Optimization system and method for distribution transformer area based on industrial data processing

Abstract

The present application relates to the technical field of optimization control of distribution transformer area, and particularly relates to a system and method for optimization of distribution transformer area based on industrial data processing, which analyzes composite power quality abnormal conditions and dynamic load distribution abnormal conditions of the distribution transformer area in combination with electrical operation data and environmental state data of the distribution transformer area; realizes pre-analysis of cross-area collaborative overload risk after different combination actions of tie switches of the distribution transformer area; analyzes abnormal risk of loop closing impact operation of the tie switches in the distribution transformer area in combination with topological switch state data and environmental state data of the distribution transformer area; realizes evaluation of cross-area global collaborative risk, and generates a Pareto optimal tie switch control scheme set; and based on the Pareto optimal tie switch control scheme set, a differentiated tie switch collaborative control scheme is executed, thereby greatly reducing protection misoperation and switch damage events, and significantly improving long-term reliability and economy of global operation of the distribution transformer area.

Description

Technical Field

[0001] This invention relates to the field of distribution transformer area optimization control technology, and in particular to a distribution transformer area optimization system and method based on industrial data processing. Background Technology

[0002] As key power supply units directly facing end users in low-voltage distribution networks, the stability of distribution transformer areas and their power quality directly impact industrial production efficiency, commercial operations, and residents' quality of life. With the large-scale integration of distributed photovoltaic power generation, electric vehicle charging stations, and new energy storage devices, distribution transformer areas are evolving from traditional unidirectional passive power supply networks to multi-source, bidirectional active distribution networks. Load characteristics are becoming increasingly complex, and the randomness of power flow is significantly enhanced. Meanwhile, the tie switches and sectionalizing switches between transformer areas provide flexible operational means for power restoration and load balancing. However, the closing-loop inrush current, transient overvoltage, and power redistribution processes triggered by switch operations can pose serious threats to equipment insulation, protection setting coordination, and power supply continuity.

[0003] However, existing technologies, when optimizing the control of distribution transformer areas, do not correlate the combined operation schemes of tie switches with the future operating states projected in the digital twin after the operation. For example, when a distribution transformer area needs to close the tie switch to transfer load due to peak load, without considering the dynamic effects of power flow distribution, voltage changes, and harmonic propagation over a period of time after the loop is closed, it is difficult to accurately determine whether the operation scheme will cause cascading overloads or power quality deterioration in adjacent distribution transformer areas, leading to misjudgments of the scheme's feasibility. Furthermore, existing technologies lack comprehensive consideration of multi-dimensional risks such as loop inrush current, switch lifespan loss, and future load fluctuations, making it difficult for the assessment results to accurately reflect the cross-regional collaborative risks and failing to provide maintenance personnel with a comprehensive and reliable basis for control schemes. Summary of the Invention

[0004] To overcome the defects and shortcomings of existing technologies, this invention provides a distribution transformer area optimization system and method based on industrial data processing. It synchronously collects electrical, environmental and topology data through IoT sensors, and uses isolated forest, multi-task LSTM, GBDT and MLP models to detect power quality and load anomalies, predict future power flow and assess the risk of loop impact, and combines digital twin simulation of multi-switch combination overload risk. It uses NSGA-II to generate Pareto optimal solution set and executes differentiated collaborative control to minimize the risk across the entire distribution transformer area.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, embodiments of the present invention provide a method for optimizing distribution transformer substations based on industrial data processing, comprising the following steps: Step S100: Simultaneously collect electrical operation data, environmental status data, and topology switch status data of the distribution transformer area through IoT sensors deployed in the cross-transformer interconnection network; Step S200: Combining the electrical operation data and environmental status data of the distribution transformer area, and based on the cross-distribution area digital twin model, analyze the composite power quality anomaly and dynamic load distribution anomaly of the distribution transformer area; and based on the analysis results of the composite power quality anomaly and dynamic load distribution anomaly, conduct a pre-analysis of the cross-distribution area coordinated overload risk after different combinations of tie switch actions in the distribution transformer area. Step S300: Combine the topology switch status data and environmental status data of the distribution transformer area to analyze the abnormal risk of the closing impact operation of the tie switch in the distribution transformer area. Step S400: Based on the cross-transformer area collaborative overload risk analysis results and the abnormal risk analysis results of the closing impact operation of the tie switch, the cross-transformer area overall collaborative risk is assessed, and a Pareto optimal tie switch control scheme set is generated. Step S500: Based on the Pareto optimal tie switch control scheme set, execute the differentiated tie switch coordinated control scheme.

[0006] In conjunction with the first aspect, this embodiment of the invention also provides a first implementation method of the first aspect, wherein step S200 combines the electrical operation data and environmental status data of the distribution transformer area, and analyzes the composite power quality anomalies and dynamic load distribution anomalies of the distribution transformer area based on a cross-distribution area digital twin model; specifically including the following: Step S201: Extract the real-time voltage, real-time current and effective values ​​of each phase of each transformer substation from the electrical operation data, and extract the ambient temperature and ambient humidity from the environmental status data to form a power quality analysis dataset. Step S202: Based on the power quality analysis dataset, calculate the voltage deviation, three-phase unbalance, and harmonic distortion index of the distribution transformer area respectively; Step S203: Combine the voltage deviation index, three-phase unbalance index, and harmonic distortion index of the transformer substation area with the ambient temperature and humidity of the transformer substation area to form a multi-dimensional feature vector of the transformer substation area; input the multi-dimensional feature vector of the transformer substation area into the pre-trained isolated forest anomaly detection model, and the model outputs the composite power quality anomaly status of the transformer substation area.

[0007] In conjunction with the first aspect, this embodiment of the invention also provides a second implementation method, which, based on a cross-substation digital twin model, analyzes the composite power quality anomalies and dynamic load distribution anomalies of the substation by combining electrical operation data and environmental status data of the substation area; specifically, it also includes the following steps: Step S204: Obtain the historical active power sequence of each phase of each transformer area under multiple historical time windows from the electrical operation data, and obtain the temperature and humidity of the corresponding historical time windows from the environmental status data to construct a load prediction dataset; divide the load prediction dataset into a load prediction training set and a load prediction validation set; Step S205: Construct a multi-task long short-term memory network prediction model. Based on the load prediction training set, use the historical active power sequence, temperature, and humidity of each historical time window as input features of the multi-task long short-term memory network prediction model. Use the active power values ​​of each transformer area and each phase at each time within a preset future time period after the corresponding historical time window in the load prediction training set as output targets of the multi-task long short-term memory network prediction model. Train the multi-task long short-term memory network prediction model to obtain an initial load prediction model. Validate the initial load prediction model using the load prediction validation set. Output an initial load prediction model with an accuracy greater than or equal to a preset second model as the load prediction model. Step S206: Input the active power sequence, temperature and humidity of each phase of each transformer area in the current time window into the load prediction model, and output the predicted active power of each phase of each transformer area at each time within a preset time period after the current time window. Step S207: Calculate the prediction coefficient of variation and spatial load imbalance based on the predicted active power value; Step S208: Combine the predicted coefficient of variation, spatial load imbalance degree and the standard deviation of the load rate of each distribution transformer area, input them into the pre-trained gradient boosting decision tree regression model, and output the dynamic load distribution anomaly of the distribution transformer area.

