates the preliminary distribution graph based on the correlation between adjacent elements, forming a user quantity knowledge graph.

[0016] The historical user aggregation information includes several aggregation samples. Each aggregation sample includes user distribution information, weather, holidays, and temperature for each charging station within 24 hours.

[0017] This application constructs a preliminary distribution map based on historical user aggregation information (including multi-dimensional samples such as changes in the number of users, weather, holidays, and temperature), and crucially introduces neighbor element correlation analysis for dynamic updates, effectively solving the problem of missing correlations caused by the interaction of multiple factors. This dynamic update mechanism can uncover and supplement cross-factor correlations that do not exist explicitly in the original data, breaking through the boundary limitations of the initial samples, and ultimately generating a more complete, more correlated, and more accurate user aggregation pattern that reflects the coupling of complex factors. This significantly improves the accuracy and practicality of the map, laying a solid foundation for subsequent accurate prediction of regional charging and discharging demand.

[0018] Existing knowledge graph construction schemes face significant challenges in transforming multi-source heterogeneous data (such as user changes, weather, holidays, and temperature) into graph elements and relationships: on the one hand, the correspondence between elements and potential relationships in massive amounts of data is unclear, making it difficult to automatically and accurately identify and associate them; on the other hand, the vast number of label types leads to a decrease in the accuracy of the element clustering process, making it impossible to effectively merge elements with similar semantics or structures, which seriously restricts the ability of the initial graph to accurately represent user distribution information and its complex associations with multi-dimensional environmental factors.

[0019] Furthermore, the preliminary distribution map construction process includes:

[0020] S1: Clean all historical user aggregation information and remove abnormal data;

[0021] S2: Convert user change information, weather, holidays, and temperature at each charging station within 24 hours into elements;

[0022] S3: A model combining LSTM encoding and CRF decoding is used to identify elements in order to establish a set of candidate relation groups containing element associations;

[0023] S4: Merge similar elements to generate a preliminary distribution map.

[0024] This application constructs a preliminary distribution map using a joint LSTM encoding and CRF decoding model, effectively overcoming the challenges of element identification and association: 1) The LSTM-CRF model can deeply capture the potential association patterns of structurally similar elements in sequence data (such as 24-hour user changes), achieving accurate identification of relationships between elements and generating a high-quality set of candidate relationship groups; 2) The model's powerful sequence labeling capability significantly improves the accuracy of identifying similar elements in a large label space, ensuring the accuracy of merging processing. The resulting preliminary distribution map not only more accurately represents the distribution information of user numbers but also clearly distinguishes and characterizes the essential associations and differences between user change information and factors of different dimensions such as weather, holidays, and temperature, laying a high-precision data foundation for subsequent dynamic updates of the map.

[0025] A key challenge in constructing a knowledge graph of user numbers is establishing the correspondence between user change information and environmental factors such as weather, holidays, and temperature. This challenge stems from the significant "many-to-many" mapping relationship in historical user aggregation information (i.e., the same user change pattern can correspond to different environmental combinations, and vice versa). This inherently uncertain mapping relationship makes the correlation between elements (user change vs. environmental factors) extremely vague and difficult to identify directly and accurately, severely hindering the generation of high-quality candidate relationship groups and becoming the main bottleneck in constructing a knowledge graph that accurately reflects user aggregation patterns.

[0026] Step S3 includes:

[0027] S31: Extract corresponding relationships from historical user aggregation information using a joint extraction model to generate a candidate relationship set;

[0028] S32: Extract the association probability of elements in the candidate relationship group set, and use the BERT relationship classification model to re-predict the association probability of element pairs. Here, the elements are user change information, weather, holidays and temperature, and the association probability is the probability that the elements are related to each other.

[0029] S33: If the newly predicted association probability is consistent with the original association probability, then retain the candidate relationship set; if they are inconsistent, then update the association probability according to the classification result.

[0030] This application innovatively employs a dual verification and correction mechanism combining a joint extraction model and a BERT relationship classification model: 1) The joint extraction model first extracts a preliminary set of candidate relationship groups from complex historical data; 2) The BERT relationship classification model, based on in-depth analysis of the correlation between adjacent clustered samples, re-predicts and verifies the association probability of element pairs (user change - environmental factors). The core advantage of this mechanism is its ability to effectively identify and dynamically correct association biases caused by small-scale randomness in historical data (updating when the BERT predicted probability differs from the original probability). Ultimately, this scheme provides a high-quality relationship data foundation for knowledge graph construction that has undergone dual verification and has more accurate and reliable association probabilities, significantly improving the accuracy and reliability of subsequent graph construction.

[0031] When constructing the correspondence between user change information and environmental factors such as weather, holidays, and temperature, the core challenge lies in the non-deterministic "many-to-many" mapping that is prevalent in historical user aggregation information (i.e., a single user change pattern can correspond to multiple environmental combinations, and vice versa). This inherent complexity leads to: (1) the potential correlation between elements (such as user change vs. weather) being extremely vague and difficult to identify and quantify directly and accurately; and (2) traditional methods failing to effectively capture the dynamic hierarchical dependence characteristics of relationships between elements (i.e., the correlation probability needs to be judged in conjunction with the meaning information of neighboring aggregated samples). This directly restricts the accuracy and reliability of the generated candidate relationship set, becoming a key obstacle to constructing a high-quality knowledge graph.

