An edge-computing-based power internet of things terminal device energy efficiency scheduling method

Abstract

The application discloses a kind of based on edge computing's electric power internet of things terminal equipment energy efficiency scheduling method, it is related to electric power internet of things technical field.The method first acquires distributed power generation data in table area, power grid operation data, terminal equipment operation data and edge node total energy consumption data, generates four-dimensional energy efficiency state data set;Then according to the energy efficiency mutual influence relationship between each dimension data constructs dynamic energy efficiency coupling matrix;Finally, based on dynamic energy efficiency coupling matrix, in energy efficiency, reliability and time delay three-dimensional multi-objective optimization space, collaborative optimization is carried out, and table area comprehensive energy efficiency scheduling scheme is generated.The application can accurately depict table area energy efficiency dynamic correlation characteristics, realize scheduling decision and real-time working condition Deep adaptation, effectively improve the overall energy utilization efficiency of table area, while guaranteeing power supply reliability and scheduling response speed, realize multi-objective collaborative optimization.

Description

Technical Field

[0001] This invention relates to the field of power Internet of Things (IoT) technology, and in particular to a power IoT terminal device energy efficiency scheduling method based on edge computing. Background Technology

[0002] The current power Internet of Things (IoT) for energy efficiency dispatching in distribution areas generally adopts a centralized data processing and decision-making model. Operational data from various power generation, consumption, and grid equipment within the distribution area are uniformly aggregated and centrally analyzed in a remote data center. The data center then generates global dispatch instructions and distributes them level by level to the distribution area terminals for execution. With the continuous expansion of distributed power source access in distribution areas and the rapid growth in the types and number of terminal electrical devices, the energy flow within the distribution areas exhibits characteristics of high-frequency fluctuations and complex interactions among multiple nodes. The scale of operational data that needs to be collected and processed is increasing exponentially.

[0003] The cross-network transmission and remote centralized processing of large-scale data will generate significant delays in scheduling decision response. These delays will be amplified as the frequency of fluctuations in the energy status of the distribution area increases, resulting in a continuous deviation between the issued scheduling instructions and the real-time energy supply and demand status of the distribution area. This makes it impossible to accurately adapt to the dynamically changing operating conditions of the distribution area, ultimately causing the overall energy utilization efficiency of the distribution area to remain at a low level for a long time and difficult to improve steadily. Summary of the Invention

[0004] This invention provides an energy efficiency scheduling method for power Internet of Things (IoT) terminal devices based on edge computing to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an energy efficiency scheduling method for power Internet of Things (IoT) terminal devices based on edge computing, comprising:

[0006] S1. Collect power generation data of distributed power sources, grid operation data, terminal equipment operation data and total energy consumption data of edge nodes within the distribution area to generate a four-dimensional energy efficiency status dataset;

[0007] S2. Construct a dynamic energy efficiency coupling matrix based on the energy efficiency interrelationships among the four-dimensional energy efficiency state datasets;

[0008] S3. Based on the dynamic energy efficiency coupling matrix, collaborative optimization is performed in a multi-objective optimization space including energy efficiency, reliability and time delay to generate a comprehensive energy efficiency scheduling scheme for the transformer area.

[0009] Preferably, the method for determining the energy efficiency interaction relationship includes:

[0010] Time synchronization processing is performed on the data of each dimension in the four-dimensional energy efficiency state dataset to obtain a time-aligned energy efficiency state sequence;

[0011] The time-aligned energy efficiency state sequence is normalized to obtain a standardized energy efficiency state sequence.

[0012] Calculate the nonlinear correlation coefficients between the data of each dimension in the standardized energy efficiency state sequence to obtain the energy efficiency mutual influence coefficient matrix;

[0013] Based on the energy efficiency mutual influence coefficient matrix, the energy efficiency mutual influence relationship between each dimension is determined.

[0014] Preferably, the calculation of the correlation coefficients between the data of each dimension in the standardized energy efficiency state sequence to obtain the energy efficiency mutual influence coefficient matrix includes:

[0015] Multi-scale wavelet decomposition was performed on the standardized energy efficiency state sequence to obtain multiple energy efficiency component sequences at different time scales.

[0016] Calculate the transfer entropy between the energy efficiency component sequences of each dimension at each time scale to obtain the directed influence intensity matrix of the corresponding time scale;

[0017] Calculate the time delay cross-correlation coefficients between the energy efficiency component sequences of each dimension at each time scale to obtain the influence symbol matrix at the corresponding time scale;

[0018] Calculate the proportion of energy in the energy efficiency component sequence to the total energy at each time scale, and use it as the weight for that time scale;

[0019] After multiplying the directed influence intensity matrix and the influence sign matrix element by element at each time scale, the matrix is ​​weighted and summed to obtain the signed directed influence coefficient matrix of energy efficiency.

[0020] Preferably, determining the energy efficiency interrelationships between different dimensions based on the energy efficiency interrelationship coefficient matrix includes:

[0021] The elements in the signed directed influence coefficient matrix of energy efficiency are threshold-filtered, and the matrix elements that meet the preset correlation strength are retained to obtain the set of directed correlation elements after filtering.

[0022] Based on the filtered set of directed related elements, a directed related topology between dimensions is constructed to obtain the dimensional related topology structure;

[0023] Based on the dimensional correlation topology, strong correlation dimension combinations and weak correlation dimension combinations are extracted to obtain the set of dimensional energy efficiency influence levels.

[0024] Based on the sign of the amplitude and the direction of transmission of the signed directed influence coefficient matrix of energy efficiency, the positive driving relationship and the negative inhibition relationship between each dimension are marked to obtain the mutual influence relationship of energy efficiency.

