Collaborative control method and system for power distribution equipment based on edge computing

Abstract

This invention relates to the field of intelligent power distribution technology, and particularly to a collaborative control method and system for power distribution equipment based on edge computing. The method includes: acquiring user electricity consumption and temperature / humidity data; generating electricity consumption curves using MATLAB spline curve fitting; constructing environmental vectors; calculating the cosine similarity between environmental vectors to determine an environmental stability index; dividing time windows for environmental stability; calculating the window environmental vector for each time window; classifying electricity consumption curves and analyzing their dispersion; dynamically adjusting the environmental stability index threshold to the optimal value; constructing a training sample set to train a fully connected neural network to obtain an electricity consumption prediction model; predicting future electricity consumption curves in real time and performing power distribution adjustments, thereby optimizing the efficiency and resource utilization of the power distribution system. This invention, by constructing an environmental stability index and dynamically adjusting thresholds, optimizes the electricity consumption prediction model, accurately matches user electricity demand, and simultaneously improves the real-time response capability and resource utilization efficiency of the power distribution system.

Description

Technical Field

[0001] This invention relates to the field of intelligent power distribution technology, and more particularly to a method and system for collaborative control of power distribution equipment based on edge computing. Background Technology

[0002] Distribution equipment management is a key component of the smart grid, responsible for monitoring and controlling various devices in the distribution network, such as transformers, switches, and smart meters, to ensure reliable power supply and efficient transmission. Its main objectives include improving power supply reliability, optimizing power quality, and reducing operating costs. However, distribution equipment management faces numerous challenges, such as handling dynamic electricity consumption behavior of residential users due to changes in temperature and humidity, and fluctuations in the power demand of agricultural equipment caused by changes in field and greenhouse environments. These complex and variable electricity demands place higher demands on the real-time performance and flexibility of distribution equipment management systems.

[0003] By introducing edge computing technology into the field of power distribution equipment management, edge computing reduces the latency of data transmission to the central server and improves the system's response speed and real-time performance by processing and analyzing data on devices at the network edge.

[0004] The existing centralized computing model has significant shortcomings in addressing these challenges. Data transmission latency and uneven distribution of computing resources result in poor real-time response capabilities, making it difficult to quickly adapt to dynamically changing electricity demands. Furthermore, the centralized computing model ignores the sensitivity differences in user environmental factors, limiting the adaptability and accuracy of electricity forecasting models. It struggles to dynamically adjust forecasting models based on actual user needs, failing to effectively meet the personalized electricity demands of different users, thus impacting the overall operational efficiency and rational utilization of resources in the power distribution system. Summary of the Invention

[0005] To address the problems of insufficient real-time response capabilities in existing centralized computing models, which are caused by data transmission delays, uneven distribution of computing resources, neglect of differences in sensitivity to user environmental factors, and difficulty in dynamically adjusting prediction models to meet personalized needs, thus affecting the operating efficiency and rational utilization of power distribution systems, this invention provides solutions in the following aspects.

[0006] In the first aspect, the edge computing-based collaborative control method for power distribution equipment includes: acquiring electricity consumption data and temperature and humidity data of each user based on time series, and using spline curves in MATLAB to perform nonlinear fitting on the electricity consumption data to obtain an electricity consumption curve; combining the temperature and humidity data as environmental vectors for each time moment; calculating the cosine similarity between environmental vectors at each time moment to construct an environmental similarity sequence; determining the environmental stability index based on the similarity calculation results; and judging whether the corresponding time period belongs to a relatively stable environmental time window based on the environmental stability index; calculating the mean of all environmental data within each time window to generate a window environmental vector, and then obtaining the environmental vector for each window. The electricity consumption curves corresponding to the environmental vectors are categorized into sets based on the same window environmental vectors. The deviation of each set of electricity consumption curves is calculated to analyze the dispersion of the sets. The environmental stability index threshold is dynamically adjusted based on the dispersion until a preset condition is met, thereby determining the most suitable environmental stability index threshold for the electricity user. A training sample set is constructed using the window environmental vector, date, and corresponding valid electricity consumption curves. A fully connected neural network is trained using the training sample set to obtain an electricity consumption prediction model. The model is input with a four-dimensional vector in real time to predict the future electricity consumption curves of the users, and real-time power distribution adjustment is performed based on the future electricity consumption curves.