[0032] Furthermore, S31 includes the following steps:

[0033] S311: Input historical user aggregation information D, D={C1, C2, C...} i ...C N}, C i Let N represent the i-th clustered sample in the historical user clustering information, and let C represent the total number of clustered samples.i ={e 1i e 2i e 3i e 4i}, e 1i e 2i e 3i e 4i Representing the clustered samples C respectively i The first, second, third, and fourth elements are: the first element is user change information, the second element is weather, the third element is holidays, and the fourth element is temperature. The user cluster information D is transformed into a vector using the BERT relationship classification model.

[0034] S312: Aggregate sample C i The input is fed into the TPLinker network to generate a candidate relation group set T, T={(e h ,r,e t )}, where e h and e t These are the forward and backward associated elements, respectively, and r is the association probability.

[0035] Furthermore, S32 includes the following steps:

[0036] S321: Input historical user aggregation information D and candidate relation group set T;

[0037] S322: Positioning e h and e t Extract the clustered sample set E containing forward-related elements and backward-related elements from the historical user clustering information D.

[0038] E={x s-k ...x s ...x e ...x e+k}, where s and e represent the starting and ending numbers of the extracted clustered samples, respectively, and k represents the extraction step size;

[0039] S323: Input the clustered sample set E into the BERT relation classification model to obtain the hidden state h of CLS. [CLS] h [CLS] ∈R;

[0040] Input=[[CLS]]⊕Context⊕[[SEP]];

[0041] Where ⊕ represents the concatenation operation, Context represents the aggregated sample set E, [CLS] represents the classification label, located at the beginning number of the input sequence, and [SEP] represents the separator, located at the end number; S32: Calculate the association probability of the corresponding elements;

[0042] P(r|e) h e t E) = Softmax(W*h [CLS] +b);

[0043] Where W represents the classification weight matrix, b represents the bias vector, and P(r|e h e t E) represents the condition given e under the constraints of the selected relation set and the clustered sample set. h and e t When aggregating the sample set E, r is used as e h and e t The probability of association of effective links between them, where Softmax represents the probability transformation function;

[0044] S325: Retain the set T of candidate relation groups whose association probability is higher than the preset value.

[0045] This application effectively solves the aforementioned challenges through a refined relation construction process that combines the TPLinker network and the BERT relation classification model: the TPLinker network efficiently extracts preliminary candidate relation groups (eh,r, et) from the original aggregated samples, laying the foundation for relation identification. The BERT relation classification model, on the other hand, is based on pairs containing target elements (e...). h e t We perform in-depth analysis on the aggregated sample set E, and recalculate and verify the association probability P(r|eh, et, E).

[0046] The core advantage of this scheme lies in its introduction of a neighbor-sample awareness mechanism (via E) into relation construction, enabling the calculation of association probabilities to dynamically adapt to the inherent randomness and dependency characteristics of neighbor samples in historical data. Thus, by selecting high-probability relation groups, this scheme can not only flexibly construct complex associations with single or multiple elements (precisely controlling relation complexity), but also generate highly accurate, context-aware correspondence graphs, providing a robust and high-precision relational data foundation for subsequent knowledge graph construction.

[0047] When constructing a preliminary distribution map based on multi-source heterogeneous historical user aggregation information (user changes, weather, holidays, temperature), the core challenges faced are map redundancy and high-dimensional complexity: a large number of semantically similar or repetitive elements in the original data (such as similar weather or similar holiday types expressed in different ways) are not effectively identified and merged, resulting in a large set of elements and information redundancy. These redundant elements further generate a large number of invalid or repetitive correspondences. These problems together lead to an excessively complex and highly dimensional set of candidate relation groups, which not only significantly increases the difficulty and computational cost of constructing the knowledge graph, but also reduces the clarity, interpretability, and efficiency of subsequent processing of the graph.

[0048] Furthermore, S4 includes the following steps:

[0049] S41: Get all the elements to obtain the element set A = {e1, e2, ..., e...} n}, where n represents the total number of elements;

[0050] S42: Generate an element database based on historical user aggregation information, and map the elements to the element database;

[0051] S43: Calculate the similarity of each element, and merge elements with similarity within a preset threshold into 1 element;

[0052] S44: Update the candidate relation group set T={(e) based on the merged elements. h ,r,e t )}, and update the candidate relation set T={(e h ,r,e t Mapping to a low-dimensional space generates a vector representation of each element and relation;

[0053] S45: Construct a preliminary distribution map based on the merged correspondences and candidate relationship sets.

[0054] This application effectively solves the graph redundancy problem by implementing a structured element merging and relation cleanup process (S41-S45). Element merging based on semantic similarity (S43) significantly simplifies the element set and eliminates information redundancy. Simultaneous cleanup and updating of the candidate relation set (S44) removes redundant or invalid relations generated by element merging. Crucially, a low-dimensional vector space mapping (S44) is introduced to transform the merged elements and relations into compact vector representations. The core advantage of this scheme lies in its simultaneous dimensionality reduction at both the element and relation levels, systematically reducing the structural complexity and information dimensionality of the knowledge graph. The resulting preliminary distribution graph not only has significantly reduced construction difficulty but also possesses higher information density, stronger interpretability, and better computational efficiency.

