Energy structure optimization method and system for integrated source-grid-load-storage combination
By collecting and processing wind and light power generation data in the microgrid system, using wavelet noise reduction and nonlinear segmentation surface division and other technologies, combined with neural network and group intelligence algorithms, anomaly detection and prediction are optimized, and the abnormal detection deviation rate in traditional systems is solved. High problems improve prediction accuracy and system efficiency.
Patent Information
- Application Number
- CN202410996896.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-07-24
AI Technical Summary
When traditional microgrid systems face output fluctuations such as wind power and photovoltaics, the detection rate and error detection rate of abnormal detection algorithms are high, resulting in a decrease in the credibility and robustness of neural network model predictions, affecting the system operation economy and environment. benefit.
By collecting historical data of wind and light power generation and weather data, using wavelet noise reduction, nonlinear segmentation surface division, isolated value calculation and weighted fusion, combining random forests and missing forests to fill vacant positions, environmental variable mapping is performed, and neural networks and group intelligence algorithms are used to fill in the vacant positions. Optimize the objective function and generate an integrated energy structure optimization plan for source, network, load and storage.
Improve the accuracy and efficiency of outlier elimination, enhance the prediction credibility and robustness of neural network models, improve the long-term and short-term prediction accuracy of wind power and photovoltaic output data, reduce the economic cost of microgrid systems, and improve the Environmental benefits and system operation efficiency.
Smart Images

Figure CN118539436B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of power grid operation, and in particular to a method and system for optimizing an energy structure of an integrated source-grid-load-storage combination. Background Art
[0002] The source-grid-load-storage system is an important development path for building a new power system. It is a distribution system composed of power sources, power grids, loads, and energy storage. The "source follows the load" control method adopted by the traditional distribution network cannot dispatch distributed power sources with random output such as wind power and photovoltaic power, making it difficult to ensure the safe and economical operation of the distribution network.
[0003] At present, the energy dispatching methods of microgrids are mainly based on multiple time scales. This type of method mainly formulates dispatching plans based on the prediction of neural network models through data collection. The collected data will generate a large amount of abnormal data due to equipment failure, wind power, photovoltaic and other output fluctuations and management. However, the detection rate and false detection rate of outliers of traditional anomaly detection algorithms may show limitations when facing massive data. The isolation forest algorithm, which is non-parametric and unsupervised and can efficiently process large amounts of data, is now adopted. However, this algorithm will cause a large anomaly detection deviation rate when randomly selecting attributes for data space segmentation, and the selection of abnormal scores is inappropriate. The abnormal scores are far higher than other scores in numerical distribution, which weakens or even destroys the credibility and robustness of the prediction of the neural network model, and reduces the real-time, short-term or ultra-short-term prediction accuracy of wind power, photovoltaic and other output data, thereby increasing the economic operation of the microgrid system and reducing environmental benefits and system operation efficiency. Summary of the invention
[0004] The disclosed embodiment provides a method and system for optimizing the energy structure of an integrated source-grid-load-storage combination, which solves the problems of large anomaly detection deviation rate, inappropriate selection of anomaly scores, and anomaly scores far exceeding other scores in numerical distribution when randomly selecting attributes for data space segmentation. The accuracy and efficiency of outlier removal are improved, the credibility and robustness of predictions made by neural network models are improved, and the long-term, short-term or ultra-short-term prediction accuracy of output data such as wind power and photovoltaics is improved, thereby reducing the operating economy of microgrid systems, improving environmental benefits and system operating efficiency, and improving the optimization capability of the energy structure of an integrated source-grid-load-storage combination, providing guidance for the actual operation of microgrid systems with integrated source-grid-load-storage.
[0005] The present invention provides a source-grid-load-storage integrated combined energy structure optimization method, comprising: collecting historical wind-solar power generation output data, historical weather data and load historical data, and obtaining denoised data through wavelet denoising; dividing the denoised data through a nonlinear segmentation surface, and obtaining tree data and path length through a segmentation ratio; calculating isolated values and tree data isolation factors through path lengths, and obtaining isolated points of the tree data through weighted fusion; obtaining abnormal data in the tree data through isolation determination for the isolated points of the tree data, and then filling in the vacant positions of the tree data through random forests and missing forests to obtain processed historical wind-solar power generation output data, processed historical weather data and processed load historical data; the processed wind and solar power generation historical output data and the processed historical weather data are mapped with environmental variables and renewable energy output power to obtain renewable energy output power mapping values and the processed load historical data through a neural network algorithm to predict the future renewable energy output power forecast value and the future load power forecast value; the objective function for optimizing the energy structure of the integrated source, grid, load and storage combination is constructed to obtain the optimized objective function; the future renewable energy output power forecast value and the future load power forecast value are solved by the swarm intelligence algorithm to obtain the optimization plan for the energy structure of the integrated source, grid, load and storage combination.
[0006] In one embodiment, dividing the denoised data by a nonlinear segmentation plane and obtaining tree data and path length by a segmentation ratio comprises:
[0007] Randomly select one of the denoised data and segment it through the nonlinear segmentation surface to obtain sub-segmentation data and sub-tree data, where the nonlinear segmentation surface is
[0008]
[0009] …
[0010]
[0011] in, Dimension is a random number For the hypothetical random numbers between the minimum and maximum values of the dimension, is a positive integer, is the point on the nonlinear segmentation surface, that is, the data point in the tree data. For the Points on the dimensional nonlinear partitioning surface;
[0012] Repeat the above sub-segmentation data operation and calculate the length of the left and right sub-tree-shaped data paths to obtain the segmentation ratio and then obtain the length of the sub-tree-shaped data path;
[0013] Then randomly select a data in the denoised data as the segmentation point and repeat the above operation of obtaining the length of the subtree data path to obtain the tree data and the path length.