[0008] In conjunction with the first aspect, this embodiment of the invention also provides a third implementation of the first aspect, and based on the analysis results of composite power quality anomalies and dynamic load distribution anomalies, conducts a pre-analysis of the cross-regional coordinated overload risk after different combinations of interconnection switches in the distribution transformer area operate; specifically including: Step S209: Construct a cross-regional digital twin model, which includes transformer parameters, line impedance parameters, topology connections, and phase assignment information of each load node for all participating distribution areas. Step S210: Enumerate all combination action schemes of tie switches and sectionalizing switches, wherein each combination action scheme of tie switches and sectionalizing switches is used to specify the closed tie switch and the open sectionalizing switch. Step S211: For each combination of tie switch and sectional switch operation scheme, perform the corresponding topology reconstruction operation in the cross-transformer area digital twin model. The predicted active power values ​​of each transformer area and each phase at each time within the future preset time period obtained in step S206 are injected into the corresponding load nodes of the reconstructed network according to the phase. Quasi-steady-state power flow calculation is performed time by time, and the load rate of each transformer and each line at each time is recorded. Step S212: For each combined action scheme of tie switch and sectional switch, calculate the cross-area coordinated overload risk of the distribution transformer area under the corresponding combined action scheme based on the quasi-steady-state power flow calculation results and combined with the abnormal power quality and dynamic load distribution conditions.

[0009] In conjunction with the first aspect, this embodiment of the invention also provides a fourth implementation of the first aspect, wherein step S300 combines the topology switch status data and environmental status data of the distribution transformer area to analyze the abnormal risk of the loop-closing impact operation of the tie switch in the distribution transformer area; including the following specific steps: Step S301: Obtain the real-time voltage amplitude, voltage phase angle difference, loop equivalent resistance, loop equivalent reactance, remaining operating life and contact resistance of both sides of the tie switch to be analyzed from the topology switch status data; and obtain the ambient temperature and ambient humidity at the corresponding tie switch installation location from the environmental status data. Step S302: Calculate the expected peak value of the closing loop impact current at the moment the tie switch to be analyzed closes, based on the real-time voltage amplitude, voltage phase angle difference, loop equivalent impedance and dynamic transient peak coefficient. Step S303: Calculate the thermal effect impact risk value based on the expected peak value of the closing loop impact current, the rated short-time withstand current of the tie switch, and the current carrying capacity derating function determined by the ambient temperature. Step S304: Calculate the risk value of operating life decay based on the remaining operating life and the rated operating life; Step S305: The thermal effect impact risk value, the operational life attenuation risk value, the ambient temperature and humidity at the installation location of the tie switch to be analyzed, and the time elapsed since the last operation of the tie switch are used as input features and input into a pre-trained multilayer perceptron neural network model. The output is the abnormal risk of the closing impact operation of the tie switch to be analyzed in the distribution transformer area.

[0010] In conjunction with the first aspect, this embodiment of the invention also provides a fifth implementation of the first aspect, wherein step S400 assesses the overall cross-regional collaborative risk based on the cross-regional coordinated overload risk analysis results and the abnormal risk analysis results of the closing-loop impact operation of the tie switch, and generates a Pareto-optimal tie switch control scheme set; including the following specific steps: Step S401: For each combined action scheme of tie switch and sectional switch, obtain the cross-regional coordinated overload risk calculated by the corresponding combined action scheme, and record it as the first risk value of the corresponding combined action scheme; obtain the maximum value among the abnormal risks of loop impact operation calculated by all action tie switches in the corresponding combined action scheme, and record it as the second risk value of the corresponding combined action scheme; and calculate the switch action attenuation index of the corresponding combined action scheme based on the remaining operating life of each switch that has taken action in the corresponding combined action scheme. Step S402: Using the first risk value, second risk value, and switch action decay index of each combination action scheme of the tie switch and segment switch as three optimization objectives, a fast non-dominated sorting genetic algorithm is used to perform multi-objective optimization search in all combination action schemes to obtain the Pareto front solution set, which constitutes the Pareto optimal tie switch control scheme set; each combination action scheme in the Pareto optimal tie switch control scheme set includes a specific switch operation sequence, the corresponding first risk value, second risk value, and switch action decay index; Step S403: The first risk value, the second risk value, and the switch action attenuation index of the corresponding combined action scheme in the Pareto optimal interconnection switch control scheme set are weighted and summed to obtain the cross-regional coordinated risk of the corresponding combined action scheme in the Pareto optimal interconnection switch control scheme set.

[0011] In conjunction with the first aspect, this embodiment of the invention also provides a sixth implementation of the first aspect, wherein step S500, based on the Pareto-optimal set of tie switch control schemes, executes a differentiated tie switch cooperative control scheme; including the following specific steps: Step S501: Extract the cross-regional full-domain collaborative risk of all combined action schemes in the Pareto optimal interconnection switch control scheme set; Step S502: The combined action scheme corresponding to the minimum risk of cross-regional full-area coordination is executed as the coordination control scheme of the interconnection switch.

[0012] Secondly, embodiments of the present invention also provide a distribution transformer area optimization system based on industrial data processing, including: The data acquisition module is used to synchronously collect electrical operation data, environmental status data, and topology switch status data of the distribution transformer area through IoT sensors deployed in the cross-area interconnection network; The switch operation simulation module is used to combine the electrical operation data and environmental status data of the distribution transformer area, and analyze the composite power quality anomaly and dynamic load distribution anomaly of the distribution transformer area based on the cross-transformer area digital twin model; and based on the analysis results of composite power quality anomaly and dynamic load distribution anomaly, to conduct a simulation analysis of the cross-transformer area coordinated overload risk after different combinations of interconnection switch actions in the distribution transformer area. The operation risk analysis module is used to analyze the abnormal risk of loop-closing impact operation of the tie switch in the distribution transformer area by combining the topology switch status data and environmental status data of the distribution transformer area. The candidate scheme output module is used to evaluate the overall risk of cross-transformer area coordination based on the cross-transformer area coordination overload risk analysis results and the tie switch closing impact operation abnormal risk analysis results, and generate a Pareto optimal tie switch control scheme set. The execution module is used to execute differentiated interconnection switch coordinated control schemes based on the Pareto optimal interconnection switch control scheme set.

[0013] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. This invention significantly improves the sensitivity and accuracy of identifying power quality degradation by detecting composite power quality anomalies and dynamic load distribution anomalies, realizes a forward-looking assessment of load fluctuation uncertainties, and effectively reduces the probability of overload tripping due to load mutations. 2. This invention is based on the time-by-time extrapolation of quasi-steady-state power flow and dynamically corrects the safe load rate threshold according to the real-time power quality deterioration and load prediction deviation, making the cross-regional collaborative overload risk prediction closer to the actual future operating boundary, avoiding the optimistic or conservative bias caused by fixed thresholds, and improving the credibility of risk prediction. 3. This invention uses a fast non-dominated sorting genetic algorithm to generate a Pareto optimal tie switch control scheme set, so that the final executed differentiated coordination scheme achieves the optimal balance between overload safety, operational impact and equipment life, which significantly improves the long-term reliability and economy of the entire distribution transformer area.