[0055] The user count knowledge graph needs to be generated using a preliminary distribution graph. In practice, the relationships in the preliminary distribution graph are too complex, meaning the constructed user count knowledge graph is an initial, balanced knowledge graph that cannot focus too much on the relationships between two elements according to the requirements. Furthermore, different extraction methods have varying degrees of concern regarding the relationships between different elements, so the preliminary distribution graph needs to be revised.

[0056] Furthermore, the method for generating the user quantity knowledge graph includes the following steps:

[0057] Z1: Obtain preliminary distribution map and new historical user clustering information G0;

[0058] Where G0 = (e0, r0, T0), e0, r0, and T0 represent elements, relations, and candidate relation sets, respectively.

[0059] Z2: Use the TransE model to train the initial distribution map embedding and calculate the implicit relationship score of each element;

[0060] Where h represents the vector of forward elements, r represents the vector of association probabilities, and t represents the vector of backward elements. Indicates the margin parameter. Indicates the range of norms;

[0061] Z3: Add relationships with implicit relationships higher than the preset value to the preliminary distribution map;

[0062] Z4: Based on the methods of S31 and S32, a set of candidate relationship groups is generated from new historical user aggregation information;

[0063] Z5: Compare the cosine similarity between the candidate relation group set of the preliminary distribution map and the newly extracted candidate relation group set. Merge the candidate relation group set with similarity less than the preset value and the newly extracted candidate relation group set into one candidate relation group set. Generate a user number knowledge graph based on all candidate relation group sets.

[0064] In the technical solution provided in this application, based on the initial knowledge graph, special historical user aggregation information is used to update the initial knowledge graph, so that a set of directional candidate relationship groups can be added on the basis of a balanced initial distribution graph.

[0065] When fusing user aggregation patterns (user quantity knowledge graph) with user charging behavior (charging habit knowledge graph), existing methods face significant challenges: 1) The knowledge structures and dynamic characteristics of the two graphs are fundamentally different (user aggregation is strongly driven by the external environment and changes rapidly, while charging habits are relatively stable), making effective alignment and deep integration difficult; 2) Traditional fusion strategies (such as simple splicing or static weighting) cannot adaptively capture and weigh the contribution of relational information at different graph levels (propagation layers), resulting in a coarse fusion process, severe information loss, and the final fused graph failing to accurately and synergistically reflect the complex relationship between user distribution information and charging volume.

[0066] Furthermore, the user habit construction module uses a preliminary distribution map scheme to generate a charging habit knowledge graph;

[0067] The fusion module integrates the charging habit knowledge graph and the user number knowledge graph according to the following steps:

[0068] Step 1: Obtain the propagation layers and element expansion set of the charging habit knowledge graph;

[0069] Step 2: Use a neural network to learn the relational information of candidate relation groups in each layer, so as to dynamically allocate weights;

[0070] Step 3: Gradually weight and aggregate the element vectors of each layer to generate hierarchical features;

[0071] Step 4: Integrate the features of each layer, construct the element link relationship, and generate the fusion map based on the element link relationship.

[0072] Furthermore, step 4 specifically includes:

[0073] Step 4 specifically includes:

[0074] Step 41: Calculate the initial features used to describe user distribution information based on the user quantity knowledge graph and the charging habit knowledge graph. and initial features used to describe the amount of charge ;

[0075] ;

[0076] , This represents the original feature set related to user distribution information. Let e ​​represent the original feature set related to the charging amount, and let 'e' represent an element in the original feature set.

[0077] Step 42: Generate an initial feature set describing user distribution information and the initial feature set describing the amount of charge ;

[0078] ;

[0079] ;

[0080] Step 43: Establish the link between the initial feature set describing user distribution information and the initial feature set describing charging amount;

[0081] Step 44: Based on the link relationships between the initial feature set describing user distribution information and the initial feature set describing charging amount, fuse the charging habit knowledge graph and the user quantity knowledge graph to generate a fused graph.

[0082] This application effectively solves the problem of heterogeneous knowledge graph fusion through an innovative hierarchical dynamic weighted fusion and dual feature set linking mechanism: 1) Hierarchical learning and dynamic weight allocation (steps 2-3): Utilizing neural networks to learn information from candidate relation groups in each propagation layer, dynamically allocating fusion weights ensures differentiated and adaptive integration of knowledge at different levels and with different dynamic characteristics; 2) Dual feature set linking (steps 41-43): Extracting and explicitly establishing the link relationship between the initial feature set describing user distribution information and the initial feature set describing charging volume, providing a structured expression basis for core associations. This scheme achieves refined and hierarchical fusion of dynamic information on user aggregation and stable information on charging habits. The resulting fused knowledge graph not only significantly improves the accuracy of the fusion of the two types of data but also accurately and collaboratively depicts the deep correlation between user distribution information and charging volume within the target area, providing solid and reliable data support for the subsequent generation of high-precision regional charging and discharging plans. Attached Figure Description

[0083] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application.