[0014] In one embodiment, the step of calculating the isolation value through the path length and the isolation factor of the tree data to obtain the isolation score of the tree data through weighted fusion includes:
[0015] The isolated value is calculated by the path length, where the isolated value for
[0016]
[0017] Tree Data Isolation Factor for
[0018]
[0019]
[0020] The isolated value and the tree data isolation factor are weighted fused to obtain the tree data isolation factor. , where isolated for
[0021]
[0022] in, is the size of the tree data is the path length of each data in the tree data on the tree data, is the average length of each data path in the tree data, is the harmonic number, is the data point in the tree data of Distance and data points in tree data To the data point in the tree data The maximum value in the interval, is the data point in the tree data The distance is less than dist The set of data points in the tree data, dist is the data point in the tree data The distances to other data points in the tree data are sorted from small to large. The distance between the data points is distance, is the fusion weight.
[0023] In one embodiment, the isolated analysis of the tree data obtains abnormal data in the tree data through isolation determination and then fills the vacant positions of the tree data through random forest and missing forest to obtain processed wind and solar power generation historical output data, processed historical weather data and processed load historical data, including:
[0024] The isolated data of the tree data is isolated and judged to get the abnormal data in the tree data, where the isolated judgment is
[0025]
[0026] After filtering out abnormal data in the tree data, random forest and missing forest are used to predict the missing positions of the tree data to obtain random forest prediction values and missing forest prediction values;
[0027] The filling reserve value is obtained by calculating the absolute value of the difference between the random forest prediction value and the missing forest prediction value;
[0028] Select the minimum value among the filling reserve values to fill the vacant position of the tree data to obtain the processed historical output data of wind and solar power generation, the processed historical weather data and the processed historical load data;
[0029] in, Isolated nodes for tree data The standard deviation of .
[0030] In one embodiment, the processing of the historical wind and solar power generation output data and the historical weather data to map the environmental variables with the renewable energy output power to obtain the renewable energy output power mapping value and the processed load historical data to obtain the future renewable energy output power prediction value and the future load power prediction value through a neural network algorithm, including:
[0031] Mapping the environmental variables and renewable energy output power of the processed historical wind and solar power generation data and the processed historical weather data to obtain the renewable energy output power mapping value;
[0032] The output power mapping values of renewable energy in different time periods every day of a week and the processed load history data are used together through the fuzzy brain emotional learning neural network to predict the output power forecast values of renewable energy and load power forecast values in different time periods every day of the next week;
[0033] The output power mapping values of renewable energy in different time periods of a day and the processed load history data are used together through the fuzzy brain emotional learning neural network to predict the output power forecast values of renewable energy and load power forecast values in different time periods of the next day;
[0034] The output power mapping value of renewable energy in one time period of a day is predicted by the recursive cerebellar model neural network to obtain the predicted output power value of renewable energy in one time period of the same day;
[0035] The processed load history data of the previous time period of the day is predicted by the recursive cerebellar model neural network to obtain the load power forecast value of the same time period of the same day.
[0036] In one embodiment, the objective function of optimizing the energy structure of the integrated source-grid-load-storage combination is constructed to obtain an optimized objective function, including:
[0037] Optimizing the objective function
[0038]
[0039] Optimize the constraints of the objective function
[0040]
[0041]
[0042]
[0043] in, is the target value of the scheduling plan, is the time period of the microgrid dispatch cycle, The economic objectives of each scheduling cycle are Environmental benefit goals Stability target of the tie line with the large power grid Stability goals within the system The weight coefficient of are the upper and lower limits of renewable energy output power, is the actual photovoltaic output power Actual wind power output is the actual output power of energy storage, The AC power between the microgrid and the large grid is the load power, is the minimum charging power for energy storage, is the maximum charging power of energy storage, is the energy storage power, Minimum state of charge for energy storage, Energy Storage The state of charge at the moment, Maximum state of charge for energy storage.
[0044] In one embodiment, the predicted output power of renewable energy and the predicted future load power are solved by a swarm intelligence algorithm to obtain an optimization plan for the source-grid-load-storage integrated energy structure, including:
[0045] The predicted values of renewable energy output power and load power at different time periods every day in the next week are solved by swarm intelligence algorithm to obtain the first energy structure optimization;
[0046] Based on the first energy structure optimization, the predicted values of renewable energy output power and load power in different time periods in the future day are solved by swarm intelligence algorithm to obtain the second energy structure optimization;
[0047] On the basis of the optimization of the secondary energy structure, the output power forecast value and load power forecast value of renewable energy in the same time period of the same day are solved by swarm intelligence algorithm to obtain the optimization plan of the integrated energy structure of source, grid, load and storage.
[0048] The present invention also provides an energy structure optimization system for an integrated source-grid-load-storage combination, including: a noise reduction unit, which collects historical output data of wind and solar power generation, historical weather data and historical load data, and obtains noise-reduced data through wavelet noise reduction; a data partitioning unit, which partitions the noise-reduced data through a nonlinear partitioning surface and obtains tree data and path length through a partitioning ratio; an isolated component calculation unit, which calculates isolated values and isolated factors of tree data through path length and obtains isolated components of tree data through weighted fusion; an isolated determination unit, which obtains abnormal data in the tree data through isolated determination for the isolated components of the tree data, and then fills in the vacant positions of the tree data through random forests and missing forests to obtain processed historical output data of wind and solar power generation, processed historical weather data and processed load historical data; a future power prediction unit, which maps the processed wind and solar power generation historical output data and the processed historical weather data with environmental variables and renewable energy output power to obtain the renewable energy output power mapping value and the processed load historical data through a neural network algorithm to obtain the future renewable energy output power prediction value and the future load power prediction value; an objective function construction unit, which constructs the objective function for optimizing the energy structure of the integrated source, grid, load and storage combination to obtain the optimized objective function; an energy structure optimization unit, which solves the optimization objective function for the future renewable energy output power prediction value and the future load power prediction value through a swarm intelligence algorithm to obtain the optimization plan for the energy structure of the integrated source, grid, load and storage combination.
[0049] The present invention also provides an electronic device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of any of the above-mentioned source-grid-load-storage integrated combined energy structure optimization methods.
[0050] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for optimizing the energy structure of the integrated source-grid-load-storage combination.