[0014] Other features and advantages of the invention will be set forth in the following description, or some features and advantages may be inferred from the description or determined without doubt, or may be learned by practicing the techniques described above.

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1This is a schematic diagram of the overall process of the distribution transformer area optimization method based on industrial data processing according to an embodiment of the present invention; Figure 2 This is a flowchart of step S200 in the distribution transformer area optimization method based on industrial data processing according to an embodiment of the present invention. Figure 3 This is a flowchart of step S300 in the distribution transformer area optimization method based on industrial data processing according to an embodiment of the present invention. Figure 4 This is a flowchart of step S400 in the distribution transformer area optimization method based on industrial data processing according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the distribution transformer area optimization system based on industrial data processing according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] Example 1

[0021] As the core unit of the low-voltage distribution network, the operation optimization of distribution transformer areas directly affects power supply reliability and power quality. Traditional distribution transformer area management relies on manual experience and offline power flow calculations, making it difficult to cope with dynamic load fluctuations and combined power quality pollution caused by distributed power source access and random charging of electric vehicles. In recent years, digital twin technology has provided a new approach to collaborative control of distribution transformer areas by constructing a virtual mirror of the physical power grid, enabling real-time mapping and advanced simulation of operating status. However, existing digital twin applications are mostly concentrated on monitoring single devices or single distribution transformer areas, lacking system solutions for multi-objective collaborative optimization in cross-distribution transformer area interconnection scenarios.

[0022] To address the aforementioned technological gaps, this invention proposes a distribution transformer area optimization method based on industrial data processing. It synchronously collects electrical, environmental, and topological data through IoT sensors, constructs a cross-transformer area digital twin model, and integrates isolated forest anomaly detection, multi-task LSTM prediction, gradient boosting decision tree regression, and multilayer perceptron classification to achieve multi-dimensional quantification of composite power quality anomalies, dynamic load distribution anomalies, and loop-closing impact operation anomalies. Furthermore, it utilizes a fast non-dominated sorting genetic algorithm to generate a Pareto-optimal tie switch control scheme set, ultimately executing differentiated collaborative control to significantly improve the overall operational safety and economy across the entire distribution transformer area.

[0023] To facilitate understanding, the distribution transformer area optimization method based on industrial data processing provided in this embodiment of the invention will be described first, referring to... Figure 1 Specifically, it includes the following steps: Step S100: Simultaneously collect electrical operation data, environmental status data, and topology switch status data of the distribution transformer area through IoT sensors deployed in the cross-transformer interconnection network; In this embodiment, the electrical operation data includes the real-time voltage, real-time current and effective values ​​of each harmonic current on the three-phase lines of each transformer area. These data are synchronously acquired by intelligent sensors installed at the low-voltage side outgoing line terminal and key branch nodes of the transformer at a sampling rate of 128 points per cycle. The sampling timestamp is guaranteed to have a cross-transformer area synchronization error of less than 1 microsecond by the IEEE 1588 precise time synchronization protocol. Environmental status data includes ambient temperature and humidity at each sensor installation point. The data is recorded once per minute by a digital temperature and humidity chip integrated in the smart terminal and aggregated to the edge computing gateway via a wireless self-organizing network.

[0024] The topology switch status data collection includes the open / close position, contact resistance, and remaining operating life of all tie switches and sectionalizing switches. The remaining operating life is calculated by accumulating the number of switch opening and closing operations and combining it with the mechanical life curve provided by the manufacturer. The contact resistance is automatically detected after each opening and closing operation using the micro-ohm level measurement circuit built into the switch.

[0025] Step S200: Combining the electrical operation data and environmental status data of the distribution transformer area, and based on the cross-distribution area digital twin model, analyze the composite power quality anomaly and dynamic load distribution anomaly of the distribution transformer area; and based on the analysis results of the composite power quality anomaly and dynamic load distribution anomaly, conduct a pre-analysis of the cross-distribution area coordinated overload risk after different combinations of tie switch actions in the distribution transformer area.

[0026] Step S300: Combine the topology switch status data and environmental status data of the distribution transformer area to analyze the abnormal risk of the closing impact operation of the tie switch in the distribution transformer area.

[0027] Step S400: Based on the cross-transformer area collaborative overload risk analysis results and the abnormal risk analysis results of the closing impact operation of the tie switch, the cross-transformer area overall collaborative risk is assessed, and a Pareto optimal tie switch control scheme set is generated.

[0028] Step S500: Based on the Pareto optimal tie switch control scheme set, execute the differentiated tie switch coordinated control scheme.

[0029] In practical implementation, step S200 combines the electrical operation data and environmental status data of the distribution transformer area, and analyzes the composite power quality anomalies and dynamic load distribution anomalies of the distribution transformer area based on the cross-transformer area digital twin model; refer to Figure 2 Specifically, it includes the following:

[0030] Step S201: Extract the real-time voltage, real-time current and effective values ​​of each phase of each transformer substation from the electrical operation data, and extract the ambient temperature and ambient humidity from the environmental status data to form a power quality analysis dataset.

[0031] Step S202: Based on the power quality analysis dataset, calculate the voltage deviation, three-phase unbalance, and harmonic distortion index of the distribution transformer area respectively;

[0032] In a specific embodiment, the voltage deviation of each phase in the distribution transformer area is calculated separately, and the maximum value of the voltage deviation of all phases is taken as the voltage deviation of the distribution transformer area; wherein, the calculation formula for the voltage deviation of each phase is: ; In the formula, Dvi represents the voltage deviation of the i-th phase, Vi represents the real-time voltage of the i-th phase, and Vb represents the nominal phase voltage. The maximum allowable absolute value of the voltage deviation is given by max(), which means taking the maximum value of the element within the parentheses. Obtain the real-time current of each phase, take the difference between the maximum and minimum values ​​of the real-time current of each phase, divide the difference by the maximum value of the real-time current of each phase, and obtain the three-phase unbalance of the distribution transformer area. Calculate the harmonic distortion index for each phase in the distribution transformer area, and take the maximum value of the harmonic distortion index for all phases as the harmonic distortion index of the distribution transformer area; the calculation formula for the harmonic distortion index of each phase is as follows: ; In the formula, Dti is the harmonic distortion index of the i-th phase. Let h be the effective value of the h-th harmonic current of the i-th phase. Let H be the effective value of the fundamental current of the i-th phase, H be the highest calculated harmonic order, and h be any term from 2 to H. Step S203: Combine the voltage deviation index, three-phase unbalance index, and harmonic distortion index of the transformer substation area with the ambient temperature and humidity of the transformer substation area to form a multi-dimensional feature vector of the transformer substation area; input the multi-dimensional feature vector of the transformer substation area into the pre-trained isolated forest anomaly detection model, and the model outputs the composite power quality anomaly status of the transformer substation area.