[0084] Furthermore, throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the elements are not necessarily drawn to scale.

[0085] In the attached diagram:

[0086] Figure 1 This is a schematic diagram of the charging station.

[0087] Figure 2 This is a schematic diagram of the power grid-connected management system based on DC superimposed energy storage. Detailed Implementation

[0088] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While some embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this application. It should be understood that the drawings and embodiments of this application are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0089] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0090] The present application will now be described in detail with reference to the accompanying drawings and embodiments.

[0091] refer to Figure 1 The DC charging station of this application generally includes a DC power storage unit, a grid-connected unit, a management unit, and a charging unit. Both the grid-connected unit and the DC power storage unit are connected to the charging unit. The charging unit supplies power to vehicles entering the charging station; the grid-connected unit is connected to the power grid and is used to obtain power from the grid and supply it to the charging unit; the DC power storage unit stores electricity and assumes part of the load when the charging unit is under high load. The electrical energy of the DC power storage unit is provided by the power grid.

[0092] The specific operation is as follows: When electricity prices are low at night, the DC-DC power storage unit draws power from the grid to fully charge its stored energy; when electricity prices are high during the day, the DC-DC power storage unit can feed power back to the grid, thereby profiting from the price difference. Furthermore, when there are many vehicles connected to the charging unit and the grid-connected unit cannot obtain a sufficiently high charging power from the grid at once, the DC-DC power storage unit can also supply power to the charging unit to supplement the remaining load gap of the charging unit.

[0093] Therefore, the energy storage units (DC cascade energy storage units) in conventional charging stations serve to enhance load capacity and store electrical energy. However, due to the large fluctuations in the number of charging vehicles and the severe disorder in charging behavior, the energy storage units mostly serve as a supplementary source of power for the charging units, making it difficult to achieve large-scale, stable reverse power transmission to the grid. Furthermore, the grid itself typically restricts the connection of such units with unstable power transmission.

[0094] Based on this, refer to Figure 2A grid-connected power management system based on DC-DC cascaded energy storage includes: a user aggregation information construction module, a user habit construction module, a fusion module, and an information output module. The user aggregation information construction module acquires user information from each charging station within a target area and constructs a user quantity knowledge graph. The user habit construction module acquires users' charging habits and constructs a charging habit knowledge graph. The fusion module generates a fused knowledge graph based on the user quantity knowledge graph and the charging habit knowledge graph. The information output module, based on pre-set charging station charging habits and the fused knowledge graph, generates charging and discharging plans for all charging stations within the target area. The charging and discharging plans include the charging amount and discharging amount for each time period.

[0095] In grid connection management, to input stored electrical energy into the grid, a request must be sent to the grid in advance to alleviate the grid's dispatching pressure. Otherwise, small-scale grid connection requests may not receive a response from the grid. However, individual charging stations have limited energy storage capacity and unstable power transmission, making it difficult to independently initiate grid connection applications. Therefore, this application merges multiple charging stations within the target area into a single entity to initiate a grid connection application. This results in a cluster of charging stations with stronger energy storage capacity, more accurate timing of power demand requests to the grid, and more accurate power supply predictions. This, in turn, ensures that the charging station's energy storage batteries receive more charge-discharge cycles, reducing their idle time.

[0096] In actual grid connection management, whether a charging station supplies power to the grid and the amount of power supplied are influenced by a combination of factors, including electricity prices, battery life, and the charging station company's management strategies. Overly rigid grid connection plans hinder companies from flexibly adjusting their strategies. Therefore, this application does not directly calculate the operating efficiency of charging stations, but rather integrates user aggregation and charging patterns into a knowledge graph. After obtaining a complete knowledge graph depicting external factors, companies can adjust their charging and discharging plans based on their own established charging habits.

[0097] In essence, the company's internal factors (management strategies) are used to set charging habits, while external factors (user behavior and environment) are used to generate a fused knowledge graph. Combining this fused knowledge graph with charging habits generates a charging and discharging plan. This plan accurately guides when charging stations are used for vehicle charging and when they are used to supply power to the grid. The key to this application lies in generating the fused knowledge graph. After generating the fused knowledge graph, the information input module can quickly locate corresponding elements and relationships from the fused knowledge graph based on the user-inputted charging station charging habits, thereby generating the necessary pre-selected information and generating the charging and discharging plan based on this pre-selected information.

[0098] Therefore, the key to this application lies in how to generate a fused knowledge graph, which requires first generating a user quantity knowledge graph and a charging habit knowledge graph. The following describes the generation process of the user quantity knowledge graph and the charging habit knowledge graph.

[0099] The user aggregation information construction module is used to generate a user quantity knowledge graph. The user aggregation information construction module includes: an initial graph construction unit and a graph information update unit.

[0100] The initial map construction unit is used to generate a preliminary distribution map based on the input historical user aggregation information.

[0101] The graph information update unit dynamically updates the initial distribution graph based on the correlation between adjacent elements, forming a user quantity knowledge graph.

[0102] Historical user aggregation information includes several aggregated samples. Each aggregated sample includes user distribution information, weather, holidays, and temperature for each charging station within 24 hours. User distribution information, weather, holidays, and temperature are elements of the aggregated sample.