[0051] Beneficial effects of the present invention: The embodiment of the present invention provides a source-grid-load-storage integrated combined energy structure optimization method and system, which collects historical wind and solar power generation output data, historical weather data and load historical data through wavelet denoising to obtain denoised data; divides the denoised data through a nonlinear segmentation surface and obtains tree data and path length through the segmentation ratio; calculates isolated values through the path length and calculates the tree data isolation factor through weighted fusion to obtain isolated points of the tree data; obtains abnormal data in the tree data through isolation judgment for the isolated points of the tree data and screens them, and then fills the vacant positions of the tree data through random forest and missing forest to obtain the processed wind and solar power generation historical output. power data, processed historical weather data and processed historical load data; the processed historical wind and solar power generation data and processed historical weather data are mapped with environmental variables and renewable energy output power to obtain the renewable energy output power mapping value and the processed historical load data through the neural network algorithm to obtain the future renewable energy output power forecast value and future load power forecast value; the objective function of the source-grid-load-storage integrated energy structure optimization is constructed to obtain the optimized objective function; the future renewable energy output power forecast value and future load power forecast value are solved by the swarm intelligence algorithm to obtain the source-grid-load-storage integrated energy structure optimization plan. The problem of large anomaly detection deviation rate, inappropriate anomaly score selection, and anomaly score far exceeding other scores in numerical distribution when randomly selecting attributes for data space segmentation is solved. Improve the accuracy and efficiency of outlier removal, improve the credibility and robustness of neural network model predictions, and improve the long-term, short-term or ultra-short-term prediction accuracy of wind power, photovoltaic and other output data, thereby reducing the operating economy of microgrid systems, improving environmental benefits and system operating efficiency, and improving the optimization capabilities of the energy structure of integrated source, grid, load and storage combinations, providing guidance for the actual operation of integrated source, grid, load and storage microgrid systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 It is a flow chart of the source-grid-load-storage integrated combined energy structure optimization method provided by the present invention;
[0053] Figure 2 It is a structural schematic diagram of the source-grid-load-storage integrated combined energy structure optimization system provided by the present invention;
[0054] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0055] Combine the following Figure 1 Figure 3 The present invention describes the source-grid-load-storage integrated combined energy structure optimization method and system.
[0056] Specifically, the present invention provides a method for optimizing the energy structure of a source-grid-load-storage integrated combination, referring to Figure 1 , Figure 1 It is a flow chart of the source-grid-load-storage integrated combined energy structure optimization method provided by the present invention.
[0057] The method steps for optimizing the energy structure of the integrated source-grid-load-storage combination provided in the embodiment of the present invention include:
[0058] Step 100: collect historical wind and solar power generation output data, historical weather data and historical load data, and obtain denoised data through wavelet denoising.
[0059] Here, step 100 includes:
[0060] Collect historical wind and solar power generation output data, historical weather data and historical load data to decompose components of different scales to obtain high-frequency components and low-frequency components;
[0061] Threshold processing is performed on the high-frequency component and the low-frequency component by soft thresholding to obtain a high-frequency threshold component and a low-frequency threshold component;
[0062] The high-frequency threshold component and the low-frequency threshold component are combined with the historical output data of wind and solar power generation, historical weather data and historical load data to reconstruct and reduce noise through inverse wavelet transform to obtain the denoised data.
[0063] The coordinated optimization of source-grid-load-storage systems refers to the optimized scheduling and active management of various distributed resources at the source level. Through a variety of interactive means, the dynamic power balance capability of the power system can be improved more economically, efficiently and safely, thereby realizing the operating mode and technology of maximizing the utilization of energy resources.
[0064] Step 200, the denoised data is divided by a nonlinear segmentation plane and tree data is obtained by segmentation ratio and path length is obtained.
[0065] Here, step 200 includes:
[0066] Randomly select one of the denoised data and segment it through the nonlinear segmentation surface to obtain sub-segmentation data and sub-tree data, where the nonlinear segmentation surface is
[0067]
[0068] …
[0069]
[0070] in, Dimension A random number between 0 and 1 For the hypothetical random numbers between the minimum and maximum values of the dimension, is a positive integer, is the point on the nonlinear segmentation surface, that is, the data point in the tree data. For the Points on the dimensional nonlinear partitioning surface;
[0071] Repeat the above sub-segmentation data operation and calculate the length of the left and right sub-tree-shaped data paths to obtain the segmentation ratio and then obtain the length of the sub-tree-shaped data path;
[0072] Then randomly select a data in the denoised data as the segmentation point and repeat the above operation of obtaining the length of the subtree data path to obtain the tree data and the path length.
[0073] Step 300, calculating the isolation value through the path length and the isolation factor of the tree data through weighted fusion to obtain the isolation score of the tree data.
[0074] Here, step 300 includes:
[0075] The isolated value is calculated by the path length, where the isolated value for
[0076]
[0077] Tree Data Isolation Factor for
[0078]
[0079]
[0080] The isolated value and the tree data isolation factor are weighted fused to obtain the tree data isolation factor. , where isolated for
[0081]
[0082] in, is the size of the tree data is the path length of each data in the tree data on the tree data, is the average length of each data path in the tree data, is the harmonic number, is the data point in the tree data of Distance and data points in tree data To the data point in the tree data The maximum value in the interval, is the data point in the tree data The distance is less than dist The set of data points in the tree data, dist is the data point in the tree data The distances to other data points in the tree data are sorted from small to large. The distance between the data points is distance, For the fusion weight, here The value is 0.5.
[0083] Step 400, the tree data is isolated and judged to obtain abnormal data in the tree data, and after screening, the vacant positions of the tree data are filled by random forest and missing forest to obtain processed wind and solar power generation historical output data, processed historical weather data and processed load historical data.
[0084] Here, step 400 includes:
[0085] The isolated data of the tree data is isolated and judged to get the abnormal data in the tree data, where the isolated judgment is
[0086]
[0087] After filtering out abnormal data in the tree data, random forest and missing forest are used to predict the missing positions of the tree data to obtain random forest prediction values and missing forest prediction values;
[0088] The filling reserve value is obtained by calculating the absolute value of the difference between the random forest prediction value and the missing forest prediction value;
[0089] Select the minimum value among the filling reserve values to fill the vacant position of the tree data, repeat the above steps until there is no vacant position in the tree data, and then obtain the processed wind and solar power generation historical output data, processed historical weather data and processed load historical data;
[0090] in, Isolated nodes for tree data The standard deviation of .