[0033] The pre-training process of the isolated forest anomaly detection model is as follows: Collect historical multidimensional feature vector sets of multiple time windows under normal operating conditions of the distribution transformer area, and divide the historical multidimensional feature vector sets into anomaly detection training set and anomaly detection validation set; construct an isolated forest model, train the isolated forest model based on the anomaly detection training set to obtain an initial isolated forest anomaly detection model; validate the initial isolated forest anomaly detection model through the anomaly detection validation set, and use the initial isolated forest anomaly detection model with an anomaly detection accuracy greater than or equal to the preset first model accuracy as the isolated forest anomaly detection model; For example, the model in this embodiment adopts the isolated forest algorithm. During the training process, a total of 4,320 samples of historical multidimensional feature vectors with a time window length of 10 minutes are collected from the historical normal operation status of the distribution transformer area (manually marked with no power quality events) for 30 consecutive days. These samples are then divided into training set and validation set in a 7:3 ratio. The model parameters are set as follows: the number of isolated trees is 200, the subsampling size of each tree is 256 samples, the maximum number of features is 5 (i.e., all features), the maximum depth of the tree is ceil(log2(256))=8, and the anomaly score threshold is determined by the receiver operating characteristic curve (ROC) on the validation set. Specifically, in this embodiment, the false positive rate is less than 5% and the true positive rate is the highest, and the final threshold is set to 0.58.

[0034] Furthermore, this embodiment does not use explicit regularization during training, and can further prevent overfitting naturally through random subsampling and random feature selection. The model validation accuracy is required to reach more than 95%, that is, the anomaly detection accuracy is ≥0.95. The model accuracy of this embodiment is verified to reach 96.2%, which meets the requirements.

[0035] In practical implementation, the electrical operation data and environmental status data of the distribution transformer area are combined, and the composite power quality anomalies and dynamic load distribution anomalies of the distribution transformer area are analyzed based on the cross-transformer area digital twin model; specifically, the following steps are also included: Step S204: Obtain the historical active power sequence of each phase of each transformer area under multiple historical time windows from the electrical operation data, and obtain the temperature and humidity of the corresponding historical time windows from the environmental status data to construct a load prediction dataset; divide the load prediction dataset into a load prediction training set and a load prediction validation set; Step S205: Construct a multi-task long short-term memory network prediction model. Based on the load prediction training set, use the historical active power sequence, temperature, and humidity of each historical time window as input features of the multi-task long short-term memory network prediction model. Use the active power values ​​of each transformer area and each phase at each time within a preset future time period after the corresponding historical time window in the load prediction training set as output targets of the multi-task long short-term memory network prediction model. Train the multi-task long short-term memory network prediction model to obtain an initial load prediction model. Validate the initial load prediction model using the load prediction validation set. Output an initial load prediction model with an accuracy greater than or equal to a preset second model as the load prediction model. For example, this embodiment constructs a multi-task long short-term memory network prediction model. The model structure is an input layer (specifically, in this embodiment, the feature dimension of the input layer is the total number of transformer areas × 3 phases + temperature and humidity 2 dimensions), the length of the input sequence is the past 6 time windows (i.e., 90 minutes of historical data), the hidden layer uses two LSTM layers, the first LSTM layer has 128 units and the second layer has 64 units, the dropout ratio between the two layers is set to 0.25 to suppress overfitting, the output layer is a fully connected layer, the number of neurons is the active power value of each phase of each transformer area at each time point within the next 15 minutes (in this embodiment, the prediction is for the next 4 hours, i.e., 16 time points), so the output dimension is the number of transformer areas × 3 × 16; Training employed the mean squared error loss function, the Adam optimizer, an initial learning rate of 0.001, which decayed to 0.9 every 10 epochs. The batch size was set to 64, and the maximum number of training epochs was 200. An early stopping strategy was used to monitor the validation set loss; training was terminated if the validation loss did not decrease for 20 consecutive epochs. The final model was considered to meet the second model accuracy requirement when its mean absolute percentage error (MAPE) on the validation set was less than 8% (in this embodiment, MAPE ≤ 10%), which actually reached 7.3%.

[0036] Step S206: Input the active power sequence, temperature and humidity of each phase of each transformer area in the current time window into the load prediction model, and output the predicted active power of each phase of each transformer area at each time within a preset time period after the current time window. Step S207: Calculate the prediction coefficient of variation and spatial load imbalance based on the predicted active power value; specifically, the prediction coefficient of variation is used to measure the overall fluctuation of the predicted active power value over a future period relative to the average level of the same period in history. The larger the value, the more drastic the predicted load change. The process involves obtaining the sum of the predicted total active power of all participating distribution areas within a future preset time period (specifically, multiple times within the future preset time period) and the average of the total active power of all participating distribution areas during the same historical time period (for example, the future preset time period and the same historical time period are both weekday afternoons from 2 PM to 6 PM). The sum of the predicted total active power at all times within the future preset time period is divided by the number of times within the future preset time period to obtain the average predicted load within the future preset time period. The absolute difference between the predicted total active power at each time point within the future preset time period and the average predicted load within the future preset time period is taken as the prediction deviation at each time point within the future preset time period. The sum of the prediction deviations at all times within the future preset time period is divided by the number of times within the future preset time period to obtain the average prediction deviation within the future preset time period. The average prediction deviation within the future preset time period is divided by the average total active power of all participating distribution areas during the same historical time period to obtain the prediction coefficient of variation. Furthermore, the spatial load imbalance degree is used to measure the uniformity of load distribution among participating coordinated distribution areas at the current moment. A value of 0 represents complete balance, while a larger value indicates that the load is more concentrated in one or a few distribution areas. The current active power of all distribution areas is added together to obtain the total power of all distribution areas at the current moment. The current active power of each distribution area is divided by the total power of all distribution areas at the current moment to obtain the actual load share of each distribution area. The value is divided by the total number of distribution areas to obtain the ideal share. The absolute difference between the actual load share and the ideal share of each distribution area is taken as the share deviation of each distribution area. The sum of the share deviations of all distribution areas and divided by the total number of distribution areas gives the spatial load imbalance degree. Step S208: Combine the predicted coefficient of variation, spatial load imbalance, and the standard deviation of the current load rate of each distribution transformer area, input them into the pre-trained gradient boosting decision tree regression model, and output the dynamic load distribution anomaly of the distribution transformer area; specifically, the standard deviation of the current load rate of each distribution transformer area is based on the transformer load rate and is used to measure the dispersion of the current operating pressure of each distribution transformer area; divide the current active power of each distribution transformer area by the corresponding rated capacity of the transformer to obtain the current load rate of the corresponding distribution transformer area; sum the current load rates of all distribution transformer areas and divide by the total number of distribution transformer areas to obtain the average load rate; take the square of the difference between the current load rate of each distribution transformer area and the average load rate, and sum the square values ​​of all distribution transformer areas and divide by the total number of distribution transformer areas, and perform a square root operation on the division result to obtain the standard deviation of the current load rate of each distribution transformer area; The pre-training process of the gradient boosting decision tree regression model is as follows: Historical operational data is obtained, and the historical prediction variation coefficient, historical spatial load imbalance degree and historical load rate standard deviation corresponding to multiple historical time windows are calculated as sample features. The dynamic load anomaly label corresponding to the historical time window is used as the regression target to construct a dynamic load distribution anomaly dataset. The dynamic load distribution anomaly dataset is divided into a dynamic load distribution anomaly training set and a dynamic load distribution anomaly validation set. A gradient boosting decision tree regression model is constructed, and the gradient boosting decision tree regression model is trained based on the dynamic load distribution anomaly training set to obtain an initial gradient boosting decision tree regression model. The initial gradient boosting decision tree regression model is validated through the dynamic load distribution anomaly validation set, and the initial gradient boosting decision tree regression model with an accuracy greater than or equal to the preset third model is used as the gradient boosting decision tree regression model. For example, in this embodiment, the historical prediction coefficient of variation, historical spatial load imbalance and historical load rate standard deviation corresponding to 3650 historical time windows (96 windows per day, sampled at a 10-minute sliding step) in the past year are collected as sample features. The binary label (0 / 1) of whether a transformer overload (load rate exceeding 90%) or line overcurrent event occurs within one hour after the corresponding window is used as the regression target. In practice, the regression form of GBDT is used to output the anomaly probability.