[0007] The benefits are as follows: By collecting user electricity consumption and temperature / humidity data over time, electricity consumption curves are obtained through nonlinear fitting of spline curves using MATLAB, and temperature / humidity data are combined into environmental vectors. An environmental similarity sequence is constructed by calculating the cosine similarity between environmental vectors, and an environmental stability index is determined to divide time windows. Within each window, the mean is calculated to generate a window environmental vector, the corresponding electricity consumption curves are obtained and categorized into sets, the degree of deviation is calculated to analyze dispersion, and the environmental stability index threshold is dynamically adjusted. Finally, a training sample set is constructed to train a fully connected neural network to obtain an electricity consumption prediction model. Real-time input data is used to predict future electricity consumption curves and adjust power distribution, thereby improving the real-time response capability, prediction accuracy, and resource utilization efficiency of the power distribution system.

[0008] Preferably, the environmental stability index includes:

[0009] Taking any time as the target time, the similarity between the environment vector at the target time and the environment vectors at each subsequent time is calculated to obtain the environment similarity sequence at the target time. The mean of all similarity sequences is used as the feature value.

[0010] The ratio between the minimum and maximum feature values ​​in all similarity sequences is used as the environmental stability index at the target time.

[0011] Its effect is that by calculating the average similarity sequence between the environmental vector at the target time and the environmental vector at subsequent times as the feature value, and deriving the environmental stability index based on the ratio of the minimum to the maximum feature value, it can effectively capture the overall stability trend of the environment, reduce the impact of data fluctuations and noise, and thus more accurately reflect the relative stability of the environment and improve the accuracy of macroscopic assessment of environmental changes.

[0012] Preferably, the environmental stability index includes:

[0013] Taking any time as the target time, the similarity between the environment vector of the target time and the environment vectors of each subsequent time is calculated to obtain the environment similarity sequence of the target time.

[0014] The ratio between the minimum and maximum environmental similarity among all environmental similarity sequences is selected as the environmental stability index at the target time.

[0015] Preferably, the step of determining whether a corresponding time period belongs to a relatively stable environmental window based on the environmental stability index includes:

[0016] When the environmental stability index is greater than the preset index threshold, the corresponding time period is determined as a relatively stable time window, and the next moment of the time window is used as the starting point for the next time window division; otherwise, if the index is less than or equal to the preset index threshold, the corresponding time period is determined not to belong to a relatively stable time window, and the environmental stability index is calculated for each subsequent moment for iterative division.

[0017] Its effects are as follows: by monitoring the environmental stability index in real time and dynamically dividing the time window of relative environmental stability, the response speed and adaptability to environmental changes are improved, the accuracy and rationality of the time window division are ensured, and a reliable basis is provided for subsequent power consumption forecasting and power distribution regulation.

[0018] Preferably, the method for calculating the degree of deviation includes:

[0019] Select any electricity consumption curve from the set of electricity consumption curves as the target curve, and calculate the difference between the target curve and the best fitted straight line;

[0020] The difference is integrated within the time window corresponding to the target curve to obtain the total deviation of the target curve from the fitted line within the corresponding time window. The absolute value of the integral is taken as the degree of deviation of the target curve.

[0021] Its effect is that by calculating the deviation of the target electricity consumption curve from the optimal fitted line, the difference between the electricity consumption curve and the overall trend is quantified, abnormal electricity consumption behavior affected by environmental factors is effectively identified, and the accuracy and robustness of the electricity consumption prediction model are improved.

[0022] Preferably, the degree of dispersion includes:

[0023] Calculate the sum of the deviations of each electricity consumption curve in the set of electricity consumption curves, then normalize the sum, and use the normalized result as the degree of dispersion of the set of electricity consumption curves. One degree of dispersion corresponds to one set of electricity consumption curves.

[0024] Preferably, the step of dynamically adjusting the environmental stability index threshold based on the degree of dispersion includes:

[0025] Calculate the mean of the dispersion of all electricity consumption sets, and use the mean as an adjustment value;

[0026] When the adjustment value is greater than the preset adjustment threshold, the environmental stability index threshold is increased by 0.01, and the dispersion calculation process is re-executed until the adjustment value is less than or equal to the preset adjustment threshold. Then, the environmental stability index threshold is determined as the optimal threshold for the electricity user, and an electricity consumption curve is obtained for each window environmental vector.