[0091] Step 500, mapping the processed historical wind and solar power generation output data and the processed historical weather data with environmental variables and renewable energy output power to obtain renewable energy output power mapping values and the processed historical load data are used together through a neural network algorithm to predict future renewable energy output power prediction values and future load power prediction values.
[0092] Here, step 500 includes:
[0093] The processed historical output data of wind and solar power generation and the processed historical weather data are used to obtain the renewable energy output power mapping value through the mapping relationship between environmental variables and renewable energy output power;
[0094] The daily time period of a week is divided into four time periods, namely morning peak, evening peak, flat period and valley period. With a resolution of 1h, the renewable energy output power mapping value and the processed load history data of each time period are respectively predicted by the fuzzy brain emotional learning neural network with strong generalization ability to obtain the renewable energy output power prediction value and load power prediction value of the four time periods of each day in the next week, namely morning peak, evening peak, flat period and valley period;
[0095] The time period in a day is divided into 96 hours, with a time period of 15 minutes. The renewable energy output power mapping value and the processed load history data with a resolution of 15 minutes and an operating cycle of 1 hour are predicted by the fuzzy brain emotional learning neural network to obtain the renewable energy output power forecast value and load power forecast value for each hour of the next day.
[0096] For each 1-minute time period of a day, with a resolution of 1 minute and an operating cycle of 15 minutes, the output power mapping value of renewable energy in the previous 15 minutes is used to predict the output power of renewable energy in the next 15 minutes of the same day in the same time period through the recursive cerebellar model neural network with strong dynamic characteristics and fast response speed;
[0097] For each 1-minute time period of a day, with 1-minute resolution and 15-minute operation cycle, the load history data processed in the previous 15 minutes is used to predict the load power forecast value of the next 15 minutes at the same time of the same day through the recursive cerebellar model neural network.
[0098] The wavelet fuzzy brain emotion learning controller is a mathematical model that simulates the human brain's description and learning process of objective things. The wavelet fuzzy brain emotion learning controller consists of five spaces: input space, which consists of input variables; memory space, which is composed of wavelet functions as basis functions. Through this space, the input variables are mapped to the fuzzy set of the brain emotion learning controller; weight space, which includes the fuzzy weights of the amygdala system and the fuzzy weights of the prefrontal lobe system; degradation space, the process of degradation and defuzzification, according to the linear change relationship, the defuzzification operator of the amygdala system and the defuzzification operator of the prefrontal lobe system are obtained; output space, the output of the wavelet fuzzy brain emotion learning controller.
[0099] In this embodiment, the renewable energy output power mapping values of each operation cycle are formed into a set , get the input variable set of amygdala fuzzy system and prefrontal fuzzy system
[0100]
[0101] The wavelet function is used as the basis function to map the input variables through the memory space to obtain the fuzzy set of the brain emotion learning controller. , where the input variables are mapped to
[0102]
[0103]
[0104] According to the fuzzy rules of the pre-established wavelet fuzzy brain emotion learning controller, the learning process of the brain's feelings and emotions is simulated, and the fuzzy weights of the amygdala system and the fuzzy weights of the prefrontal lobe system are updated through adaptive learning rules and supervised learning methods. The fuzzy rules of the wavelet fuzzy brain emotion learning controller are:
[0105] Amygdala system: If is and is is , then (for )
[0106] Prefrontal system: If is and is is , then (for )
[0107] The initial fuzzy weight, initial learning rate and emotional signal adjustment parameters of the amygdala system and the initial fuzzy weight, initial learning rate and emotional signal adjustment parameters of the prefrontal lobe system are pre-set, wherein
[0108] Fuzzy weights of the amygdala system for
[0109]
[0110] Fuzzy weights of the prefrontal system for
[0111]
[0112] According to the physiological process of human brain emotion learning, the amygdala system and the prefrontal lobe system are adjusted according to the following parameter adaptive learning rules, where the fuzzy weight of the amygdala system in parameter adaptive learning is
[0113]
[0114] Fuzzy weights of the prefrontal system in the brain with adaptive parameter learning
[0115]
[0116] Adaptive supervised learning for updating the fuzzy weights of the amygdala system
[0117]
[0118] Adaptive supervised learning for updating fuzzy weights in the prefrontal system
[0119]
[0120] Updated amygdala system fuzzy weights The linear relationship between the sum and the fuzzy set is used to obtain the defuzzification operator of the updated amygdala system
[0121]
[0122] Updated fuzzy weights of the prefrontal system The linear relationship between the fuzzy set and the defuzzification operator of the prefrontal cortex system is updated
[0123]
[0124] Based on the defuzzification operator of the updating prefrontal system and update the defuzzification operator of the amygdala system The output result is obtained. According to the defuzzified output result, the simulation result of the brain emotion learning model is obtained and used to simulate the actual things to obtain the predicted value of renewable energy output power.
[0125]
[0126] in, is the input dimension, is the total output dimension, is the ith input variable, is the ith output, is the weight value of the amygdala fuzzy set of the oth output, is the weight value of the o-th output prefrontal fuzzy set, is the output of the oth amygdala fuzzy set, is the output of the oth prefrontal fuzzy set, and is the learning rate, Adjust parameters for emotional signals.
[0127] Similarly, it can be known that in this embodiment, the processed load history data of each operating cycle is formed into a set and the learning process of the brain's feelings and emotions is simulated according to the established fuzzy rules. The fuzzy weights of the amygdala system and the linear relationship of the fuzzy weights of the prefrontal lobe system are updated through adaptive learning rules and supervised learning methods to obtain the updated defuzzification operator of the amygdala system and the fuzzy weights of the prefrontal lobe system and the linear relationship of the fuzzy set, and the defuzzification operator of the prefrontal lobe system and the defuzzification operator of the prefrontal lobe system are updated to obtain the load power prediction value.