[0037] The dataset was split into a 7:3 ratio for training and validation. The model parameters were as follows: 300 trees, maximum depth of 5, learning rate of 0.05, subsampling ratio of 0.8, minimum number of leaf node samples of 10, and a squared loss function. During training, the parameters were tuned using 5-fold cross-validation. The coefficient of determination on the validation set had to be greater than 0.85 (the third model required a coefficient of determination greater than or equal to 0.85, which was achieved in practice at 0.89). The output value was the probability of dynamic load distribution anomalies, representing the abnormal situation of dynamic load distribution.

[0038] In practical implementation, based on the analysis results of composite power quality anomalies and dynamic load distribution anomalies, a pre-analysis of the cross-regional coordinated overload risk after different combinations of interconnection switches in the distribution transformer area is conducted; specifically including: Step S209: Construct a cross-regional digital twin model. The cross-regional digital twin model includes transformer parameters (rated capacity, short-circuit impedance, no-load loss), line impedance parameters (resistance and reactance per unit length, length), topology connection relationship (connection table of each switch and node), and phase assignment information of each load node for all participating coordinated regions. Step S210: Enumerate all combination action schemes of tie switches and sectionalizing switches, wherein each combination action scheme of tie switches and sectionalizing switches is used to specify the closed tie switch and the open sectionalizing switch; for example, this embodiment involves 3 tie switches and 5 sectionalizing switches, with a total of 256 combination action schemes. After removing invalid schemes that do not meet the radial constraints, there are 180 valid combination action schemes of tie switches and sectionalizing switches. Step S211: For each combination of tie switch and sectional switch operation scheme, perform the corresponding topology reconstruction operation in the cross-transformer area digital twin model. The predicted active power values ​​of each transformer area and each phase at each time within the future preset time period obtained in step S206 are injected into the corresponding load nodes of the reconstructed network according to the phase. Quasi-steady-state power flow calculation is performed time by time, and the load rate of each transformer and each line at each time is recorded. In a specific embodiment, the specific process of quasi-steady-state power flow calculation is as follows:

[0039] The future time span covered by the load forecast is evenly divided into several time segments according to the same time step (i.e., time interval) as the load forecast, with each segment corresponding to a discrete time point; thus, the continuous future time process is transformed into a batch of steady-state moments to be calculated one by one. For each time segment, the predicted active power and reactive power of each phase in each transformer area on the corresponding segment are obtained, and the reactive power is estimated based on the historical power factor. According to the phase assignment of the corresponding load node in the cross-transformer area digital twin model, the predicted active power and reactive power of each phase in each transformer area on the corresponding segment are converted into a constant power load model and injected into the corresponding node of the network after the topology reconstruction is completed. The active power is directly taken from the predicted value, while the reactive power is estimated based on the power factor of the same period in history. This is the existing technology and will not be described in detail here. The high-voltage side busbar of each transformer area is set as the balancing node, and the corresponding voltage amplitude is defined as the system nominal voltage. In a specific embodiment, since the reconstructed low-voltage distribution network still maintains a radial structure, this embodiment uses the forward-backward substitution method to solve the problem. Specifically, the forward-pushing process starts from the end of the network, accumulates the load power of each node and the loss power of the downstream branch segment by segment, and pushes towards the root node to obtain the power distribution of each branch. The back-substitution process starts from the root node, and uses the known root node voltage and the power of each branch to calculate the voltage drop segment by segment towards the end, and updates the voltage of each node.

[0040] The above forward and backward substitution process is iterated repeatedly until the node voltage difference between two consecutive iterations is less than the preset convergence threshold. Then the power flow calculation is considered to have converged. After the power flow converges, the load rate of each transformer and each line at the corresponding time segment is recorded; further, the load rate of the transformer is the ratio of the apparent power actually passed by the transformer to the rated capacity of the transformer; the load rate of the line is the ratio of the current actually flowing through the corresponding line to the safe current carrying capacity of the line. Extract the load rate of each transformer and each line across all time segments to obtain the load rate sequence of each evaluated device at each discrete moment within a preset future time period; Step S212: For each combined action scheme of tie switch and sectional switch, calculate the cross-area coordinated overload risk of the distribution transformer area under the corresponding combined action scheme based on the quasi-steady-state power flow calculation results and combined with the composite power quality anomaly and dynamic load distribution anomaly. Specifically, for each transformer and each line in each time section, take the load rate of the corresponding evaluated equipment at the corresponding time, subtract the corrected safe load rate threshold, divide the difference by the difference between the value and the safe load rate threshold, and extract the larger value between the division result and 0. Perform a power operation on the larger value by a preset penalty exponent to obtain the single equipment overload risk factor of the corresponding evaluated equipment. The correction process for the revised safe load rate threshold is as follows: Obtain the composite power quality anomaly status and dynamic load distribution anomaly status of each participating coordinated distribution area at the current moment, and take the maximum value of the composite power quality anomaly status and the dynamic load distribution anomaly status as the global power quality deterioration coefficient and the dynamic load distribution anomaly index at the current moment; Subtract the global power quality degradation coefficient at the current moment from the value one to obtain the power quality correction coefficient at the current moment; subtract the dynamic load distribution anomaly index at the current moment from the value one to obtain the load forecast correction coefficient at the current moment.

[0041] Based on the aforementioned power quality correction coefficient and load forecast correction coefficient, the preset safe load rate threshold is dynamically corrected to obtain the corrected safe load rate threshold. Specifically, the power quality correction coefficient and load forecast correction coefficient are weighted and summed to obtain the correction coefficient. The preset safe load rate threshold is multiplied by the correction coefficient to obtain the corrected safe load rate threshold.

[0042] For each time segment, the single equipment overload risk factors of all transformers and lines involved in the evaluation are weighted and summed according to the preset equipment importance weights to obtain the system overload risk value at the corresponding time. The system overload risk values ​​calculated above for all future time segments are compared one by one, and the maximum value is taken as the cross-regional coordinated overload risk of the corresponding interconnection switch combination operation scheme.