[0027] Its effect is that by dynamically adjusting the environmental stability index threshold, and by comparing the mean of the dispersion (adjustment value) with the preset threshold, the threshold is gradually optimized to ensure that the environmental vector of each window accurately corresponds to the effective electricity consumption curve, thereby improving the adaptability and accuracy of the electricity consumption prediction model.

[0028] Preferably, the step of training the fully connected neural network includes:

[0029] The four-dimensional vector formed by the combination of window environment vector and date in the training sample set is used as the input feature of the fully connected neural network. The corresponding effective electricity consumption curve is used as the target output label. The training sample set is input into the fully connected neural network. A preset loss function is used to quantify the difference between the predicted output of the fully connected neural network and the actual target output. The network parameters are iteratively updated according to the optimization algorithm. When the loss function is less than the preset value or the preset number of training times is reached, the training is completed and the electricity consumption prediction model is obtained.

[0030] Preferably, the electricity consumption prediction model is constructed based on a fully connected neural network, and its network structure is as follows: the input layer receives a four-dimensional vector consisting of a window environment vector and a date; the hidden layer consists of multiple layers of fully connected neurons, each layer is equipped with an activation function to introduce nonlinear features; and the output layer outputs the predicted electricity consumption curve.

[0031] Secondly, an edge computing-based power distribution equipment collaborative control system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned edge computing-based power distribution equipment collaborative control method is implemented.

[0032] The present invention has the following effects:

[0033] 1. This invention constructs an environmental stability index, selects a relatively stable time window, and dynamically adjusts the threshold of the environmental stability index by combining the calculation of deviation and dispersion, thereby optimizing the training data of the electricity consumption prediction model, improving the prediction accuracy of the model, and making the prediction results more in line with the actual electricity consumption needs of users.

[0034] 2. This invention reduces data transmission delays and uneven distribution of computing resources by collecting and analyzing real-time power consumption and temperature and humidity data, effectively improving the rapid response capability to changes in the power distribution system and ensuring the real-time performance and sensitivity of the power distribution system. Attached Figure Description

[0035] Figure 1 This is a flowchart of steps S1-S4 in the edge computing-based collaborative control method for power distribution equipment according to an embodiment of the present invention.

[0036] Figure 2 This is a schematic diagram of the power consumption prediction model in the edge computing-based collaborative control method for power distribution equipment according to an embodiment of the present invention.

[0037] Figure 3 This is a structural block diagram of a power distribution equipment collaborative control system based on edge computing according to an embodiment of the present invention. Detailed Implementation

[0038] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0039] Reference Figure 1 The edge computing-based collaborative control method for power distribution equipment includes steps S1-S4, as detailed below:

[0040] S1: Obtain electricity consumption data and temperature and humidity data for each user based on time series, and use spline curves in MATLAB to perform nonlinear fitting on the electricity consumption data to obtain the electricity consumption curve. Combine the temperature and humidity data as the environmental vector for each time point.

[0041] It should be noted that, taking a household electricity user as an example, the real-time electricity consumption of the user is collected using an electricity meter (edge ​​device), and the real-time indoor temperature and humidity data are obtained using temperature and humidity sensors deployed indoors, with a collection frequency of once every 30 seconds.

[0042] The environmental vectors include, but are not limited to, temperature and humidity data. Depending on the actual application, other environmental factors, such as rainfall and light intensity, may also need to be considered.

[0043] S2: Calculate the cosine similarity between environmental vectors at each time point, construct an environmental similarity sequence, determine the environmental stability index based on the similarity calculation results, and determine whether the corresponding time period belongs to a relatively stable time window based on the environmental stability index.

[0044] Environmental stability index, including:

[0045] Taking any time as the target time, the similarity between the environment vector at the target time and the environment vectors at each subsequent time is calculated to obtain the environment similarity sequence at the target time. The mean of all similarity sequences is used as the feature value.

[0046] The ratio between the minimum and maximum feature values ​​in all similarity sequences is used as the environmental stability index at the target time.

[0047] Specifically, the environmental stability index satisfies the following polynomial:

[0048] ;

[0049] ;

[0050] ;

[0051] In the formula, Indicates the first Time and the The degree of environmental similarity at any given moment Represents the cosine similarity function. Indicates the first The environment vector at any given time. Indicates the first The environment vector at any given time. Indicates the first A sequence of environmental similarities at different times. Indicates the first Time and the The degree of environmental similarity at any given moment Indicates the first Time to the Environmental stability index at any given time Describes the minimum value function. Represents the maximum value function. Indicates the first The mean of environmental similarity sequences at time points. Indicates the first The mean of environmental similarity sequences at time points. ,and .