[0025] Preferably, the step of constructing a dynamic energy efficiency coupling matrix based on the energy efficiency interrelationships among the four-dimensional energy efficiency state datasets includes:

[0026] Based on the direction of energy efficiency influence between the dimensions in the four-dimensional energy efficiency state dataset, determine the sign of the element at the corresponding position in the matrix;

[0027] Based on the intensity of energy efficiency influence between each dimension, the magnitude of the matrix elements is determined, and the sign and magnitude are combined to obtain the initial matrix elements with directional attributes;

[0028] The initial energy efficiency coupling matrix is ​​obtained by constructing the matrix rows and columns based on the dimensional order.

[0029] The initial energy efficiency coupling matrix is ​​weighted and fused with the real-time influence weighting factor to obtain the dynamic energy efficiency coupling matrix.

[0030] Preferably, the method for constructing the real-time influence weighting factor includes:

[0031] By performing connectivity statistics on the dimensional relationship topology, a set of connectivity degrees for influence transmission between dimensions is obtained.

[0032] Based on combinations of strongly correlated and weakly correlated dimensions, a priority sequence of dimensional influence is generated.

[0033] The amplitude of the positive driving relationship and the negative inhibition relationship is normalized to obtain the baseline value of the relationship strength;

[0034] Based on the set of influence transmission connectivity and the priority sequence of dimensional influence, the baseline value of relationship strength is weighted and corrected to obtain the real-time influence weight factor.

[0035] Preferably, the step of collaboratively optimizing and generating a comprehensive energy efficiency scheduling scheme for the power distribution area based on a dynamic energy efficiency coupling matrix within a three-dimensional multi-objective optimization space encompassing energy efficiency, reliability, and latency includes:

[0036] By performing dimensional correlation decoupling on the dynamic energy efficiency coupling matrix, multi-objective optimization input parameters are obtained;

[0037] Based on the dimensional correlation topology and real-time influence weighting factors, a multi-objective optimization constraint boundary is constructed.

[0038] Multi-objective optimization input parameters are mapped to constraint boundaries to form a calibration optimization space;

[0039] Perform non-dominated sorting within the calibrated optimization space to obtain a set of non-dominated candidate scheduling schemes;

[0040] Based on the positive driving relationship and the negative inhibition relationship, the set of non-dominated candidate scheduling schemes is screened to obtain the integrated energy efficiency scheduling scheme for the transformer area.

[0041] Preferably, the step of performing non-dominated sorting within the calibration optimization space to obtain a set of non-dominated candidate scheduling schemes includes:

[0042] The feasible candidate scheduling schemes in the calibration optimization space are compared pairwise to generate a set of dominance relationships among the candidate scheduling schemes.

[0043] Based on the dimension-related topology, the set of dominance relations is modified to generate a set of dominance relations under topological constraints.

[0044] Based on the set of dominance relationships under topological constraints, feasible candidate scheduling schemes are hierarchically divided to obtain a hierarchical candidate scheduling scheme set.

[0045] Hierarchical calibration of the hierarchical candidate scheduling scheme set is performed based on real-time impact weighting factors;

[0046] The causal consistency of the calibrated hierarchical candidate scheduling scheme set is checked, and the candidate scheduling schemes that violate the energy efficiency causal relationship are eliminated, resulting in the non-dominated candidate scheduling scheme set.

[0047] Preferably, the hierarchical calibration of the hierarchical candidate scheduling scheme set based on real-time impact weight factors includes:

[0048] The dimension priority of the real-time impact weight factors is extracted to obtain the dimension impact priority vector;

[0049] The dimensional contribution of each candidate scheduling scheme in the hierarchical candidate scheduling scheme set is calculated to obtain the dimensional contribution of each candidate scheduling scheme.

[0050] The dimensional contribution of each candidate scheduling scheme is weighted based on the dimensional influence priority vector to obtain the comprehensive calibration value of each candidate scheduling scheme.

[0051] Based on the comprehensive calibration value, the candidate scheduling schemes within the same level are sorted to obtain the calibrated hierarchical candidate scheduling scheme set.

[0052] Preferably, the step of filtering the set of non-dominated candidate scheduling schemes based on positive driving relationships and negative inhibition relationships to obtain a comprehensive energy efficiency scheduling scheme for the distribution area includes:

[0053] Based on the set of dimensional energy efficiency impact levels, the candidate scheduling schemes in the set of non-dominated candidate scheduling schemes are prioritized to obtain a priority sequence.

[0054] Candidate scheduling schemes are selected in descending order of priority. The currently selected candidate scheduling scheme is mapped to the dimension-related topology to verify the causal transmission consistency of the scheme. If it passes, it is adopted as the integrated energy efficiency scheduling scheme for the distribution area. If it fails, the next priority candidate scheduling scheme is selected for verification until a scheme that passes the verification is selected.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. By constructing a dynamic energy efficiency coupling matrix and performing collaborative optimization in the three-dimensional multi-objective optimization space of energy efficiency, reliability, and time delay based on this matrix, the dynamic correlation characteristics between various energy efficiency states within the distribution area can be accurately captured. This enables deep adaptation between scheduling decisions and the real-time operating conditions of the distribution area, effectively reducing the deviation between scheduling instructions and actual energy supply and demand, significantly improving the overall energy utilization efficiency of the distribution area, and ensuring scheduling response speed and power supply stability.

[0057] 2. By decomposing energy efficiency components at multiple scales and accurately quantifying the intensity and direction of directional influences, combined with non-dominated sorting and causal consistency verification under dimensional correlation topological constraints, the strength levels and transmission paths of energy efficiency influences can be clearly delineated, further improving the rationality and accuracy of the multi-objective optimization process. This ensures that the generated scheduling scheme fully conforms to the inherent laws of energy efficiency transmission in the transformer area, achieving deep synergy and overall optimization of various optimization objectives. Attached Figure Description

[0058] Figure 1 A flowchart illustrating an energy efficiency scheduling method for power Internet of Things terminal devices based on edge computing, provided by the present invention;

[0059] Figure 2 A schematic diagram illustrating the process of constructing the effective influence coefficient matrix provided by this invention. Detailed Implementation

[0060] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0062] Reference Figures 1-2 As shown, in this embodiment, a power Internet of Things (IoT) terminal device energy efficiency scheduling method based on edge computing includes:

[0063] S1. Collect power generation data of distributed power sources, grid operation data, terminal equipment operation data and total energy consumption data of edge nodes within the distribution area to generate a four-dimensional energy efficiency status dataset.