[0043] In practical implementation, step S300 combines the topology switch status data and environmental status data of the distribution transformer area to analyze the abnormal risk of loop-closing impact operation of the tie switches in the distribution transformer area; refer to Figure 3 This includes the following specific steps: Step S301: Obtain the real-time voltage amplitude, voltage phase angle difference, loop equivalent resistance, loop equivalent reactance, remaining operating life, and contact resistance of both sides of the tie switch to be analyzed from the topology switch status data; obtain the ambient temperature and ambient humidity at the installation location of the corresponding tie switch from the environmental status data; in this embodiment, the loop equivalent resistance and loop equivalent reactance are the sum of the resistance and reactance of all equipment on the entire closed loop current flow path corresponding to the tie switch to be analyzed; Step S302: Calculate the expected peak value of the closing loop impact current at the moment the tie switch to be analyzed closes, based on the real-time voltage amplitude, voltage phase angle difference, loop equivalent impedance and dynamic transient peak coefficient. The formula for calculating the dynamic transient peak value is: ; In the formula, k is the current real-time dynamic transient peak value coefficient, R is the loop equivalent resistance, and X is the loop equivalent reactance;

[0044] The formula for calculating the expected peak value of the loop impact current is: ; In the formula, Ic is the expected peak value of the closing loop impact current at the moment the tie switch is closed, and Zh is the equivalent impedance of the loop. Specifically, in this embodiment, the equivalent impedance of the loop is the sum of the impedances of all equipment through which the current flows in the entire closed loop after the tie switch is closed. These are the real-time voltage phasors on both sides of the interconnecting switch; specifically, in this embodiment, the real-time voltage phasors... The rectangular coordinate form is: ; In the formula, This represents the real part of the voltage phasor, i.e., the projected component of the voltage onto the horizontal axis, measured in volts. is the imaginary part of the voltage phasor, that is, the projected component of the voltage on the vertical axis, with the unit being volts, and j being the imaginary unit;

[0045] Step S303: Calculate the thermal effect shock risk value based on the expected peak closing-loop impact current, the rated short-time withstand current of the tie switch, and the current carrying capacity derating function determined by ambient temperature. Specifically, obtain the derating coefficient corresponding to the ambient temperature at the tie switch installation location in the current carrying capacity derating function determined by ambient temperature. In a specific embodiment, the current carrying capacity derating function determined by ambient temperature is a dimensionless derating coefficient calculated based on the real-time ambient temperature at the switch installation location, obtained from the temperature derating curve or derating coefficient table provided in the tie switch product manual. Multiply the derating coefficient corresponding to the ambient temperature at the tie switch installation location by the rated short-time withstand current of the tie switch, and divide the expected peak closing-loop impact current by the multiplication result to obtain the thermal effect shock risk value of the tie switch to be analyzed.

[0046] Step S304: Calculate the operating life attenuation risk value based on the remaining operating life and the rated life; specifically, subtract the ratio of the remaining operating life to the rated life from the value to obtain the operating life attenuation risk value of the interconnecting switch to be analyzed.

[0047] Step S305: Take the thermal effect impact risk value, the operating life decay risk value, the ambient temperature and humidity of the corresponding tie switch installation location, and the time elapsed since the last action of the tie switch as input features, input them into the pre-trained multilayer perceptron neural network model, and output the abnormal risk of the loop-closing impact operation of the tie switch to be analyzed in the distribution transformer area.

[0048] The pre-training process of the multilayer perceptron neural network model is as follows: Historical loop-closing operation records of the tie switch are obtained. From these records, the thermal effect impact risk value, operational life decay risk value, ambient temperature, ambient humidity, and the elapsed time since the last operation are extracted as training sample features. A classification label is used to determine whether any of the following events occurred after the loop-closing operation: protection malfunction, switch damage, or voltage drop. A loop-closing impact training set and a loop-closing impact validation set are constructed. A multilayer perceptron neural network model is constructed and trained using the loop-closing impact training set to obtain an initial multilayer perceptron neural network model. This initial model is then validated using the loop-closing impact validation set. An initial multilayer perceptron neural network model with an accuracy greater than or equal to a preset fourth model is used as the final multilayer perceptron neural network model. For example, the training process of the model in this embodiment is as follows: Collect the historical loop closing operation records of the corresponding tie switch over the past 3 years, totaling 500 valid operations. For each operation, record the above 5 feature values ​​before the operation, and whether any one of the following three abnormal events occurred after the operation: protection malfunction (such as instantaneous tripping of overcurrent protection), switch damage (such as contact welding), or voltage drop (amplitude below 0.8pu for more than 20ms). These are used as binary classification labels (1 for the occurrence of any event, 0 otherwise). The dataset was divided into a training set (400 samples) and a validation set (100 samples) in an 8:2 ratio. The model structure consisted of an input layer with 5 neurons, three hidden layers with 128, 64, and 32 neurons respectively, each followed by a ReLU activation function, and a dropout layer with dropout ratios of 0.2, 0.3, and 0.2 respectively. The output layer was a single neuron that used a Sigmoid activation function to output probability values. The loss function is binary cross-entropy, the optimizer is Adam, the initial learning rate is 0.0005, the batch size is 16, the maximum number of iterations is 500, and the early stopping strategy is based on the validation set AUC (area under the curve). If the AUC does not improve for 30 consecutive iterations, the training stops. Furthermore, to address sample imbalance (abnormal events account for about 15%), the positive class weight is set to 5 in the loss function. During training, the validation set accuracy needs to reach the preset fourth model accuracy (set to 0.92 in this embodiment), and the AUC is required to be greater than 0.85. The model accuracy in this embodiment reached 93.1%, and the AUC was 0.89.

[0049] In practical implementation, step S400 assesses the overall cross-regional collaborative overload risk based on the cross-regional transformer area collaborative overload risk analysis results and the abnormal risk analysis results of the loop-closing impact operation of the tie switch, and generates a Pareto-optimal tie switch control scheme set; referring to Figure 4 This includes the following specific steps: Step S401: For each combined action scheme of tie switch and sectional switch, obtain the cross-regional coordinated overload risk calculated by the corresponding combined action scheme, and record it as the first risk value of the corresponding combined action scheme; obtain the maximum value among the abnormal risks of loop impact operation calculated by all action tie switches in the corresponding combined action scheme, and record it as the second risk value of the corresponding combined action scheme; and calculate the switch action attenuation index of the corresponding combined action scheme based on the remaining operating life of each switch that has taken action in the corresponding combined action scheme. Step S402: Using the first risk value, second risk value, and switch action decay index of each combination action scheme of the tie switch and segment switch as three optimization objectives, a fast non-dominated sorting genetic algorithm is used to perform multi-objective optimization search in all combination action schemes to obtain the Pareto front solution set. Each combination action scheme in the Pareto front solution set satisfies the following condition: without worsening at least one objective value, it is impossible to improve any other objective value, thus forming a Pareto optimal tie switch control scheme set. Each combination action scheme in the Pareto optimal tie switch control scheme set includes a specific switch operation sequence, the corresponding first risk value, second risk value, and switch action decay index. For example, this embodiment uses the fast non-dominated sorting genetic algorithm to perform multi-objective optimization search in all effective combination action schemes; the algorithm parameters are set as follows: population size is 100, maximum number of iterations is 200, crossover probability is 0.9, mutation probability is 0.1, simulated binary crossover distribution exponent is 20, and multinomial mutation distribution exponent is 20. The initial population is randomly generated to satisfy the radial topological constraints. In each generation, a non-dominated sort is first performed to stratify individuals according to Pareto level. Within the same stratum, individuals are sorted according to crowding distance to maintain population diversity. After multiple generations of evolution, a Pareto front solution set is obtained. Each scheme in the Pareto front solution set satisfies the following condition: without worsening any other objective value, it is impossible to improve any other objective value. That is, these schemes do not dominate each other and constitute the Pareto optimal tie switch control scheme set.