[0128] The recursive cerebellar neural network is a fast, generalized, and local approximation neural network based on neurophysiology, including an input layer, an associative memory layer, a receiving field layer, a weight memory layer, and an output layer. In this embodiment, the output power mapping values of renewable energy sources in the first 15 minutes are divided into a training set and a test set;
[0129] The training set is trained through the associative memory layer of the recursive cerebellar model neural network to obtain the associative memory layer output result
[0130]
[0131] Recurrent cerebellar neural network model using gauss function
[0132]
[0133] Associative memory layer output results Then the output layer is obtained through the receiving field layer and the weight memory layer Train to obtain the Gauss function M that makes the model converge. , weight and The trained model is obtained by using the same parameters, where the output layer for
[0134]
[0135] The test set is passed through the trained model through the output layer to obtain the neural network output result, and then weighted smoothing is used to obtain the predicted value of renewable energy output power for the next 15 minutes in the same day and time period. , where the weighted processing is
[0136]
[0137] in, is the mean of the Gaussian function, i.e. the center point, is the variance of the Gaussian function, i.e., the width, is the input feature vector , m is the dimension of the feature vector, and are the weights of the associative memory layer and the weight memory layer, n is the resolution of the input feature vector, t is the time series, is the output result of the associative memory layer, is the output layer result, is a constant.
[0138] Step 600, constructing an objective function for optimizing the energy structure of the integrated source-grid-load-storage combination to obtain an optimized objective function.
[0139] Here, step 600 includes:
[0140] Optimizing the objective function
[0141]
[0142] Optimize the constraints of the objective function
[0143]
[0144]
[0145]
[0146] in, is the target value of the scheduling plan, is the time period of the microgrid dispatch cycle, The economic objectives of each scheduling cycle are Environmental benefit goals Stability target of the tie line with the large power grid Stability goals within the system The weight coefficient of are the upper and lower limits of renewable energy output power, is the actual photovoltaic output power Actual wind power output is the actual output power of energy storage, The AC power between the microgrid and the large grid is the load power, is the minimum charging power for energy storage, is the maximum charging power of energy storage, is the energy storage power, Minimum state of charge for energy storage, Energy Storage The state of charge at the moment, Maximum state of charge for energy storage.
[0147] Step 700, the future renewable energy output power forecast value and the future load power forecast value are solved by the swarm intelligence algorithm to obtain the source-grid-load-storage integrated combined energy structure optimization plan.
[0148] The historical output data of wind and solar power generation, historical weather data and historical load data are collected and the denoised data are obtained by wavelet denoising; the denoised data are divided by nonlinear segmentation and the tree data is obtained by segmentation ratio and the path length is obtained; the isolated value is calculated by the path length and the isolation factor of the tree data is calculated, and the isolated points of the tree data are obtained by weighted fusion; the abnormal data in the tree data is obtained by isolation judgment for the isolated points of the tree data, and the vacant positions of the tree data are filled by random forest and missing forest to obtain the processed historical output data of wind and solar power generation, the processed historical weather data and the processed historical load data; the processed historical output data of wind and solar power generation and the processed historical weather data are mapped with environmental variables and renewable energy output power, and the renewable energy output power mapping value is obtained, and the processed load historical data is predicted together by neural network algorithm to obtain the future renewable energy output power prediction value and future load power prediction value; the objective function of optimizing the energy structure of the integrated combination of source, grid, load and storage is constructed to obtain the optimization objective function; the future renewable energy output power prediction value and the future load power prediction value are solved by the swarm intelligence algorithm to obtain the optimization objective function to obtain the optimization plan of the integrated combination of source, grid, load and storage. The problem of large anomaly detection deviation rate, inappropriate anomaly score selection, and anomaly scores far exceeding other scores in numerical distribution when randomly selecting attributes for data space segmentation is solved. The accuracy and efficiency of outlier removal are improved, the credibility and robustness of neural network model prediction are improved, and the long-term, short-term or ultra-short-term prediction accuracy of wind power, photovoltaic and other output data is improved, thereby reducing the operation economy of microgrid systems, improving environmental benefits and system operation efficiency, and improving the optimization capability of the energy structure of the integrated source, grid, load and storage combination, providing guidance for the actual operation of the integrated source, grid, load and storage microgrid system.
[0149] Here, step 700 includes:
[0150] The predicted values of renewable energy output power and load power for the four time periods every day in the next week, namely the morning peak, evening peak, flat period and valley period, are solved by the swarm intelligence algorithm to obtain the first energy structure optimization by solving the optimization objective function, so that the net power curve volatility of the average power of "source", "grid", "load" and "storage" in each time period is minimized, reducing the impact of microgrids on the large grid; at the same time, the optimal operation economy of the microgrid is formulated: by purchasing electricity from the large grid, the loss of energy storage charging and discharging, and the income obtained from the microgrid's electricity sales, new energy is absorbed as much as possible, so as to optimize the first energy structure with the minimum interaction volatility with the "grid";
[0151] On the basis of the first energy structure optimization, it is set that "source, load, and storage" do not interact with the "grid" side to reduce the impact of the microgrid on the large grid. The predicted values of renewable energy output power and load power per hour in the next day are solved by the swarm intelligence algorithm to obtain the second energy structure optimization, reduce the loss of battery charging and discharging in energy storage, and the net power of the average power of "source", "load", and "storage" in each hour is the minimum, wind power generation and photovoltaic power generation are fully output, and the load that can be migrated is the minimum.