[0064] In practice, data acquisition devices are installed at the output ports of distributed power sources within the distribution area, such as at the AC output of photovoltaic inverters and the grid connection interface of small wind turbines, to collect real-time power generation and electricity generation data. The acquisition interval can be adjusted according to the fluctuation characteristics of the corresponding energy source.

[0065] Data acquisition devices are installed on the incoming and outgoing sides of the transformer in the distribution area to collect voltage, current and power factor data of the power grid. The incoming side collects the power supply parameters of the power grid, and the outgoing side collects the total load parameters of the distribution area. If voltage fluctuations exceed the normal range, the start and end times of the fluctuations can be recorded simultaneously.

[0066] Data collection programs are deployed on various terminal devices within the distribution area, such as smart meters, home charging piles, and small industrial power terminals, to collect real-time power consumption, runtime, and start / stop status data. If a terminal device is in a shutdown state, its shutdown duration and the average power data of its last operation are recorded.

[0067] An energy consumption acquisition device is installed at the power input terminal of the edge node to collect the total energy consumption data of the edge node, which can be further divided into computing energy consumption, communication energy consumption and standby energy consumption. Computing energy consumption corresponds to the energy consumption when the edge node processes the collected data, and communication energy consumption corresponds to the energy consumption when the edge node transmits data with the terminal device.

[0068] The data collected in the above four categories are organized according to a unified time granularity. For example, all data are timestamped at the same time interval and stored in the dataset as four independent dimensions. Each dimension contains complete collected data of all devices in the corresponding category, and finally a four-dimensional energy efficiency status dataset is generated.

[0069] S2. Construct a dynamic energy efficiency coupling matrix based on the energy efficiency interaction relationships among the four-dimensional energy efficiency state datasets.

[0070] As one embodiment of the present invention, the method for determining the energy efficiency interaction relationship includes:

[0071] Time synchronization processing is performed on the data of each dimension in the four-dimensional energy efficiency state dataset to obtain a time-aligned energy efficiency state sequence;

[0072] The time-aligned energy efficiency state sequence is normalized to obtain a standardized energy efficiency state sequence.

[0073] Calculate the nonlinear correlation coefficients between the data of each dimension in the standardized energy efficiency state sequence to obtain the energy efficiency mutual influence coefficient matrix;

[0074] Based on the energy efficiency mutual influence coefficient matrix, the energy efficiency mutual influence relationship between each dimension is determined.

[0075] As one embodiment of the present invention, the correlation coefficients between the data of each dimension in the standardized energy efficiency state sequence are calculated to obtain the energy efficiency mutual influence coefficient matrix, including:

[0076] Multi-scale wavelet decomposition was performed on the standardized energy efficiency state sequence to obtain multiple energy efficiency component sequences at different time scales.

[0077] Calculate the transfer entropy between the energy efficiency component sequences of each dimension at each time scale to obtain the directed influence intensity matrix of the corresponding time scale;

[0078] Calculate the time delay cross-correlation coefficients between the energy efficiency component sequences of each dimension at each time scale to obtain the influence symbol matrix at the corresponding time scale;

[0079] Calculate the proportion of energy in the energy efficiency component sequence to the total energy at each time scale, and use it as the weight for that time scale;

[0080] After multiplying the directed influence intensity matrix and the influence sign matrix element by element at each time scale, the matrix is ​​weighted and summed to obtain the signed directed influence coefficient matrix of energy efficiency.

[0081] As one embodiment of the present invention, the energy efficiency mutual influence relationship between each dimension is determined based on the energy efficiency mutual influence coefficient matrix, including:

[0082] The elements in the signed directed influence coefficient matrix of energy efficiency are threshold-filtered, and the matrix elements that meet the preset correlation strength are retained to obtain the set of directed correlation elements after filtering.

[0083] Based on the filtered set of directed related elements, a directed related topology between dimensions is constructed to obtain the dimensional related topology structure;

[0084] Based on the dimensional correlation topology, strong correlation dimension combinations and weak correlation dimension combinations are extracted to obtain the set of dimensional energy efficiency influence levels.

[0085] Based on the sign of the amplitude and the direction of transmission of the signed directed influence coefficient matrix of energy efficiency, the positive driving relationship and the negative inhibition relationship between each dimension are marked to obtain the mutual influence relationship of energy efficiency.

[0086] As one embodiment of the present invention, a dynamic energy efficiency coupling matrix is ​​constructed based on the energy efficiency interrelationships among four-dimensional energy efficiency state datasets, including:

[0087] Based on the direction of energy efficiency influence between the dimensions in the four-dimensional energy efficiency state dataset, determine the sign of the element at the corresponding position in the matrix;

[0088] Based on the intensity of energy efficiency influence between each dimension, the magnitude of the matrix elements is determined, and the sign and magnitude are combined to obtain the initial matrix elements with directional attributes;

[0089] The initial energy efficiency coupling matrix is ​​obtained by constructing the matrix rows and columns based on the dimensional order.

[0090] The initial energy efficiency coupling matrix is ​​weighted and fused with the real-time influence weighting factor to obtain the dynamic energy efficiency coupling matrix.

[0091] As one embodiment of the present invention, the method for constructing weighting factors in real time includes:

[0092] By performing connectivity statistics on the dimensional relationship topology, a set of connectivity degrees for influence transmission between dimensions is obtained.

[0093] Based on combinations of strongly correlated and weakly correlated dimensions, a priority sequence of dimensional influence is generated.

[0094] The amplitude of the positive driving relationship and the negative inhibition relationship is normalized to obtain the baseline value of the relationship strength;

[0095] Based on the set of influence transmission connectivity and the priority sequence of dimensional influence, the baseline value of relationship strength is weighted and corrected to obtain the real-time influence weight factor.