[0050] Step S403: The first risk value, the second risk value, and the switch action attenuation index of the corresponding combined action scheme in the Pareto optimal interconnection switch control scheme set are weighted and summed to obtain the cross-regional coordinated risk of the corresponding combined action scheme in the Pareto optimal interconnection switch control scheme set. The cross-regional coordinated risk is a continuous value between 0 and 1.

[0051] In its specific implementation, step S500 executes a differentiated coordinated control scheme for the tie switches based on the Pareto-optimal tie switch control scheme set; this includes the following specific steps: Step S501: Extract the cross-regional full-domain collaborative risk of all combined action schemes in the Pareto optimal interconnection switch control scheme set; Step S502: The combined action scheme corresponding to the minimum risk of cross-regional full-area coordination is executed as the coordination control scheme of the interconnection switch.

[0052] It should be noted that the method for determining the weights and thresholds not specifically described in this embodiment is as follows: Electrical operation data, environmental status data, and topology switch status data are synchronously collected during the actual operation of 2000 sets of distribution transformer areas. The cross-area coordinated overload risk, the abnormal risk of loop-closing impact operation of each operating switch, and the attenuation index of the scheme switch operation under each candidate interconnection switch combination scheme are calculated according to the aforementioned steps, and the overall coordinated risk is further calculated. Simultaneously, the judgment results of whether abnormal events such as transformer overload tripping, line overcurrent protection operation, switch contact welding, or voltage drop actually occurred under the corresponding operating conditions are obtained for 2000 sets. The calculation results of various indicators and the corresponding abnormal event judgment results during the operation of 2000 sets are imported into the MATLAB Curve Fitting Toolbox fitting software. Binary logistic regression is used to perform maximum likelihood estimation fitting for each single indicator, and the weights and thresholds of each model are determined with the highest determination coefficient on the validation set as the objective.

[0053] Example 2

[0054] This embodiment provides a distribution transformer area optimization system based on industrial data processing, referring to... Figure 5 ,include: The data acquisition module is used to synchronously collect electrical operation data, environmental status data, and topology switch status data of the distribution transformer area through IoT sensors deployed in the cross-area interconnection network; The switch operation simulation module is used to combine the electrical operation data and environmental status data of the distribution transformer area, and analyze the composite power quality anomaly and dynamic load distribution anomaly of the distribution transformer area based on the cross-transformer area digital twin model; and based on the analysis results of composite power quality anomaly and dynamic load distribution anomaly, to conduct a simulation analysis of the cross-transformer area coordinated overload risk after different combinations of interconnection switch actions in the distribution transformer area. The operation risk analysis module is used to analyze the abnormal risk of loop-closing impact operation of the tie switch in the distribution transformer area by combining the topology switch status data and environmental status data of the distribution transformer area. The candidate scheme output module is used to evaluate the overall risk of cross-transformer area coordination based on the cross-transformer area coordination overload risk analysis results and the tie switch closing impact operation abnormal risk analysis results, and generate a Pareto optimal tie switch control scheme set. The execution module is used to execute differentiated interconnection switch coordinated control schemes based on the Pareto optimal interconnection switch control scheme set.

[0055] The steps for implementing the corresponding functions of each parameter and each unit module in the distribution transformer area optimization system based on industrial data processing of the present invention can be referred to the parameters and steps in the embodiments of the distribution transformer area optimization method based on industrial data processing mentioned above, and will not be repeated here.

[0056] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the embodiments for IoT devices and media are relatively simple in description because they are fundamentally similar to the method embodiments; relevant parts can be referred to the descriptions in the method embodiments.

[0057] The systems, media, and methods provided in the embodiments of the present invention are in one-to-one correspondence. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0058] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0059] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0060] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0061] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0062] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0063] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0064] It should also be noted that the terms include, encompass, or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitations, the inclusion of an element by a statement that includes a… does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0065] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for optimizing distribution transformer substations based on industrial data processing, characterized in that, Includes the following steps: Step S100: Simultaneously collect electrical operation data, environmental status data, and topology switch status data of the distribution transformer area through IoT sensors deployed in the cross-transformer interconnection network; Step S200: Combining the electrical operation data and environmental status data of the distribution transformer area, and based on the cross-distribution area digital twin model, analyze the composite power quality anomaly and dynamic load distribution anomaly of the distribution transformer area; and based on the analysis results of the composite power quality anomaly and dynamic load distribution anomaly, conduct a pre-analysis of the cross-distribution area coordinated overload risk after different combinations of tie switch actions in the distribution transformer area. Step S300: Combine the topology switch status data and environmental status data of the distribution transformer area to analyze the abnormal risk of the closing impact operation of the tie switch in the distribution transformer area. Step S400: Based on the cross-transformer area collaborative overload risk analysis results and the abnormal risk analysis results of the closing impact operation of the tie switch, the cross-transformer area overall collaborative risk is assessed, and a Pareto optimal tie switch control scheme set is generated. Step S500: Based on the Pareto optimal tie switch control scheme set, execute the differentiated tie switch coordinated control scheme.

2. The method for optimizing distribution transformer areas based on industrial data processing according to claim 1, characterized in that, In step S200, the electrical operation data and environmental status data of the distribution transformer area are combined with the cross-transformer area digital twin model to analyze the composite power quality anomalies and dynamic load distribution anomalies of the distribution transformer area; specifically, this includes the following: Step S201: Extract the real-time voltage, real-time current and effective values ​​of each phase of each transformer substation from the electrical operation data, and extract the ambient temperature and ambient humidity from the environmental status data to form a power quality analysis dataset. Step S202: Based on the power quality analysis dataset, calculate the voltage deviation, three-phase unbalance, and harmonic distortion index of the distribution transformer area respectively; Step S203: Combine the voltage deviation index, three-phase unbalance index, and harmonic distortion index of the transformer substation area with the ambient temperature and humidity of the transformer substation area to form a multi-dimensional feature vector of the transformer substation area; input the multi-dimensional feature vector of the transformer substation area into the pre-trained isolated forest anomaly detection model, and the model outputs the composite power quality anomaly status of the transformer substation area.