[0152] On the basis of the secondary energy structure optimization, the forecasted output power of renewable energy and the forecasted load power for the next 15 minutes in the same time period of the same day are solved by the swarm intelligence algorithm to obtain the optimization objective function to obtain the source-grid-load-storage integrated energy structure optimization plan. Only the first time period is executed to reduce the loss of battery charging and discharging in energy storage as much as possible, and use the energy storage components with high power density to smooth power fluctuations as much as possible. The net power of the average power of "source", "load" and "storage" in each time period during the day is the minimum, the wind and solar power are fully output, and the load that can be migrated is the minimum. The source-grid-load-storage integrated energy structure optimization plan;
[0153] In this embodiment, the swarm intelligence algorithm uses the optimal point set theory to encode the predicted value of renewable energy output power and the predicted value of load power as follows: The population of individual positions is initialized to form the initial population. , where t is the number of iterations;
[0154] By optimizing the objective function Calculate the fitness and position of individuals
[0155]
[0156] Optimize the constraints of the objective function and screen to obtain appropriate individual fitness and position
[0157]
[0158]
[0159]
[0160] The fitness and position of the individual are updated through reverse differential evolution learning and adaptive distribution mutation individual optimization, and the optimal position is determined through the objective function, where the individual global search target position for
[0161]
[0162] If the objective function value of an individual's new position improves at that position, then the position is accepted. In this type of update, called efficient update, the algorithm is prevented from moving to non-optimal areas to obtain the global individual position.
[0163]
[0164] The global convergence to a better point in the hunting area increases the local search capability and the global search capability. The individual is modeled as
[0165]
[0166] Update individual attack target individual position for
[0167] =
[0168] Combine the individual position with each round of reverse learning update iteration to determine whether to update the updated individual, reversely calculate the individual position and select the updated individual, and obtain an optimal solution set after selecting all individuals in each direction. and reverse population The fitness of the solution is to select a better candidate solution. is the objective function, which is used to measure the quality of the candidate solution. , then use some to replace Otherwise, continue using After continuous searching, the variable scope of the current generation Much smaller than the initial variable range Get the updated population, where the reverse population is
[0169]
[0170] The individuals in the updated population are calculated through the objective function to obtain the fitness value and compare the position of the optimal individual, and the optimal individual position is updated through adaptive distribution mutation.
[0171]
[0172] Determine whether the current iteration number is greater than the maximum iteration number. If not, loop through the above steps of step 700. If it is greater, output the optimal individual position to obtain the optimal renewable energy output power prediction value and load power prediction value.
[0173] in, Based on this stage Only individuals in The new location of the For prey in The location of the dimension, is its objective function value, is a number that can be randomly equal to 1 or 2, For the The new state of the individual, is the objective function value based on this stage, is the maximum number of iterations, Based on this stage Only individuals in The new location of the is a constant, for The neighborhood radius of For the The new position of each individual, is the objective function value based on this stage, and are population and reverse population respectively Vector Dimensional component, is the group size, is the adaptive e distribution variation disturbance after The optimal position is the variation before disturbance The optimal solution position of iter By number of iterations Iter is the e distribution of the degree of freedom parameter. The optimization objective function is solved by the Pelican optimization method of reverse differential evolution learning and adaptive distribution mutation to improve the global search performance and the ability to escape the local optimum. It is one of the representative algorithms of swarm intelligence optimization algorithm, with the advantages of simple structure, easy practical application, fast solution speed and low computational cost.
[0174] Figure 2 This is a schematic diagram of the structure of the source-grid-load-storage integrated energy structure optimization system provided by the present invention, referring to Figure 2 The embodiment of the present invention provides a source-grid-load-storage integrated combined energy structure optimization system, the system embodiment and Figure 1 The method embodiment shown corresponds to the embodiment shown. The system can be specifically applied to various electronic devices. The system including this embodiment includes: a noise reduction unit S201, a data division unit S202, an isolated point calculation unit S203, an isolated determination unit S204, a future power prediction unit S205, an objective function construction unit S206 and an energy structure optimization unit S207.
[0175] The noise reduction unit S201 collects historical output data of wind and solar power generation, historical weather data and historical load data and obtains noise-reduced data by performing wavelet noise reduction;
[0176] The data partitioning unit S202 partitions the denoised data using a nonlinear partitioning plane and obtains tree data and path length using a partitioning ratio;
[0177] An isolated score calculation unit S203 calculates the isolated value through the path length and the isolated factor of the tree data and obtains the isolated score of the tree data through weighted fusion;
[0178] An isolation determination unit S204 performs isolation determination on the isolated data of the tree structure data to obtain abnormal data in the tree structure data and then fills the vacant positions of the tree structure data with a random forest and a missing forest to obtain processed historical output data of wind and solar power generation, processed historical weather data and processed historical load data;
[0179] The future power prediction unit S205 performs environmental variable and renewable energy output power mapping on the processed historical output data of wind and solar power generation and the processed historical weather data to obtain the renewable energy output power mapping value and the processed historical load data through a neural network algorithm to obtain the future renewable energy output power prediction value and the future load power prediction value;
[0180] An objective function construction unit S206 constructs an objective function for optimizing the energy structure of the integrated source-grid-load-storage combination to obtain an optimized objective function;
[0181] Energy structure optimization unit S207, future renewable energy output power forecast value and future load power forecast value are solved by swarm intelligence algorithm to obtain the source-grid-load-storage integrated combined energy structure optimization plan.
[0182] The source-grid-load-storage integrated combined energy structure optimization system provided by the embodiment of the present invention collects historical output data of wind and solar power generation, historical weather data and historical load data, and obtains denoised data through wavelet denoising; divides the denoised data through a nonlinear segmentation surface, obtains tree data through a segmentation ratio and obtains a path length; calculates isolated values through path lengths and calculates isolated factors of tree data through weighted fusion to obtain isolated points of tree data; obtains abnormal data in the tree data through isolation judgment for isolated points of tree data, and then fills in the vacant positions of the tree data through random forests and missing forests to obtain processed historical output data of wind and solar power generation, processed Historical weather data and processed load historical data; the processed wind and solar power generation historical output data and processed historical weather data are mapped with environmental variables and renewable energy output power to obtain the renewable energy output power mapping value and the processed load historical data, and the future renewable energy output power prediction value and future load power prediction value are predicted together through the neural network algorithm; the objective function of the source-grid-load-storage integrated energy structure optimization is constructed to obtain the optimized objective function; the future renewable energy output power prediction value and the future load power prediction value are solved by the swarm intelligence algorithm to obtain the source-grid-load-storage integrated energy structure optimization plan. Improve the accuracy and efficiency of outlier removal, improve the credibility and robustness of the prediction of the neural network model, and improve the long-term, short-term or ultra-short-term prediction accuracy of wind power, photovoltaic and other output data, thereby reducing the operation economy of the microgrid system, improving environmental benefits and system operation efficiency, and improving the optimization capability of the source-grid-load-storage integrated energy structure, providing guidance for the actual operation of the source-grid-load-storage integrated microgrid system.