[0096] In practice, based on the four-dimensional energy efficiency status dataset generated by the aforementioned steps, the discrete data points collected by different devices in each dimension are aligned according to timestamps, using a unified time base as a reference. If a device does not collect data at a certain moment, the intermediate value of the data collected by the device at two adjacent moments before and after is used to complete the data. After completion, the data in each dimension contains the same number of time points, and the times of all time points are completely consistent, thus obtaining a time-aligned energy efficiency status sequence.

[0097] Then, based on the newly generated time-aligned energy efficiency state sequence, the maximum and minimum values ​​in the energy efficiency state sequence of each dimension are extracted. The minimum value of that dimension is subtracted from each data point in the sequence, and then divided by the difference between the maximum and minimum values ​​of that dimension. In this way, the data of all dimensions are mapped to the same numerical range, while eliminating the influence of the difference in the scale of data in different dimensions, and finally obtaining the standardized energy efficiency state sequence.

[0098] Furthermore, based on the newly generated standardized energy efficiency state sequence, it is subjected to multi-scale wavelet decomposition, which can decompose it into three different time scales, corresponding to short-term fluctuations at the minute level, medium-term fluctuations at the hour level, and long-term fluctuations at the day level. During the decomposition process, the original data trend and fluctuation characteristics at each time scale are preserved. Each dimension of the standardized energy efficiency state sequence will be decomposed into a corresponding number of energy efficiency component sequences, and finally, multiple energy efficiency component sequences at different time scales are obtained.

[0099] Then, based on the energy efficiency component sequences obtained at multiple different time scales, for each time scale, for any two-dimensional energy efficiency component sequences, first, the probability of the occurrence of the subsequent value state of the other sequence under different value states of one sequence is calculated, and then the probability of the occurrence of the subsequent value state of the other sequence itself under different value states is calculated. The difference between these two probabilities is accumulated to obtain the directed influence strength between the two dimensions. After calculating the directed influence strength between all pairs of dimensions in turn, they are arranged in matrix form according to the dimensional order to obtain the directed influence strength matrix of the corresponding time scale.

[0100] Meanwhile, based on the directional influence intensity matrix obtained for each time scale, for any two energy efficiency component sequences at the same time scale, one of the sequences is shifted backward by different time steps, and the correlation between the two sequences after the shift is calculated.

[0101] The calculation first calculates the average value of each of the two sequences. Then, for each time point, the difference between the value of the first sequence at that time point and its average value is calculated, and the difference between the value of the second sequence at the corresponding time point after the shift and its average value is calculated. The two differences are multiplied together and the product of all time points is accumulated. Finally, the accumulated result is divided by the square root of the product of the squares of the differences between the two sequences to obtain the correlation value at that shift step size.

[0102] Then find the translation direction and value corresponding to the maximum correlation. If the translation direction is positive and the correlation value is positive, mark it as positive. If the translation direction is positive and the correlation value is negative, mark it as negative. After completing the sign marking between all pairs of dimensions, arrange them into a matrix according to the dimensional order to obtain the influence sign matrix of the corresponding time scale.

[0103] Then, based on the directed influence intensity matrix and influence symbol matrix of all time scales obtained earlier, the total energy of all dimensional energy efficiency component sequences at each time scale is calculated first, and then the total energy of all time scales is calculated. The total energy of each time scale is divided by the total energy of all time scales to obtain the energy proportion of that time scale. This proportion is used as the weight of that time scale. The sum of the weights of all time scales equals the total weight of the whole.

[0104] Further, based on the directional influence intensity matrix, influence sign matrix, and corresponding weights obtained for each time scale, for each time scale, the value at each position in the directional influence intensity matrix is ​​multiplied by the sign at the corresponding position in the influence sign matrix to obtain the signed influence intensity matrix for that time scale. Then, the signed influence intensity matrix of each time scale is multiplied by the weight of the corresponding time scale. Finally, the values ​​at corresponding positions in the weighted matrices of all time scales are added together to obtain the signed energy efficiency directional influence coefficient matrix.

[0105] Then, based on the newly generated signed directed energy efficiency influence coefficient matrix, the preset correlation strength can be determined according to the statistical results of historical transformer area operation data. The effective correlation strength range between each dimension in the historical data is statistically analyzed, and the lower limit of this range is taken as the screening threshold. If the absolute value of an element in the signed directed energy efficiency influence coefficient matrix is ​​greater than or equal to the screening threshold, the element is retained; if it is less than the screening threshold, the element is removed. Finally, the set of directed correlation elements after screening is obtained.

[0106] Meanwhile, based on the newly obtained set of directed related elements, the four dimensions of the four-dimensional energy efficiency state dataset are used as nodes of the topology. Each element in the set of directed related elements corresponds to a directed edge between nodes. The direction of the directed edge is consistent with the direction of influence of the element, and the length of the directed edge is proportional to the absolute value of the element. Finally, a dimensional related topology containing four nodes and several directed edges is formed.

[0107] Then, based on the newly generated dimensional association topology, the number and total strength of directed edges connected to each node are counted. Node combinations with higher total strength of directed edges are classified as strongly associated dimensional combinations, and node combinations with lower total strength of directed edges are classified as weakly associated dimensional combinations. All dimensional combinations are then sorted from high to low according to their association strength to form different influence levels, and finally, a set of dimensional energy efficiency influence levels is obtained.

[0108] Furthermore, based on the newly obtained signed directed energy efficiency influence coefficient matrix and dimensional correlation topology, if the sign of an element in the signed directed energy efficiency influence coefficient matrix is ​​positive, then the two corresponding dimensions are marked as having a positive driving relationship, that is, an increase in the value of one dimension will lead to an increase in the value of the other dimension. If the sign is negative, then the two corresponding dimensions are marked as having a negative inhibiting relationship, that is, an increase in the value of one dimension will lead to a decrease in the value of the other dimension. Finally, the complete energy efficiency mutual influence relationship is obtained.