3. The method for optimizing distribution transformer substations based on industrial data processing according to claim 2, characterized in that, The method combines electrical operation data and environmental status data of distribution transformer areas, and analyzes the combined power quality anomalies and dynamic load distribution anomalies of distribution transformer areas based on a cross-regional digital twin model; specifically, it also includes the following steps: Step S204: Obtain the historical active power sequence of each phase of each transformer area under multiple historical time windows from the electrical operation data, and obtain the temperature and humidity of the corresponding historical time windows from the environmental status data to construct a load prediction dataset; divide the load prediction dataset into a load prediction training set and a load prediction validation set; Step S205: Construct a multi-task long short-term memory network prediction model. Based on the load prediction training set, use the historical active power sequence, temperature, and humidity of each historical time window as input features of the multi-task long short-term memory network prediction model. Use the active power values ​​of each transformer area and each phase at each time within a preset future time period after the corresponding historical time window in the load prediction training set as output targets of the multi-task long short-term memory network prediction model. Train the multi-task long short-term memory network prediction model to obtain an initial load prediction model. Validate the initial load prediction model using the load prediction validation set. Output an initial load prediction model with an accuracy greater than or equal to a preset second model as the load prediction model. Step S206: Input the active power sequence, temperature and humidity of each phase of each transformer area in the current time window into the load prediction model, and output the predicted active power of each phase of each transformer area at each time within a preset time period after the current time window. Step S207: Calculate the prediction coefficient of variation and spatial load imbalance based on the predicted active power value; Step S208: Combine the predicted coefficient of variation, spatial load imbalance degree and the standard deviation of the load rate of each distribution transformer area, input them into the pre-trained gradient boosting decision tree regression model, and output the dynamic load distribution anomaly of the distribution transformer area.

4. The method for optimizing distribution transformer substations based on industrial data processing according to claim 3, characterized in that, Based on the analysis results of composite power quality anomalies and dynamic load distribution anomalies, a pre-analysis of the cross-regional coordinated overload risk after different combinations of interconnection switches in the distribution transformer area is conducted; specifically including: Step S209: Construct a cross-regional digital twin model, which includes transformer parameters, line impedance parameters, topology connections, and phase assignment information of each load node for all participating distribution areas. Step S210: Enumerate all combination action schemes of tie switches and sectionalizing switches, wherein each combination action scheme of tie switches and sectionalizing switches is used to specify the closed tie switch and the open sectionalizing switch. Step S211: For each combination of tie switch and sectional switch operation scheme, perform the corresponding topology reconstruction operation in the cross-transformer area digital twin model. The predicted active power values ​​of each transformer area and each phase at each time within the future preset time period obtained in step S206 are injected into the corresponding load nodes of the reconstructed network according to the phase. Quasi-steady-state power flow calculation is performed time by time, and the load rate of each transformer and each line at each time is recorded. Step S212: For each combined action scheme of tie switch and sectional switch, calculate the cross-area coordinated overload risk of the distribution transformer area under the corresponding combined action scheme based on the quasi-steady-state power flow calculation results and combined with the abnormal power quality and dynamic load distribution conditions.

5. The method for optimizing distribution transformer substations based on industrial data processing according to claim 4, characterized in that, In step S300, the abnormal risk of loop-closing impact operation of the tie switch in the distribution transformer area is analyzed by combining the topology switch status data and environmental status data of the distribution transformer area. The specific steps include the following: Step S301: Obtain the real-time voltage amplitude, voltage phase angle difference, loop equivalent resistance, loop equivalent reactance, remaining operating life and contact resistance of both sides of the tie switch to be analyzed from the topology switch status data; and obtain the ambient temperature and ambient humidity at the corresponding tie switch installation location from the environmental status data. Step S302: Calculate the expected peak value of the closing loop impact current at the moment the tie switch to be analyzed closes, based on the real-time voltage amplitude, voltage phase angle difference, loop equivalent impedance and dynamic transient peak coefficient. Step S303: Calculate the thermal effect impact risk value based on the expected peak value of the closing loop impact current, the rated short-time withstand current of the tie switch, and the current carrying capacity derating function determined by the ambient temperature. Step S304: Calculate the risk value of operating life decay based on the remaining operating life and the rated operating life; Step S305: The thermal effect impact risk value, the operational life attenuation risk value, the ambient temperature and humidity at the installation location of the tie switch to be analyzed, and the time elapsed since the last operation of the tie switch are used as input features and input into a pre-trained multilayer perceptron neural network model. The output is the abnormal risk of the closing impact operation of the tie switch to be analyzed in the distribution transformer area.

6. The method for optimizing distribution transformer areas based on industrial data processing according to claim 5, characterized in that, In step S400, based on the cross-transformer area collaborative overload risk analysis results and the abnormal risk analysis results of the closing impact operation of the tie switch, the cross-transformer area collaborative risk is evaluated, and a Pareto optimal tie switch control scheme set is generated. The specific steps include the following: Step S401: For each combined action scheme of tie switch and sectional switch, obtain the cross-regional coordinated overload risk calculated by the corresponding combined action scheme, and record it as the first risk value of the corresponding combined action scheme. The maximum value among the abnormal risks of the closing impact operation calculated from all action interconnection switches in the corresponding combined action scheme is recorded as the second risk value of the corresponding combined action scheme. And based on the remaining operating life of each switch that has acted in the corresponding combined action scheme, calculate the switch action attenuation index of the corresponding combined action scheme; Step S402: Using the first risk value, second risk value, and switch action decay index of each combination action scheme of the tie switch and segment switch as three optimization objectives, a fast non-dominated sorting genetic algorithm is used to perform multi-objective optimization search in all combination action schemes to obtain the Pareto front solution set, which constitutes the Pareto optimal tie switch control scheme set; each combination action scheme in the Pareto optimal tie switch control scheme set includes a specific switch operation sequence, the corresponding first risk value, second risk value, and switch action decay index; Step S403: The first risk value, the second risk value, and the switch action attenuation index of the corresponding combined action scheme in the Pareto optimal interconnection switch control scheme set are weighted and summed to obtain the cross-regional coordinated risk of the corresponding combined action scheme in the Pareto optimal interconnection switch control scheme set.

7. The method for optimizing distribution transformer substations based on industrial data processing according to claim 6, characterized in that, In step S500, a differentiated coordinated control scheme for the tie switch is executed based on the Pareto optimal tie switch control scheme set. The specific steps include the following: Step S501: Extract the cross-regional full-domain collaborative risk of all combined action schemes in the Pareto optimal interconnection switch control scheme set; Step S502: The combined action scheme corresponding to the minimum risk of cross-regional full-area coordination is executed as the coordination control scheme of the interconnection switch.

8. A distribution transformer area optimization system based on industrial data processing, implemented based on the distribution transformer area optimization method based on industrial data processing as described in any one of claims 1-7, characterized in that, The system includes: The data acquisition module is used to synchronously collect electrical operation data, environmental status data, and topology switch status data of the distribution transformer area through IoT sensors deployed in the cross-area interconnection network; The switch operation simulation module is used to combine the electrical operation data and environmental status data of the distribution transformer area, and analyze the composite power quality anomaly and dynamic load distribution anomaly of the distribution transformer area based on the cross-transformer area digital twin model; and based on the analysis results of composite power quality anomaly and dynamic load distribution anomaly, to conduct a simulation analysis of the cross-transformer area coordinated overload risk after different combinations of interconnection switch actions in the distribution transformer area. The operation risk analysis module is used to analyze the abnormal risk of loop-closing impact operation of the tie switch in the distribution transformer area by combining the topology switch status data and environmental status data of the distribution transformer area. The candidate scheme output module is used to evaluate the overall risk of cross-transformer area coordination based on the cross-transformer area coordination overload risk analysis results and the tie switch closing impact operation abnormal risk analysis results, and generate a Pareto optimal tie switch control scheme set. The execution module is used to execute differentiated interconnection switch coordinated control schemes based on the Pareto optimal interconnection switch control scheme set.

Images