[0103] User distribution information, weather, holidays, and temperature are all coded. User distribution information represents the number of charging users at each charging station within the target area per hour over a 24-hour period. This information reflects user aggregation within the target area at different times. Weather refers to the weather with the longest daytime duration. Weather tags include rain, overcast, sunny, and cloudy. Temperature primarily refers to the average daily temperature. Holidays indicate whether the day is a workday, such as weekdays, weekends, winter vacation, summer vacation, or National Day holiday. Holidays and weather are discrete coded tags, while temperature is a continuous coded tag. These three elements—weather, holidays, and temperature—affect user distribution information. Additionally, information such as the number of shopping malls and companies in the target area can be incorporated. This application considers weather, holidays, and temperature. These three factors are most correlated with users' electricity consumption habits and therefore require priority consideration. Based on this, a preliminary distribution map is constructed, which clearly shows the relationship between user distribution information, weather, holidays, and temperature.

[0104] User distribution information, weather, holidays, and temperature are converted into codes using existing technology. Weather, holidays, and temperature are one-dimensional data. The specific conversion methods will not be elaborated here.

[0105] User distribution information is three-dimensional data, consisting of time, quantity, and charging station coordinates. Therefore, it is transformed into a three-dimensional data matrix, which is then encoded. For example, the three-dimensional user distribution information (time, quantity, and charging station coordinates) is structured as a three-dimensional data matrix (dimensions: time point × charging station location × number of users). Encoding techniques (such as neural networks and tensor decomposition) are then used to automatically learn and extract the spatiotemporal distribution features contained in this matrix, providing a low-dimensional, dense vector representation for subsequent knowledge graph construction or fusion.

[0106] Furthermore, the preliminary distribution map construction process includes:

[0107] S1: Clean all historical user aggregation information and remove abnormal data.

[0108] Abnormal data mainly refers to data that lacks key elements, such as weather or temperature. These incomplete data need to be removed to avoid affecting subsequent weighting.

[0109] S2: Convert user change information, weather, holidays, and temperature at each charging station within 24 hours into elements.

[0110] The conversion to elements here actually involves encoding user change information, weather, holidays, and temperature. The specific encoding method has been described earlier.

[0111] S3: A model combining LSTM encoding and CRF decoding is used to identify elements in order to establish a set of candidate relation groups containing element associations;

[0112] Step S3 includes the following steps:

[0113] S31: Extract corresponding relationships from historical user aggregation information through a joint extraction model to generate a set of candidate relationship groups.

[0114] S31 mainly extracts the corresponding relationships from each element, thereby obtaining a set of candidate element relationship groups. The set of candidate element relationship groups includes (forward elements, relationships, and backward elements).

[0115] In this application, the forward element is user change information, and the backward element is one of "weather, holidays and temperature", with the relationship being the probability of the two.

[0116] S31 includes the following steps:

[0117] S311: Input historical user aggregation information D, D={C1, C2, C...} i ...C N}, C iLet N represent the i-th clustered sample in the historical user clustering information, and let C represent the total number of clustered samples. i ={e 1i e 2i e 3i e 4i}, e 1i e 2i e 3i e 4i Representing the clustered samples C respectively i The first, second, third, and fourth elements are: the first element is user change information, the second element is weather, the third element is holidays, and the fourth element is temperature. The user cluster information D is transformed into a vector using the BERT relationship classification model.

[0118] S312: Aggregate sample C i The input is fed into the TPLinker network to generate a candidate relation group set T, T={(e h ,r,e t )}, where e h and e t These are the forward and backward associated elements, respectively, and r is the association probability.

[0119] In this approach, the BERT relation classification model transforms the encoding into vectors, which are then used to classify words. In this approach, C is first... i The BERT relation classification model is used to convert the data into vectors. Then, the TPLinker network is used to generate EH-to-ET, SH-to-OH, and ST-to-OT matrices. Finally, the intersection of the three matrices is analyzed to generate a set of candidate relation groups.

[0120] The essential function of the TPLinker network is to identify correspondences. The principle is to extract e from historical user aggregation information D. 1i e 2i e 3i e 4i The probability of two elements being in the same cluster sample is used to generate the association probability. Therefore, it is possible to transform multivariate cluster samples into association samples with only three elements, thus reducing the dimensionality of the data.

[0121] S32: Extract the association probability of elements in the candidate relationship group set, and use the BERT relationship classification model to re-predict the association probability of element pairs. Here, the elements are user change information, weather, holidays and temperature, and the association probability is the probability that the elements are related to each other.

[0122] S32 includes the following steps:

[0123] S321: Input historical user aggregation information D and candidate relation group set T.

[0124] S322: Positioning e h and e t Extract the clustered sample set E containing forward-related elements and backward-related elements from the historical user clustering information D.

[0125] E={x s-k ...x s ...x e ...x e+k}, where s and e represent the starting and ending numbers of the extracted clustered samples, respectively, and k represents the extraction step size.

[0126] The historical user cluster information D is not sorted randomly, but according to the coding order, such as the coding order of temperature. Therefore, adjacent cluster samples in historical user cluster information D have high similarity. Then, a cluster sample set E containing forward and backward association elements is extracted. In this set E, the cluster samples are relatively similar.