[0183] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the method, which includes: collecting historical output data of wind and solar power generation, historical weather data and historical load data, and obtaining denoised data through wavelet denoising; dividing the denoised data through a nonlinear segmentation surface, and obtaining tree data through the segmentation ratio and path length; calculating the isolated value through the path length and the isolated factor of the tree data, and obtaining the isolated points of the tree data through weighted fusion; obtaining abnormal data in the tree data through isolation determination for the isolated points of the tree data, and then filling the vacant positions of the tree data through random forest and missing forest to obtain the processed historical output data of wind and solar power generation, Processed historical weather data and processed historical load data; mapping environmental variables and renewable energy output power of processed historical wind and solar power generation data and processed historical weather data to obtain renewable energy output power mapping values and the processed historical load data are predicted together through a neural network algorithm to obtain future renewable energy output power prediction values and future load power prediction values; constructing an objective function for optimizing the energy structure of the integrated combination of source, grid, load and storage to obtain an optimized objective function; solving the optimization objective function for the future renewable energy output power prediction values and future load power prediction values through a swarm intelligence algorithm to obtain an optimization plan for the energy structure of the integrated combination of source, grid, load and storage.
[0184] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, ReadOnly Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0185] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented by a processor to execute the methods provided by the above methods when executed, the methods comprising: collecting historical output data of wind and solar power generation, historical weather data and historical load data, and obtaining denoised data through wavelet denoising; dividing the denoised data through a nonlinear segmentation surface and obtaining tree data and path length through the segmentation ratio; calculating isolated values and tree data isolation factors through path lengths, and obtaining isolated points of the tree data through weighted fusion; obtaining abnormal data in the tree data through isolation judgment for the isolated points of the tree data, and filtering and filtering the missing positions of the tree data through random forests and missing forests. Fill in to obtain processed historical output data of wind and solar power generation, processed historical weather data and processed historical load data; perform environmental variable and renewable energy output power mapping on the processed historical output data of wind and solar power generation and the processed historical weather data to obtain renewable energy output power mapping values and predict the future renewable energy output power forecast values and future load power forecast values together with the processed historical load data through a neural network algorithm; construct an objective function for optimizing the energy structure of the integrated combination of source, grid, load and storage to obtain an optimized objective function; solve the optimization objective function for the future renewable energy output power forecast values and future load power forecast values through a swarm intelligence algorithm to obtain an optimization plan for the energy structure of the integrated combination of source, grid, load and storage.
[0186] The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.
[0187] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0188] The above are only preferred implementations of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.
Claims
1. A method for optimizing the energy structure of a source-grid-load-storage integrated combination, characterized in that: The method comprises: Collect historical wind and solar power generation output data, historical weather data and historical load data and obtain denoised data through wavelet denoising; The denoised data is divided by nonlinear segmentation and the tree data is obtained by segmentation ratio and the path length is obtained; The isolated value is calculated by path length and the isolated factor of tree data is calculated, and the isolated score of tree data is obtained by weighted fusion; The isolated data of the tree data is isolated and judged to obtain abnormal data in the tree data. After screening, the vacant positions of the tree data are filled by random forest and missing forest to obtain the processed historical output data of wind and solar power generation, the processed historical weather data and the processed historical load data. The processed historical output data of wind and solar power generation and the processed historical weather data are mapped with environmental variables and renewable energy output power to obtain the renewable energy output power mapping value and the processed historical load data together through the neural network algorithm to obtain the future renewable energy output power forecast value and the future load power forecast value; The objective function of optimizing the energy structure of the integrated source-grid-load-storage combination is constructed to obtain the optimized objective function; The future renewable energy output power forecast value and future load power forecast value are solved by swarm intelligence algorithm to obtain the optimization plan of source-grid-load-storage integrated energy structure. The method of dividing the denoised data by a nonlinear segmentation surface and obtaining tree data and path length by a segmentation ratio comprises: Randomly select one of the denoised data and segment it through the nonlinear segmentation surface to obtain sub-segmentation data and sub-tree data, where the nonlinear segmentation surface is … in, Dimension is a random number For the hypothetical random numbers between the minimum and maximum values of the dimension, is a positive integer, is the point on the nonlinear segmentation surface, that is, the data point in the tree data. For the Points on the dimensional nonlinear partitioning surface; Repeat the above sub-segmentation data operation and calculate the length of the left and right sub-tree-shaped data paths to obtain the segmentation ratio and then obtain the length of the sub-tree-shaped data path; Then randomly select one of the denoised data as the split point and repeat the above operation of obtaining the length of the subtree data path to obtain the tree data and the path length; The method of calculating the isolated value by path length and the isolated factor of the tree data by weighted fusion to obtain the isolated score of the tree data includes: The isolated value is calculated by the path length, where the isolated value for Tree Data Isolation Factor for The isolated value and the tree data isolation factor are weighted fused to obtain the tree data isolation factor. , where isolated for in, is the size of the tree data is the path length of each data in the tree data, is the average length of each data path in the tree data, is the harmonic number, is the data point in the tree data of Distance and data points in tree data To the data point in the tree data The maximum value in the interval, is the data point in the tree data The distance is less than dist The set of data points in the tree data, dist is the data point in the tree data The distances to other data points in the tree data are sorted from small to large. The distance between the data points is distance, is the fusion weight.
2. The method according to claim 1, characterized in that: The isolated analysis of the tree data obtains abnormal data in the tree data through isolated determination, and after screening, fills the vacant positions of the tree data through random forest and missing forest to obtain processed wind and solar power generation historical output data, processed historical weather data and processed load historical data, including: The isolated data of the tree data is isolated and judged to get the abnormal data in the tree data, where the isolated judgment is After filtering out abnormal data in the tree data, random forest and missing forest are used to predict the missing positions of the tree data to obtain random forest prediction values and missing forest prediction values; The filling reserve value is obtained by calculating the absolute value of the difference between the random forest prediction value and the missing forest prediction value; Select the minimum value among the filling reserve values to fill the vacant position of the tree data to obtain the processed historical output data of wind and solar power generation, the processed historical weather data and the processed historical load data; in, Isolated nodes for tree data The standard deviation of .