[0109] Then, based on the newly determined complete energy efficiency interaction relationship, the rows of the matrix correspond to the initiating dimension of the influence, and the columns correspond to the receiving dimension of the influence. If the direction of the influence between two dimensions is from the row dimension to the column dimension, then the element symbol at that row and column position in the matrix is ​​consistent with the symbol of the corresponding relationship in the energy efficiency interaction relationship. The element symbols at all positions in the matrix are determined in turn.

[0110] Simultaneously, based on the element signs of all positions in the newly determined matrix, the absolute value of the corresponding element in the signed energy efficiency directional influence coefficient matrix is ​​used as the amplitude of the initial matrix element. Then, the previously determined sign is combined with this amplitude to obtain the initial matrix element with directional attributes. Each initial matrix element contains both the direction and intensity information of the influence.

[0111] Then, based on all the initial matrix elements with directional attributes that were just obtained, the rows and columns of the matrix are arranged in the order of distributed power generation data, grid operation data, terminal equipment operation data, and edge node total energy consumption data. All the initial matrix elements with directional attributes are filled into the corresponding positions in the matrix, and finally the initial energy efficiency coupling matrix is ​​obtained.

[0112] Furthermore, based on the newly generated dimensional association topology, the shortest path length between any two nodes is calculated. The shorter the path length, the higher the connectivity. The connectivity between all pairs of nodes is calculated in turn, and all connectivity values ​​are arranged in the order of dimension pairs to obtain the set of inter-dimensional influence transmission connectivity.

[0113] Then, based on the newly obtained set of dimensional energy efficiency impact levels, dimensions in strongly correlated dimensional combinations are assigned higher priority, and dimensions in weakly correlated dimensional combinations are assigned lower priority. Then, by combining the number of times each dimension appears in all correlated combinations and the total correlation strength, the priority of all dimensions is sorted, and finally, a dimensional impact priority sequence is generated.

[0114] Simultaneously, based on the newly determined complete energy efficiency interaction relationship, the amplitudes of all positive driving relationships and negative inhibiting relationships are extracted, the maximum amplitude is found, and the amplitude of each relationship is divided by this maximum amplitude to obtain the normalized relationship strength value. These values ​​are used as the baseline values ​​for relationship strength.

[0115] Then, based on the newly obtained set of inter-dimensional influence transmission connectivity, dimensional influence priority sequence, and relationship strength benchmark, the real-time influence weight factor is calculated using the following formula:

[0116]

[0117] In the formula, Indicates at time From a dimension To dimension The real-time impact weighting factor, whose value changes dynamically over time, is used to reflect the differences in the intensity of impact between dimensions at different times;

[0118] It is obtained from the set of influence transmission connectivity, representing the dimension. With dimension The influence between them is transmitted through connectivity;

[0119] It is obtained from the dimensional impact priority sequence, representing the dimension that initiated the impact. Priority coefficient;

[0120] and It is a preset weighting coefficient used to balance the weight of the connectivity priority term and the influence intensity normalization term, and the sum of the two is 1;

[0121] It is obtained from the relation strength benchmark value, representing the dimension. To dimension The baseline value for the strength of the relationship;

[0122] It is an influencing dimension The total of all other dimensions is used to normalize the baseline value of relation strength;

[0123] It is a time-varying adjustment function, the value of which is based on the dimension. At any moment The deviation between the real-time status data and the historical average status is dynamically determined to reflect the specific impact requirements at the current moment.

[0124] The calculation first calculates the connectivity priority term, then multiplies the influence propagation connectivity of the dimension pair by the priority coefficient of the initiating dimension, and finally multiplies it by a preset weighting coefficient. Then, the influence strength normalization term is calculated by dividing the baseline value of the relation strength corresponding to the dimension pair by the sum of the baseline values ​​of the relation strengths pointing to the receiving dimension, multiplying by the time-varying adjustment function, and finally multiplying by the preset weighting coefficient. The two calculation results are added together to obtain the corrected weight values. All the corrected weight values ​​are arranged in dimensional order to obtain the real-time impact weight factors.

[0125] Finally, based on the newly generated initial energy efficiency coupling matrix and the real-time influence weight factor, the element at each position in the initial energy efficiency coupling matrix is ​​multiplied by the value at the corresponding position in the real-time influence weight factor to obtain the fused matrix elements. All the fused matrix elements are arranged in the original row and column order to finally obtain the dynamic energy efficiency coupling matrix.

[0126] In summary, this embodiment extracts signed directional influence coefficients by combining multi-scale decomposition with transfer entropy and time-delay cross-correlation analysis. Based on the correlation strength, it constructs dimensional correlation topology and divides energy efficiency influence levels. Then, it integrates dimensional connectivity, influence priority, and time-varying adjustment factors to generate real-time weights, which are weighted with the initial matrix to obtain a dynamic energy efficiency coupling matrix. This approach can accurately characterize the causal influence strength and positive and negative effects between dimensions at different time scales and clearly distinguish between strong and weak correlation dimension combinations.

[0127] Meanwhile, this solution enables dynamic real-time updates of energy efficiency coupling relationships, providing a precise and reliable foundation for subsequent multi-objective collaborative optimization. This effectively enhances the scientific nature and adaptability of the integrated energy efficiency scheduling scheme for power distribution areas, and ensures the collaborative optimization effect across the three dimensions of energy efficiency, reliability, and latency.

[0128] S3. Based on the dynamic energy efficiency coupling matrix, collaborative optimization is performed in a multi-objective optimization space including energy efficiency, reliability and time delay to generate a comprehensive energy efficiency scheduling scheme for the transformer area.

[0129] As one embodiment of the present invention, based on a dynamic energy efficiency coupling matrix, a comprehensive energy efficiency scheduling scheme for power distribution areas is generated through collaborative optimization within a three-dimensional multi-objective optimization space encompassing energy efficiency, reliability, and latency. This includes:

[0130] By performing dimensional correlation decoupling on the dynamic energy efficiency coupling matrix, multi-objective optimization input parameters are obtained;

[0131] Based on the dimensional correlation topology and real-time influence weighting factors, a multi-objective optimization constraint boundary is constructed.