[0052] Calculating the average can smooth out short-term environmental fluctuations, making the environmental stability index more stable and preventing individual fluctuations from affecting the overall judgment. For example, when monitoring indoor temperature and humidity, there may be some transient interference factors, such as the opening and closing of doors and windows for a short period of time. These factors may cause temporary abnormal fluctuations in similarity. Calculating the average can reduce the impact of such interference. It can better adapt to the randomness and noise that may exist in environmental data, reduce judgments caused by local abnormal data, and at the same time, reduce the impact of extreme similarity values ​​on the environmental stability index to a certain extent.

[0053] In addition, another embodiment includes:

[0054] Taking any time as the target time, the similarity between the environment vector of the target time and the environment vectors of each subsequent time is calculated to obtain the environment similarity sequence of the target time.

[0055] The ratio between the minimum and maximum environmental similarity among all environmental similarity sequences is selected as the environmental stability index at the target time.

[0056] Specifically, the environmental stability index satisfies the following polynomial:

[0057] ;

[0058] ;

[0059] ;

[0060] In the formula, Indicates the first Time and the The degree of environmental similarity at any given moment Represents the cosine similarity function. Indicates the first The environment vector at any given time. Indicates the first The environment vector at any given time. Indicates the first A sequence of environmental similarities at different times. Indicates the first Time and the The degree of environmental similarity at any given moment Indicates the first Time to the Environmental stability index at any given time Describes the minimum value function. Represents the maximum value function. Indicates the first A sequence of environmental similarities at different times. Indicates the first A sequence of environmental similarities at different times. ,and .

[0061] By calculating the environmental stability index directly using minimum and maximum similarity without calculating the mean, it is more sensitive to minute changes in the environment and can quickly detect extreme environmental changes. This makes it suitable for scenarios requiring timely responses to sudden environmental changes. For example, in experimental environments with strict requirements for environmental change control, timely detection of any environmental mutations that may affect the experiment is crucial.

[0062] When the environmental stability index is greater than the preset index threshold, the corresponding time period is determined as a relatively stable time window, and the next moment of this time window is used as the starting point for the next time window division; otherwise, if it is less than or equal to the preset index threshold, the corresponding time period is determined not to belong to a relatively stable time window, and the environmental stability index is calculated for each subsequent moment for iterative division.

[0063] For example, the preset index threshold is 0.5, which can be adjusted according to specific circumstances. This preset index threshold is calculated by calculating the environmental stability index from historical data and then fine-tuning it in combination with the actual situation.

[0064] Furthermore, when dividing the environmental time window, the environmental stability index threshold used essentially reflects the user's sensitivity to the environment (such as temperature and humidity). Specifically, a higher environmental stability index threshold means that the temperature and humidity data within the environmental time window fluctuate more, which may lead to significant deviations in the subsequent training of the user's electricity consumption prediction model; while setting a lower threshold increases the number of training samples, thereby increasing computational costs. Therefore, it is necessary to dynamically adjust the environmental stability index threshold based on the dispersion of users' electricity consumption under similar environmental conditions, in order to optimize threshold settings, improve prediction accuracy, and balance computational costs.

[0065] S3: Calculate the mean of all environmental data within each time window to generate a window environment vector, and then obtain the electricity consumption curve corresponding to each window environment vector. Based on the same window environment vector, classify the corresponding electricity consumption curves into a set of electricity consumption curves, calculate the deviation of each set of electricity consumption curves to analyze the dispersion of the set of electricity consumption curves, and dynamically adjust the environmental stability index threshold according to the dispersion until the preset conditions are met, thereby determining the most suitable environmental stability index threshold for the electricity user.

[0066] Methods for calculating the degree of deviation include:

[0067] Select any electricity consumption curve from the set of electricity consumption curves as the target curve, and calculate the difference between the target curve and the best fitted straight line;

[0068] The difference is integrated within the time window corresponding to the target curve to obtain the total deviation of the target curve from the fitted line within the corresponding time window. The absolute value of the integral is taken as the degree of deviation of the target curve.

[0069] Specifically, the degree of deviation satisfies the following relationship:

[0070] ;

[0071] In the formula, Represents the first in the set of electricity consumption curves The degree of deviation of the electricity consumption curve, This represents the best-fitting straight line for the set of electricity consumption curves. Represents the first in the set of electricity consumption curves A power consumption curve, Represents the first in the set of electricity consumption curves The electricity consumption curve corresponds to the start time of the environmental time window. Represents the first in the set of electricity consumption curves The electricity consumption curve corresponds to the end time of the environmental time window.