3. The method according to claim 2, characterized in that The processing of the processed historical wind and solar power generation output data and the processed historical weather data with environmental variables and renewable energy output power to obtain the renewable energy output power mapping value and the processed historical load data together with the processed load historical data through a neural network algorithm to obtain the future renewable energy output power prediction value and the future load power prediction value, including: Mapping the environmental variables and renewable energy output power of the processed historical wind and solar power generation data and the processed historical weather data to obtain the renewable energy output power mapping value; The output power mapping values of renewable energy in different time periods every day of the week and the processed load history data are used together through the fuzzy brain emotional learning neural network to predict the output power forecast values of renewable energy and load power forecast values in different time periods every day of the next week; The output power mapping values of renewable energy in different time periods of a day and the processed load history data are used together through the fuzzy brain emotional learning neural network to predict the output power forecast values of renewable energy and load power forecast values in different time periods of the next day; The output power mapping value of renewable energy in one time period of a day is predicted by the recursive cerebellar model neural network to obtain the predicted output power value of renewable energy in one time period of the same day; The processed load history data of the previous time period of the day is predicted by the recursive cerebellar model neural network to obtain the load power forecast value of the same time period of the same day.
4. The method according to claim 3, characterized in that The objective function of constructing the source-grid-load-storage integrated combined energy structure optimization is obtained by optimizing the objective function, including: Optimizing the objective function Optimize the constraints of the objective function in, is the target value of the scheduling plan, is the time period of the microgrid dispatch cycle, The economic objectives of each scheduling cycle are Environmental benefit goals Stability target of the tie line with the large power grid Stability goals within the system The weight coefficient of are the upper and lower limits of renewable energy output power, is the actual photovoltaic output power Actual wind power output is the actual output power of energy storage, The AC power between the microgrid and the large grid is the load power, is the minimum charging power for energy storage, is the maximum charging power of energy storage, is the energy storage power, Minimum state of charge for energy storage, Energy Storage The state of charge at the moment, Maximum state of charge for energy storage.
5. The method according to claim 4, characterized in that The future renewable energy output power forecast value and the future load power forecast value are solved by the swarm intelligence algorithm to obtain the source-grid-load-storage integrated combined energy structure optimization plan, including: The predicted values of renewable energy output power and load power at different time periods every day in the next week are solved by swarm intelligence algorithm to obtain the first energy structure optimization; Based on the first energy structure optimization, the predicted values of renewable energy output power and load power in different time periods in the future day are solved by swarm intelligence algorithm to obtain the second energy structure optimization; On the basis of the optimization of the secondary energy structure, the output power forecast value and load power forecast value of renewable energy in the same time period of the same day are solved by swarm intelligence algorithm to obtain the optimization plan of the integrated energy structure of source, grid, load and storage.
6. A source-grid-load-storage integrated energy structure optimization system, characterized in that: The system comprises: The noise reduction unit collects historical output data of wind and solar power generation, historical weather data and historical load data and obtains noise-reduced data through wavelet noise reduction; A data partitioning unit, which partitions the denoised data through a nonlinear partitioning plane and obtains tree data and path length through a partitioning ratio; An isolated score calculation unit calculates the isolated value through the path length and the isolated factor of the tree data, and obtains the isolated score of the tree data through weighted fusion; An isolation determination unit is used to determine the isolated data of the tree data, obtain the abnormal data in the tree data through isolation determination, and then fill the vacant positions of the tree data through random forest and missing forest to obtain the processed historical output data of wind and solar power generation, the processed historical weather data and the processed historical load data; The future power prediction unit maps the processed historical output data of wind and solar power generation and the processed historical weather data with the output power of renewable energy, obtains the output power mapping value of renewable energy, and predicts the future output power prediction value of renewable energy and the future load power prediction value together with the processed historical load data through a neural network algorithm; An objective function construction unit is used to construct an objective function for optimizing the energy structure of the integrated source-grid-load-storage combination to obtain an optimized objective function; Energy structure optimization unit, the future renewable energy output power forecast value and future load power forecast value are solved by swarm intelligence algorithm to obtain the source-grid-load-storage integrated energy structure optimization plan; The method of dividing the denoised data by a nonlinear segmentation surface and obtaining tree data and path length by a segmentation ratio comprises: Randomly select one of the denoised data and segment it through the nonlinear segmentation surface to obtain sub-segmentation data and sub-tree data, where the nonlinear segmentation surface is … in, Dimension is a random number For the hypothetical random numbers between the minimum and maximum values of the dimension, is a positive integer, is the point on the nonlinear segmentation surface, that is, the data point in the tree data. For the Points on the dimensional nonlinear partitioning surface; Repeat the above sub-segmentation data operation and calculate the length of the left and right sub-tree-shaped data paths to obtain the segmentation ratio and then obtain the length of the sub-tree-shaped data path; Then randomly select one of the denoised data as the split point and repeat the above operation of obtaining the length of the subtree data path to obtain the tree data and the path length; The method of calculating the isolated value by path length and the isolated factor of the tree data by weighted fusion to obtain the isolated score of the tree data includes: The isolated value is calculated by the path length, where the isolated value for Tree Data Isolation Factor for The isolated value and the tree data isolation factor are weighted fused to obtain the tree data isolation factor. , where isolated for in, is the size of the tree data is the path length of each data in the tree data, is the average length of each data path in the tree data, is the harmonic number, is the data point in the tree data of Distance and data points in tree data To the data point in the tree data The maximum value in the interval, is the data point in the tree data The distance is less than dist The set of data points in the tree data, dist is the data point in the tree data The distances to other data points in the tree data are sorted from small to large. The distance between the data points is distance, is the fusion weight.
7. Electronic equipment, including: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the program.
8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.