[0132] Multi-objective optimization input parameters are mapped to constraint boundaries to form a calibration optimization space;

[0133] Perform non-dominated sorting within the calibrated optimization space to obtain a set of non-dominated candidate scheduling schemes;

[0134] Based on the positive driving relationship and the negative inhibition relationship, the set of non-dominated candidate scheduling schemes is screened to obtain the integrated energy efficiency scheduling scheme for the transformer area.

[0135] As one embodiment of the present invention, non-dominated sorting is performed within a calibrated optimization space to obtain a set of non-dominated candidate scheduling schemes, including:

[0136] The feasible candidate scheduling schemes in the calibration optimization space are compared pairwise to generate a set of dominance relationships among the candidate scheduling schemes.

[0137] Based on the dimension-related topology, the set of dominance relations is modified to generate a set of dominance relations under topological constraints.

[0138] Based on the set of dominance relationships under topological constraints, feasible candidate scheduling schemes are hierarchically divided to obtain a hierarchical candidate scheduling scheme set.

[0139] Hierarchical calibration of the hierarchical candidate scheduling scheme set is performed based on real-time impact weighting factors;

[0140] The causal consistency of the calibrated hierarchical candidate scheduling scheme set is checked, and the candidate scheduling schemes that violate the energy efficiency causal relationship are eliminated, resulting in the non-dominated candidate scheduling scheme set.

[0141] As one embodiment of the present invention, hierarchical calibration of the hierarchical candidate scheduling scheme set based on real-time influence weighting factors includes:

[0142] The dimension priority of the real-time impact weight factors is extracted to obtain the dimension impact priority vector;

[0143] The dimensional contribution of each candidate scheduling scheme in the hierarchical candidate scheduling scheme set is calculated to obtain the dimensional contribution of each candidate scheduling scheme.

[0144] The dimensional contribution of each candidate scheduling scheme is weighted based on the dimensional influence priority vector to obtain the comprehensive calibration value of each candidate scheduling scheme.

[0145] Based on the comprehensive calibration value, the candidate scheduling schemes within the same level are sorted to obtain the calibrated hierarchical candidate scheduling scheme set.

[0146] As one embodiment of the present invention, the set of non-dominated candidate scheduling schemes is screened based on positive driving relationships and negative inhibition relationships to obtain a comprehensive energy efficiency scheduling scheme for the distribution area, including:

[0147] Based on the set of dimensional energy efficiency impact levels, the candidate scheduling schemes in the set of non-dominated candidate scheduling schemes are prioritized to obtain a priority sequence.

[0148] Candidate scheduling schemes are selected in descending order of priority. The currently selected candidate scheduling scheme is mapped to the dimension-related topology to verify the causal transmission consistency of the scheme. If it passes, it is adopted as the integrated energy efficiency scheduling scheme for the distribution area. If it fails, the next priority candidate scheduling scheme is selected for verification until a scheme that passes the verification is selected.

[0149] In practice, based on the dynamic energy efficiency coupling matrix generated in the aforementioned steps, dimensional correlation decoupling is performed. During decoupling, according to the strongly correlated and weakly correlated dimensional combinations in the dimensional correlation topology, the mutually coupled dimensions are split into independent input items. The strongly correlated dimensional combinations are used as the overall input items, and the weakly correlated dimensions are used as individual input items. The split input items are then mapped to the three optimization objectives of energy efficiency, reliability, and time delay, respectively, and finally, multi-objective optimization input parameters are obtained.

[0150] Then, based on the aforementioned generated dimensional correlation topology and real-time influence weight factors, a multi-objective optimization constraint boundary is constructed. During the construction, the basic value range of each optimization objective is first determined according to the influence transmission path in the dimensional correlation topology. Then, the tightness of the constraint is adjusted in combination with the real-time influence weight factors. For example, for dimensions with higher real-time influence weights, the corresponding constraint boundary is set more strictly, while for dimensions with lower real-time influence weights, the corresponding constraint boundary is set relatively loosely, ultimately forming a complete constraint boundary covering the three optimization objectives.

[0151] Furthermore, the obtained multi-objective optimization input parameters are mapped to the above-mentioned constraint boundaries. During the mapping, the original value range of each input parameter is mapped to the effective interval within the constraint boundary. If the value of an input parameter exceeds the constraint boundary, it is adjusted to the limit value at the boundary, and the relative proportional relationship between the input parameters is preserved to ensure that the mapped parameters can accurately reflect the characteristics of the original data, and finally form a calibration optimization space.

[0152] Then, all feasible candidate scheduling schemes are generated within the calibration optimization space. These feasible candidate scheduling schemes are compared pairwise. During the comparison, the performance of the two schemes in three dimensions, namely energy efficiency, reliability and time delay, is evaluated. If a scheme performs no worse than the other scheme in all dimensions and performs better in at least one dimension, then the former is determined to dominate the latter. After completing the pairwise comparison of all candidate scheduling schemes in sequence, all dominance relationships are sorted out, and finally a set of dominance relationships between candidate scheduling schemes is generated.

[0153] Simultaneously, based on the aforementioned generated dimensional association topology, the set of dominance relations is modified. During the modification, it is checked whether each dominance relation conforms to the influence transmission logic in the dimensional association topology. If a dominance relation violates the directed influence path in the topology, the dominance relation is removed. If a dominance relation conforms to the influence transmission logic in the topology, the dominance relation is retained, and finally, a set of dominance relations under topological constraints is generated.

[0154] Then, based on the set of dominance relationships under topological constraints, feasible candidate scheduling schemes are hierarchically divided. During the division, candidate scheduling schemes that are not dominated by any other schemes are divided into the first level, candidate scheduling schemes that are dominated only by the first level schemes are divided into the second level, and so on, until all candidate scheduling schemes are divided into the corresponding levels, and finally a hierarchical candidate scheduling scheme set is obtained.