[0127] S323: Input the clustered sample set E into the BERT relation classification model to obtain the hidden state h of CLS. [CLS] h [CLS] ∈R;

[0128] Input=[[CLS]]⊕Context⊕[[SEP]];

[0129] Where ⊕ represents the concatenation operation, Context represents the aggregated sample set E, [CLS] represents the classification label, located at the beginning number of the input sequence, and [SEP] represents the separator, located at the end number;

[0130] S324: Calculate the association probability of corresponding elements;

[0131] P(r|e) h e t E) = Softmax(W*h [CLS] +b);

[0132] Where W represents the classification weight matrix, b represents the bias vector, and P(r|e h e t E) represents the condition given e under the constraints of the selected relation set and the clustered sample set. h and e t When aggregating the sample set E, r is used as e h and e t The probability of association of effective links between them, where Softmax represents the probability transformation function;

[0133] S325: Retain relationships with an association probability higher than a preset value, including the association probability of element pairs.

[0134] S33: If the newly predicted association probability is consistent with the original association probability, then retain the candidate relationship set; if they are inconsistent, then update the association probability according to the classification result.

[0135] In S31 and S32, S31 first determines which information in the database belongs to the elements, reducing the information dimensionality when calculating the correspondence in the subsequent process. Then, in S2, the Softmax function is used to further calculate the relationships between the elements, thereby obtaining accurate correspondences and updating the candidate relationship set.

[0136] S4: Merge similar elements to generate a preliminary distribution map.

[0137] S4 includes the following steps:

[0138] S41: Get all the elements to obtain the element set A = {e1, e2, ..., e...} n}, where n represents the total number of elements.

[0139] In S3, all candidate relation groups were obtained. Extracting each element from these candidate relation groups yielded an element set. The words in this element set were the words that needed to be merged based on similarity.

[0140] S42: Generate an element database based on historical user aggregation information, and map the elements to the element database.

[0141] The element database here is actually based on the repetition rate of each element in the historical user aggregation information. For each element, after mapping to the element database, a new vector representation can be obtained.

[0142] S43: Calculate the similarity of each element, and merge elements with similarity within a preset threshold into one element.

[0143] S44: Update the candidate relation group set T={(e) based on the merged elements. h ,r,e t )}, and update the candidate relation set T={(e h ,r,e t Mapping to a low-dimensional space generates a vector representation of each element and relation. Then, by calculating the cosine similarity of the elements and comparing it to a similarity threshold between elements, it can be determined whether to retain an element. For example, if the cosine similarity between element A and element B is greater than a preset threshold, then element A and element B represent the same meaning, and thus element A and element B are merged into one element.

[0144] S45: Construct a preliminary distribution map based on the merged correspondences and candidate relationship sets.

[0145] Merging elements in S44 further reduces the number of candidate relation sets. All candidate relation sets are then used to construct a preliminary distribution graph. Using candidate relation sets to construct a knowledge graph is an existing technique; therefore, the specific construction scheme will not be further elaborated here.

[0146] The method for generating the user count knowledge graph includes the following steps:

[0147] Z1: Obtain new corpus information from the preliminary distribution map;

[0148] Where G0 = (e0, r0, T0), e0, r0, and T0 represent elements, relations, and candidate relation sets, respectively.

[0149] The new historical user cluster information is actually the extracted historical user cluster information containing T0. In essence, it requires further analysis of the relationship between e0 and T0.

[0150] For example, T0 indicates that the correlation between user change information a and weather b is 20%, while G0 treats T0 as a whole and re-analyzes the correspondence between temperature and T0.

[0151] Z2: Use the TransE model to train the initial distribution map embedding and calculate the implicit relationship score of each element;

[0152] Where h represents the vector of forward elements, r represents the vector of association probabilities, and t represents the vector of backward elements. Indicates the margin parameter. Indicates the range of norms;

[0153] Z3: Add relationships with implicit relationships higher than the preset value to the preliminary distribution map;

[0154] Z4: Based on the methods of S31 and S32, a set of candidate relation groups is generated from new corpus information;

[0155] Z5: Compare the cosine similarity between the candidate relation group set of the preliminary distribution map and the newly extracted candidate relation group set. Merge the candidate relation group set with similarity less than the preset value and the newly extracted candidate relation group set into one candidate relation group set. Generate a user number knowledge graph based on all candidate relation group sets.

[0156] The user habit building module uses a preliminary distribution map scheme to generate a charging habit knowledge graph.

[0157] The key to building the user habit knowledge graph lies in extracting the average electricity consumption of users from their charging data. For example, when the temperature is low, vehicles consume more electricity, resulting in vehicles arriving at charging stations with lower battery levels and requiring more charging per vehicle. The data acquired when building the charging habit knowledge graph is based on historical charging data with the added dimension of charging volume. Therefore, the charging habit knowledge graph has the same structure as the user quantity knowledge graph.