[0072] In other words, it is the absolute value of the total deviation between the target electricity consumption curve and the best-fit straight line. It characterizes the degree of difference between the target curve and the overall electricity consumption trend, reflecting the magnitude of the fluctuation of the target curve relative to the overall trend within that time period. The greater the deviation, the more significant the deviation between the electricity consumption curve and the overall electricity consumption behavior, which may be affected by environmental factors (such as changes in temperature and humidity) or changes in electricity consumption behavior.

[0073] Discreteness, including:

[0074] Calculate the sum of the deviations of each electricity consumption curve in the set of electricity consumption curves, then normalize the sum, and use the normalized result as the degree of dispersion of the set of electricity consumption curves. Each set of electricity consumption curves corresponds to a unique degree of dispersion.

[0075] Calculate the mean of the dispersion of all electricity consumption sets, and use the mean as an adjustment value;

[0076] When the adjustment value is greater than the preset adjustment threshold, the environmental stability index threshold is increased by 0.01, and the dispersion calculation process is re-executed until the adjustment value is less than or equal to the preset adjustment threshold. Then, the environmental stability index threshold is determined as the optimal threshold for the electricity user, and an electricity consumption curve is obtained for each window environmental vector.

[0077] For example, the threshold is adjusted to 0.4, but this can be adjusted according to the actual situation.

[0078] S4: Construct a training sample set using the window environment vector, date, and corresponding effective electricity consumption curve. Use the training sample set to train a fully connected neural network to obtain an electricity consumption prediction model. Real-time acquisition of four-dimensional vector input to the model to predict the future electricity consumption curve of users. Real-time power distribution adjustment is performed based on the future electricity consumption curve.

[0079] The steps for training a fully connected neural network include:

[0080] The four-dimensional vector formed by the combination of window environment vector and date in the training sample set is used as the input feature of the fully connected neural network. The corresponding effective electricity consumption curve is used as the target output label. The training sample set is input into the fully connected neural network. A preset loss function is used to quantify the difference between the predicted output of the fully connected neural network and the actual target output. The network parameters are iteratively updated according to the optimization algorithm. When the loss function is less than the preset value or the preset number of training times is reached, the training is completed and the electricity consumption prediction model is obtained.

[0081] The loss function is the mean squared error function, which satisfies the following relationship:

[0082] ;

[0083] In the formula, Represents the loss function. Indicates the number of samples in the set. Indicates the first The predicted output for each sample, Indicates the first The actual output of each sample.

[0084] The electricity consumption prediction model is built on a fully connected neural network, and its network structure is as follows:

[0085] The input layer consists of four neurons that receive a four-dimensional vector consisting of a window environment vector and a date. The hidden layer consists of multiple fully connected layers, each followed by an activation function to introduce non-linear features. The activation function is the ReLU function. The output layer consists of one neuron that predicts the electricity consumption for a given date and then constructs an electricity consumption curve for each date.

[0086] Reference Figure 2 This is a schematic diagram of the electricity consumption prediction model in an embodiment of this application. In the diagram, A represents the input layer, where... , For the window context vector, , The input is a four-dimensional vector representing the date (month and day) corresponding to the window. B is a hidden layer, consisting of two fully connected layers, used to extract features from the thought vector. C is the output layer, which maps the feature extraction results to electricity consumption. .

[0087] This invention also provides a collaborative control system for power distribution equipment based on edge computing. For example... Figure 3 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the edge computing-based collaborative control method for power distribution equipment according to the first aspect of the present invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface, the setup and functions of which are known in the art and will not be described further here.