[0155] Furthermore, based on the aforementioned real-time impact weight factors, the hierarchical candidate scheduling scheme set is calibrated. First, the dimension priority of the real-time impact weight factors is extracted, and the three dimensions of energy efficiency, reliability and latency are sorted according to the weight to obtain the dimension impact priority vector.

[0156] Then, the dimensional contribution of each candidate scheduling scheme in the hierarchical candidate scheduling scheme set is calculated. During the calculation, the degree of improvement of the scheme relative to the baseline state in each dimension is counted, and the degree of improvement in each dimension is normalized to obtain the dimensional contribution of the candidate scheduling scheme in each dimension.

[0157] Meanwhile, the dimensional contribution of each candidate scheduling scheme is weighted based on the dimensional influence priority vector. The contribution of each dimension is multiplied by the priority coefficient corresponding to that dimension, and the weighted results of the three dimensions are added together to obtain the comprehensive calibration value of each candidate scheduling scheme.

[0158] Then, the candidate scheduling schemes within the same level are sorted according to the comprehensive calibration value. The candidate scheduling scheme with the higher the comprehensive calibration value is ranked higher. After sorting all levels in turn, the final set of calibrated hierarchical candidate scheduling schemes is obtained.

[0159] Further, the causal consistency of the calibrated hierarchical candidate scheduling scheme set is checked. During the check, it is checked whether the execution logic of each candidate scheduling scheme conforms to the causal transmission relationship in the dimensional association topology. If a candidate scheduling scheme has a logical contradiction that the result precedes the cause, or violates the positive driving relationship and negative inhibition relationship, the candidate scheduling scheme is removed. All candidate scheduling schemes that pass the check are retained, and finally the set of non-dominated candidate scheduling schemes is obtained.

[0160] Then, based on the aforementioned positive driving relationship and negative inhibition relationship, the set of non-dominated candidate scheduling schemes is screened. First, according to the previously generated set of dimensional energy efficiency influence levels, the candidate scheduling schemes in the set of non-dominated candidate scheduling schemes are prioritized. Schemes that can make full use of the positive driving relationship of strong correlation dimensions are given priority, while schemes that can effectively avoid the negative inhibition relationship of weak correlation dimensions are given priority, and finally, a priority sequence is obtained.

[0161] Then, candidate scheduling schemes are selected in descending order of priority. The currently selected candidate scheduling scheme is mapped to the dimension-related topology to verify the causal transmission consistency of the scheme. During the verification, it is checked whether the influence of each adjustment action in the scheme can be correctly transmitted along the directed edges in the topology and whether the results after transmission conform to the definition of positive driving relationship and negative inhibition relationship.

[0162] If the candidate scheduling scheme passes the causal transmission consistency verification, it will be used as the integrated energy efficiency scheduling scheme for the distribution area. If it fails the verification, the next priority candidate scheduling scheme will be selected for verification until a scheme that passes the verification is selected, and the integrated energy efficiency scheduling scheme for the distribution area will be generated.

[0163] In summary, this scheme decouples dimensional relationships based on a dynamic energy efficiency coupling matrix and constructs a calibration optimization space that integrates topological constraints and real-time weights. It combines non-dominated sorting with dimensional priority calibration and causal consistency verification, and then selects the final scheduling scheme based on the energy efficiency impact level and the positive and negative relationships between dimensions. This effectively compresses the ineffective solution space, avoids infeasible schemes that violate the causal logic of energy efficiency from the root, significantly improves the rationality and accuracy of non-dominated sorting, and ensures that the selected scheduling scheme meets the requirements of multi-objective collaborative optimization.

[0164] Furthermore, the settings in this embodiment fully conform to the actual energy efficiency transmission law of the distribution area, greatly narrowing the gap between theoretical optimization and actual operation effect, achieving the optimal balance between distribution area energy efficiency improvement, power supply reliability guarantee and dispatch delay control, and comprehensively enhancing the practicality and adaptability of the integrated energy efficiency dispatch scheme.

[0165] In the several embodiments provided by this invention, it should be understood that the disclosed method can be implemented in other ways.

[0166] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0167] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, and technology that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A power Internet of Things (IoT) terminal device energy efficiency scheduling method based on edge computing, characterized in that, The method includes: S1. Collect power generation data of distributed power sources, grid operation data, terminal equipment operation data and total energy consumption data of edge nodes within the distribution area to generate a four-dimensional energy efficiency status dataset; S2. Construct a dynamic energy efficiency coupling matrix based on the energy efficiency interrelationships among the four-dimensional energy efficiency state datasets; S3. Based on the dynamic energy efficiency coupling matrix, collaborative optimization is performed in a multi-objective optimization space including energy efficiency, reliability and time delay to generate a comprehensive energy efficiency scheduling scheme for the transformer area.

2. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 1, characterized in that, The method for determining the energy efficiency interaction relationship includes: Time synchronization processing is performed on the data of each dimension in the four-dimensional energy efficiency state dataset to obtain a time-aligned energy efficiency state sequence; The time-aligned energy efficiency state sequence is normalized to obtain a standardized energy efficiency state sequence. Calculate the nonlinear correlation coefficients between the data of each dimension in the standardized energy efficiency state sequence to obtain the energy efficiency mutual influence coefficient matrix; Based on the energy efficiency mutual influence coefficient matrix, the energy efficiency mutual influence relationship between each dimension is determined.

3. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 1, characterized in that, The correlation coefficients between the data of each dimension in the standardized energy efficiency state sequence are calculated to obtain the energy efficiency mutual influence coefficient matrix, including: Multi-scale wavelet decomposition was performed on the standardized energy efficiency state sequence to obtain multiple energy efficiency component sequences at different time scales. Calculate the transfer entropy between the energy efficiency component sequences of each dimension at each time scale to obtain the directed influence intensity matrix of the corresponding time scale; Calculate the time delay cross-correlation coefficients between the energy efficiency component sequences of each dimension at each time scale to obtain the influence symbol matrix at the corresponding time scale; Calculate the proportion of energy in the energy efficiency component sequence to the total energy at each time scale, and use it as the weight for that time scale; After multiplying the directed influence intensity matrix and the influence sign matrix element by element at each time scale, the matrix is ​​weighted and summed to obtain the signed directed influence coefficient matrix of energy efficiency.

4. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 3, characterized in that, The determination of the energy efficiency interrelationships between different dimensions based on the energy efficiency interrelationship coefficient matrix includes: The elements in the signed directed influence coefficient matrix of energy efficiency are threshold-filtered, and the matrix elements that meet the preset correlation strength are retained to obtain the set of directed correlation elements after filtering. Based on the filtered set of directed related elements, a directed related topology between dimensions is constructed to obtain the dimensional related topology structure; Based on the dimensional correlation topology, strong correlation dimension combinations and weak correlation dimension combinations are extracted to obtain the set of dimensional energy efficiency influence levels. Based on the sign of the amplitude and the direction of transmission of the signed directed influence coefficient matrix of energy efficiency, the positive driving relationship and the negative inhibition relationship between each dimension are marked to obtain the mutual influence relationship of energy efficiency.

5. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 4, characterized in that, The construction of a dynamic energy efficiency coupling matrix based on the energy efficiency interrelationships among the four-dimensional energy efficiency state datasets includes: Based on the direction of energy efficiency influence between the dimensions in the four-dimensional energy efficiency state dataset, determine the sign of the element at the corresponding position in the matrix; Based on the intensity of energy efficiency influence between each dimension, the magnitude of the matrix elements is determined, and the sign and magnitude are combined to obtain the initial matrix elements with directional attributes; The initial energy efficiency coupling matrix is ​​obtained by constructing the matrix rows and columns based on the dimensional order. The initial energy efficiency coupling matrix is ​​weighted and fused with the real-time influence weighting factor to obtain the dynamic energy efficiency coupling matrix.

6. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 5, characterized in that, The method for constructing the real-time impact weighting factor includes: By performing connectivity statistics on the dimensional relationship topology, a set of connectivity degrees for influence transmission between dimensions is obtained. Based on combinations of strongly correlated and weakly correlated dimensions, a priority sequence of dimensional influence is generated. The amplitude of the positive driving relationship and the negative inhibition relationship is normalized to obtain the baseline value of the relationship strength; Based on the set of influence transmission connectivity and the priority sequence of dimensional influence, the baseline value of relationship strength is weighted and corrected to obtain the real-time influence weight factor.

7. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 6, characterized in that, The method for collaboratively optimizing and generating a comprehensive energy efficiency scheduling scheme for power distribution areas within a three-dimensional multi-objective optimization space based on a dynamic energy efficiency coupling matrix includes: By performing dimensional correlation decoupling on the dynamic energy efficiency coupling matrix, multi-objective optimization input parameters are obtained; Based on the dimensional correlation topology and real-time influence weighting factors, a multi-objective optimization constraint boundary is constructed. Multi-objective optimization input parameters are mapped to constraint boundaries to form a calibration optimization space; Perform non-dominated sorting within the calibrated optimization space to obtain a set of non-dominated candidate scheduling schemes; Based on the positive driving relationship and the negative inhibition relationship, the set of non-dominated candidate scheduling schemes is screened to obtain the integrated energy efficiency scheduling scheme for the transformer area.

8. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 7, characterized in that, The step of performing non-dominated sorting within the calibration optimization space to obtain a set of non-dominated candidate scheduling schemes includes: The feasible candidate scheduling schemes in the calibration optimization space are compared pairwise to generate a set of dominance relationships among the candidate scheduling schemes. Based on the dimension-related topology, the set of dominance relations is modified to generate a set of dominance relations under topological constraints. Based on the set of dominance relationships under topological constraints, feasible candidate scheduling schemes are hierarchically divided to obtain a hierarchical candidate scheduling scheme set. Hierarchical calibration of the hierarchical candidate scheduling scheme set is performed based on real-time impact weighting factors; The causal consistency of the calibrated hierarchical candidate scheduling scheme set is checked, and the candidate scheduling schemes that violate the energy efficiency causal relationship are eliminated, resulting in the non-dominated candidate scheduling scheme set.

9. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 8, characterized in that, The hierarchical calibration of the hierarchical candidate scheduling scheme set based on real-time impact weight factors includes: The dimension priority of the real-time impact weight factors is extracted to obtain the dimension impact priority vector; The dimensional contribution of each candidate scheduling scheme in the hierarchical candidate scheduling scheme set is calculated to obtain the dimensional contribution of each candidate scheduling scheme. The dimensional contribution of each candidate scheduling scheme is weighted based on the dimensional influence priority vector to obtain the comprehensive calibration value of each candidate scheduling scheme. Based on the comprehensive calibration value, the candidate scheduling schemes within the same level are sorted to obtain the calibrated hierarchical candidate scheduling scheme set.

10. The energy efficiency scheduling method for power Internet of Things terminal equipment based on edge computing as described in claim 7, characterized in that, The process of filtering the set of non-dominated candidate scheduling schemes based on positive driving and negative inhibiting relationships yields a comprehensive energy efficiency scheduling scheme for the distribution area, including: Based on the set of dimensional energy efficiency impact levels, the candidate scheduling schemes in the set of non-dominated candidate scheduling schemes are prioritized to obtain a priority sequence. Candidate scheduling schemes are selected in descending order of priority. The currently selected candidate scheduling scheme is mapped to the dimension-related topology to verify the causal transmission consistency of the scheme. If it passes, it is adopted as the integrated energy efficiency scheduling scheme for the distribution area. If it fails, the next priority candidate scheduling scheme is selected for verification until a scheme that passes the verification is selected.

Images