[0158] Therefore, there is a significant correlation between the charging habit knowledge graph and the user quantity knowledge graph, but these relationships are difficult to integrate into the user quantity knowledge graph. Therefore, it is necessary to generate the link relationships between elements. This includes the following steps:

[0159] The fusion module integrates the charging habit knowledge graph and the user number knowledge graph according to the following steps:

[0160] Step 1: Obtain the propagation layers and element extension set of the charging habit knowledge graph.

[0161] ;

[0162] in, This represents the set of candidate relation groups for the knowledge graph of charging habits. Let L represent the set of elements in the (l-1)th layer, L represent the layer number of the charging habit knowledge graph, and l represent the index of the layer in the charging habit knowledge graph.

[0163] Step 2: Use a neural network to learn the relational information of candidate relation groups in each layer, so as to dynamically allocate weights;

[0164] Specifically: Calculate the attention weights for each element:

[0165] ;

[0166] ;

[0167] Among them, h i Represents the forward element, r i Represents a relation vector, t i The back element is represented by W1, W2, and W3, which represent three learnable matrices of different sizes. || represents vector concatenation, and ReLU represents the rectified linear activation function. Represents the S activation function;

[0168] Step 3: Gradually weight and aggregate the element vectors of each layer to generate hierarchical features;

[0169] ;

[0170] in, , This represents the size of the candidate relation set at level l. This represents the normalized attention weights.

[0171] Step 4: Integrate the features of each layer, construct the element link relationships, and generate a fusion map based on the element link relationships.

[0172] Step 4 specifically includes:

[0173] Step 41: Calculate the initial features used to describe user distribution information based on the user quantity knowledge graph and the charging habit knowledge graph. and initial features used to describe the amount of charge ;

[0174] ;

[0175] , This represents the original feature set related to user distribution information. Let e ​​represent the original feature set related to the charging quantity, and let e represent an element in the original feature set.

[0176] The set of original features related to user distribution information refers to the underlying features related to user distribution information, such as weather and temperature.

[0177] Step 42: Generate an initial feature set describing user distribution information and the initial feature set describing the amount of charge ;

[0178] ;

[0179] ;

[0180] Step 43: Establish the link between the initial feature set describing user distribution information and the initial feature set describing charging quantity.

[0181] Step 44: Based on the link relationships between the initial feature set describing user distribution information and the initial feature set describing charging amount, fuse the charging habit knowledge graph and the user quantity knowledge graph to generate a fused graph.

[0182] The information output module generates a charging and discharging plan based on the fusion graph. For example, if a charging station's charging habits are to reduce its electricity load on users in order to obtain greater economic benefits, then the charging station will reduce its maximum output power during peak electricity price periods, allowing the DC-DC energy storage unit to supply power to the grid to increase economic efficiency.

[0183] Thus, after setting the charging habits of all charging stations within the target area, the system can accurately predict the most likely user number trends and charging demand for the corresponding time period based on a fused knowledge graph and by inputting real-time temperature, weather, and holiday information. Furthermore, by combining the grid's time-of-use pricing strategy (especially off-peak pricing), an optimized charging and discharging plan can be generated.

[0184] The above description is merely a selection of preferred embodiments of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this application.

Claims

1. A power grid-connected management system based on DC-DC cascaded energy storage, applied in a charging station, wherein the charging station includes a DC-DC cascaded energy storage unit, a grid-connected unit, and a management unit; The charging unit is used to supply power to vehicles entering the charging station; the grid-connected unit is connected to the power grid and is used to obtain power from the power grid and supply it to the charging unit. The DC cascade storage unit is used to store electricity and takes on part of the load when the charging unit is under high load. Its features are, include: The user aggregation information construction module is used to obtain user information of each charging station in the target area and construct a user quantity knowledge graph. The user habit building module is used to obtain users' charging habits and build a charging habit knowledge graph. The fusion module generates a fused knowledge graph based on the user quantity knowledge graph and the charging habit knowledge graph. The information output module generates charging and discharging plans for all charging stations within the target area based on pre-set charging station charging habits and a fused knowledge graph. The charge / discharge plan includes the amount of charge and the amount of discharge for each time period. The user aggregation information construction module includes: The initial map construction unit is used to generate a preliminary distribution map based on the input historical user aggregation information. The graph information update unit dynamically updates the preliminary distribution graph based on the correlation between adjacent elements, forming a user quantity knowledge graph. Among them, the historical user cluster information includes several cluster samples, and each cluster sample includes user distribution information, weather, holidays and temperature of each charging station within 24 hours; The preliminary distribution map construction process includes: S1: Clean all historical user aggregation information and remove abnormal data; S2: Convert user change information, weather, holidays, and temperature at each charging station within 24 hours into elements; S3: A model combining LSTM encoding and CRF decoding is used to identify elements in order to establish a set of candidate relation groups containing element associations; S4: Merge similar elements to generate a preliminary distribution map; S3 includes: S31: Extract corresponding relationships from historical user aggregation information using a joint extraction model to generate a candidate relationship set; S32: Extract the association probability of elements in the candidate relationship group set, and use the BERT relationship classification model to re-predict the association probability of element pairs. Here, the elements are user change information, weather, holidays and temperature, and the association probability is the probability that the elements are related to each other. S33: If the newly predicted association probability is consistent with the original association probability, then retain the candidate relationship set; if they are inconsistent, then update the association probability according to the classification result. S4 includes the following steps: S41: Get all the elements to obtain the element set A = {e1, e2, ..., e...} n }, where n represents the total number of elements; S42: Generate an element database based on historical user aggregation information, and map the elements to the element database; S43: Calculate the similarity of each element, and merge elements with similarity within a preset threshold into 1 element; S44: Update the candidate relation group set T={(e) based on the merged elements. h ,r,e t )}, and update the candidate relation set T={(e h ,r,e t Mapping to a low-dimensional space generates a vector representation of each element and relation; S45: Construct a preliminary knowledge graph based on the merged correspondences and candidate relation sets; The method for generating the user count knowledge graph includes the following steps: Z1: Obtain a preliminary knowledge graph and new historical user aggregation information G0; Where G0 = (e0, r0, T0), e0, r0, and T0 represent elements, relations, and candidate relation sets, respectively; Z2: Use the TransE model to train the initial knowledge graph embedding and calculate the implicit relationship score of each element; Where h represents the vector of forward elements, r represents the vector of association probabilities, and t represents the vector of backward elements. Indicates the margin parameter. Indicates the range of norms; Z3: Add relationships with implicit relationships higher than the preset value to the preliminary knowledge graph; Z4: Based on the methods of S31 and S32, a set of candidate relationship groups is generated from new historical user aggregation information; Z5: Compare the cosine similarity between the candidate relation group set of the preliminary knowledge graph and the newly extracted candidate relation group set. Merge the candidate relation group set with similarity less than the preset value and the newly extracted candidate relation group set into 1 candidate relation group set. Generate a user number knowledge graph based on all candidate relation group sets. The user habit building module uses a preliminary knowledge graph approach to generate a charging habit knowledge graph; The fusion module integrates the charging habit knowledge graph and the user number knowledge graph according to the following steps: Step 1: Obtain the propagation layers and element expansion set of the charging habit knowledge graph; Step 2: Use a neural network to learn the relational information of candidate relation groups in each layer, so as to dynamically allocate weights; Step 3: Gradually weight and aggregate the element vectors of each layer to generate hierarchical features; Step 4: Integrate the features of each layer, construct the element link relationship, and generate the fusion map based on the element link relationship.