[0088] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A collaborative control method for power distribution equipment based on edge computing, characterized in that, include: The power consumption data and temperature and humidity data of each user are obtained based on time series, and the power consumption data are nonlinearly fitted using spline curves in MATLAB to obtain the power consumption curve. The temperature and humidity data are combined as the environmental vectors corresponding to each time point. The cosine similarity between environmental vectors at each time point is calculated to construct an environmental similarity sequence. Then, the environmental stability index is determined based on the similarity calculation results. When the environmental stability index is greater than a preset index threshold, the corresponding time period is determined as a relatively stable time window, and the next time point of the time window is used as the starting point for the next time window division. Conversely, if the index is less than or equal to the preset index threshold, the corresponding time period is determined not to belong to a relatively stable time window, and the environmental stability index is calculated for each subsequent time point for iterative division. The environmental stability index includes: taking any given time as the target time, calculating the similarity between the environmental vector at the target time and the environmental vectors at subsequent times to obtain a corresponding environmental similarity sequence for the target time, and using the mean of all similarity sequences as a feature value; using the ratio between the minimum and maximum feature values ​​among all similarity sequences as the environmental stability index for the target time; or, The environmental stability index includes: taking any time as the target time, calculating the similarity between the environmental vector at the target time and the environmental vectors at subsequent times to obtain the corresponding environmental similarity sequence at the target time; and selecting the ratio between the minimum and maximum environmental similarity among all environmental similarity sequences as the environmental stability index at the target time. The mean of all environmental data within each time window is calculated to generate a window environment vector, and then the electricity consumption curve corresponding to each window environment vector is obtained. Based on the same window environment vector, the corresponding electricity consumption curves are classified into electricity consumption curve sets. The deviation of each electricity consumption curve set is calculated to analyze the dispersion of the electricity consumption curve sets. The environmental stability index threshold is dynamically adjusted based on the dispersion until a preset condition is met, thereby determining the most suitable environmental stability index threshold for the electricity user. The dynamic adjustment of the environmental stability index threshold based on the dispersion includes: calculating the mean of the dispersion of all electricity consumption sets and using the mean as an adjustment value; in response to the adjustment value being greater than the preset adjustment threshold, the environmental stability index threshold is increased by 0.01, and the dispersion calculation process is re-executed until the adjustment value is less than or equal to the preset adjustment threshold, at which point the environmental stability index threshold is determined as the optimal threshold for the electricity user, resulting in an electricity consumption curve corresponding to each window environment vector. A training sample set is constructed using the window environment vector, date, and corresponding effective electricity consumption curve. The fully connected neural network is trained using the training sample set to obtain an electricity consumption prediction model. The four-dimensional vector is collected in real time and input into the model to predict the future electricity consumption curve of the electricity user. Real-time power distribution adjustment is carried out based on the future electricity consumption curve. The loss function used when training a fully connected neural network is as follows: ; In the formula, Represents the loss function. Indicates the number of samples in the set. Indicates the first The predicted output for each sample, Indicates the first The actual output of each sample.

2. The collaborative control method for power distribution equipment based on edge computing according to claim 1, characterized in that, The method for calculating the degree of deviation includes: Select any electricity consumption curve from the set of electricity consumption curves as the target curve, and calculate the difference between the target curve and the best fitted straight line; The difference is integrated within the time window corresponding to the target curve to obtain the total deviation of the target curve from the fitted line within the corresponding time window. The absolute value of the integral is taken as the degree of deviation of the target curve.

3. The collaborative control method for power distribution equipment based on edge computing according to claim 1, characterized in that, The degree of dispersion includes: Calculate the sum of the deviations of each electricity consumption curve in the set of electricity consumption curves, then normalize the sum, and use the normalized result as the degree of dispersion of the set of electricity consumption curves. Each set of electricity consumption curves corresponds to a unique degree of dispersion.

4. The collaborative control method for power distribution equipment based on edge computing according to claim 1, characterized in that, The steps for training the fully connected neural network include: The four-dimensional vector formed by the combination of window environment vector and date in the training sample set is used as the input feature of the fully connected neural network. The corresponding effective electricity consumption curve is used as the target output label. The training sample set is input into the fully connected neural network. A preset loss function is used to quantify the difference between the predicted output of the fully connected neural network and the actual target output. The network parameters are iteratively updated according to the optimization algorithm. When the loss function is less than the preset value or the preset number of training times is reached, the training is completed and the electricity consumption prediction model is obtained.

5. The collaborative control method for power distribution equipment based on edge computing according to claim 1, characterized in that, The electricity consumption prediction model is built on a fully connected neural network, and its network structure is as follows: the input layer receives a four-dimensional vector consisting of a window environment vector and a date; the hidden layer consists of multiple layers of fully connected neurons, each layer is equipped with an activation function to introduce nonlinear features; and the output layer outputs the predicted electricity consumption curve.

6. A collaborative control system for power distribution equipment based on edge computing, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the edge computing-based power distribution equipment collaborative control method according to any one of claims 1-5.

Images