2. The power grid connection management system based on DC superimposed energy storage according to claim 1, characterized in that: S31 includes the following steps: S311: Input historical user aggregation information D, D={C1, C2, C...} i ...C N }, C i Let N represent the i-th clustered sample in the historical user clustering information, and let C represent the total number of clustered samples. i ={e 1i e 2i e 3i e 4i }, e 1i e 2i e 3i e 4i Representing the clustered samples C respectively i The first, second, third, and fourth elements are: the first element is user change information, the second element is weather, the third element is holidays, and the fourth element is temperature. The user cluster information D is transformed into a vector using the BERT relationship classification model. S312: Aggregate sample C i The input is fed into the TPLinker network to generate a candidate relation group set T, T={(e h ,r,e t )}, where e h and e t These are the forward and backward associated elements, respectively, and r is the association probability.

3. The power grid connection management system based on DC superimposed energy storage according to claim 2, characterized in that: S32 includes the following steps: S321: Input historical user aggregation information D and candidate relation group set T; S322: Positioning e h and e t Extract the clustered sample set E containing forward-related elements and backward-related elements from the historical user clustering information D. E={x s-k ...x s ...x e ...x e+k }, where s and e represent the starting and ending numbers of the extracted clustered samples, respectively, and k represents the extraction step size; S323: Input the clustered sample set E into the BERT relation classification model to obtain the hidden state h of CLS. [CLS] h [CLS] ∈R; Input=[[CLS]]⊕Context⊕[[SEP]]; Where ⊕ represents the concatenation operation, Context represents the aggregated sample set E, [CLS] represents the classification label, located at the beginning number of the input sequence, and [SEP] represents the separator, located at the end number; S324: Calculate the association probability of corresponding elements; P(r|e h ,e t ,E)=Softmax(W*h [CLS] +b); Where W represents the classification weight matrix, b represents the bias vector, and P(r|e h e t E) represents the condition given e under the constraints of the candidate relation set and the clustered sample set. h and e t When aggregating the sample set E, r is used as e h and e t The probability of association of effective links between them, where Softmax represents the probability transformation function; S325: Retain the set T of candidate relation groups whose association probability is higher than the preset value.

4. The power grid connection management system based on DC superimposed energy storage according to claim 1, characterized in that: Step 4 specifically includes: Step 41: Calculate the initial features used to describe user distribution information based on the user quantity knowledge graph and the charging habit knowledge graph. and initial features used to describe the amount of charge ; ； , This represents the original feature set related to user distribution information. Let e ​​represent the original feature set related to the charging amount, and let 'e' represent an element in the original feature set. Step 42: Generate an initial feature set describing user distribution information and the initial feature set describing the amount of charge ; ； ； Step 43: Establish the link between the initial feature set describing user distribution information and the initial feature set describing charging amount; Step 44: Based on the link relationships between the initial feature set describing user distribution information and the initial feature set describing charging amount, fuse the charging habit knowledge graph and the user quantity knowledge graph to generate a fused graph.

Images