Energy scheduling optimization method and system based on energy internet of things

Abstract

The application provides an energy scheduling optimization method and system based on an energy internet of things, which first acquires a real-time energy monitoring data set and a historical energy scheduling record set in an energy internet of things system, the former containing energy production, consumption and transmission data of a plurality of energy nodes, and the latter containing energy scheduling strategies and execution result data of a historical time period, then performs correlation analysis on the two types of data to generate energy supply and demand correlation characteristics and network load characteristics, then calls a pre-constructed energy scheduling optimization model to perform joint optimization processing to generate a preliminary energy scheduling strategy set, dynamically adjusts the preliminary strategy set based on the network load characteristics to obtain target energy scheduling instructions containing energy distribution parameters and transmission path selection parameters of each energy node, and finally issues the instructions to an energy control terminal and updates the historical energy scheduling record set, thereby realizing optimization of energy scheduling and improving the efficiency and reliability of energy scheduling.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to an energy dispatch optimization method and system based on the Internet of Things for Energy. Background Technology

[0002] In the energy sector, with the rapid development of the Internet of Things (IoT), the scale and complexity of energy systems are increasing daily. Traditional energy dispatching methods mainly rely on human experience and simple rule settings, making it difficult to cope with the multiple nodes, multiple data sources, and dynamically changing energy supply and demand in IoT systems.

[0003] While existing energy dispatching systems can acquire some energy data, they often view real-time energy monitoring data and historical energy dispatching records in isolation, lacking in-depth analysis of the correlation between the two. For example, in terms of the dynamic matching of energy production and consumption, they cannot accurately grasp the production and consumption patterns of different energy nodes at different times, leading to frequent energy supply and demand imbalances. Simultaneously, the load status of energy transmission networks cannot be perceived and analyzed in real time and comprehensively, easily causing transmission path congestion or resource waste. Furthermore, traditional dispatching strategy generation methods are relatively simple, lacking flexibility and adaptability, making it difficult to dynamically adjust dispatching strategies according to real-time changes in network load. This affects the efficiency and reliability of energy dispatching, failing to meet the requirements of efficient and stable operation of energy Internet of Things (IoT) systems. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an energy dispatch optimization method based on the energy Internet of Things, the method comprising: Acquire a set of real-time energy monitoring data and a set of historical energy dispatch records from an energy Internet of Things (IoT) system. The set of real-time energy monitoring data includes energy production data, energy consumption data, and energy transmission data from multiple energy nodes. The set of historical energy dispatch records includes energy dispatch strategies and execution result data for corresponding historical time periods. The real-time energy monitoring data set and the historical energy dispatch record set are subjected to correlation analysis to generate energy supply and demand correlation characteristics and network load characteristics. The energy supply and demand correlation characteristics are used to characterize the dynamic matching relationship between energy production data and energy consumption data, and the network load characteristics are used to characterize the load status of each transmission path in the energy transmission network. The pre-built energy scheduling optimization model is invoked to jointly optimize the energy supply and demand correlation characteristics and the network load characteristics to generate a preliminary energy scheduling strategy set, which includes multiple candidate scheduling schemes and corresponding scheduling priority parameters. Based on the network load characteristics, the preliminary energy scheduling strategy set is dynamically adjusted to obtain the target energy scheduling instruction, which includes the energy allocation parameters and transmission path selection parameters of each energy node. The target energy dispatch instruction is sent to the energy control terminal in the energy Internet of Things system, and the target energy dispatch instruction and the corresponding execution result data are updated to the historical energy dispatch record set.

[0005] In another aspect, embodiments of the present invention also provide an energy dispatch optimization system based on the Internet of Energy, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0006] Based on the above, this invention, by acquiring real-time energy monitoring data sets and historical energy dispatch record sets from the energy Internet of Things (IoT) system, comprehensively integrates real-time information and historical experience data from multiple dimensions such as energy production, consumption, and transmission. It performs correlation analysis on these two types of data to generate energy supply and demand correlation characteristics and network load characteristics, accurately depicting the dynamic matching relationship between energy production and consumption and the load status of the energy transmission network. This enables a deep understanding of the energy system's operational status. A pre-built energy dispatch optimization model is invoked for joint optimization, generating a preliminary energy dispatch strategy set containing multiple candidate dispatch schemes and corresponding dispatch priority parameters. This fully considers the complex relationship between energy supply and demand and network load, providing diverse dispatch options. Based on network load characteristics, the preliminary energy dispatch strategy set is dynamically adjusted to obtain the target energy dispatch instruction. This allows for real-time adaptation to the dynamic changes in the energy system, ensuring the rationality and effectiveness of the dispatch strategy. The target energy dispatch instruction is then sent to the energy control terminal, and the historical energy dispatch record set is updated in a timely manner. This significantly improves the efficiency, accuracy, and reliability of energy dispatch, effectively promoting the efficient and stable operation of the energy IoT system. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the execution flow of the energy dispatch optimization method based on the energy Internet of Things provided in the embodiments of the present invention.

[0008] Figure 2 This is a schematic diagram of exemplary hardware and software components of an energy dispatch optimization system based on the Internet of Energy provided in an embodiment of the present invention. Detailed Implementation

[0009] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1This is a flowchart illustrating an energy dispatch optimization method based on the Internet of Energy (IoT) according to an embodiment of the present invention. The following is a detailed description of the energy dispatch optimization method based on the Internet of Energy.

[0010] Step S110: Obtain the real-time energy monitoring data set and the historical energy dispatch record set from the energy Internet of Things system. The real-time energy monitoring data set includes energy production data, energy consumption data and energy transmission data of multiple energy nodes. The historical energy dispatch record set includes energy dispatch strategies and execution result data for the corresponding historical time period.

[0011] This embodiment uses an energy Internet of Things (IoT) system comprising multiple energy production nodes, energy consumption nodes, and transmission paths connecting them as an example. The system includes energy production nodes such as wind power plants F1 and F2, solar power plants S1 and S2, and thermal power plant H1; energy consumption nodes such as residential areas R1 and R2, industrial parks I1 and I2, and commercial center B1; and transmission paths including L1 connecting F1 and R1, L2 connecting F1 and I1, L3 connecting F2 and R2, L4 connecting F2 and I2, L5 connecting S1 and B1, L6 connecting S1 and I1, L7 connecting S2 and R1, L8 connecting S2 and B1, L9 connecting H1 and R2, and L10 connecting H1 and I2.

[0012] In step S110, the real-time energy monitoring data set is first acquired. This data set is obtained through monitoring equipment deployed at each energy node and along the transmission path. Each energy production node is equipped with a corresponding monitoring device for energy production data. Wind power plants F1 and F2 are equipped with wind speed sensors and power generation metering devices to collect power generation data per unit time in real time; solar power plants S1 and S2 are equipped with light intensity sensors and power generation metering devices to collect power generation data per unit time in real time; and thermal power plant H1 is equipped with fuel consumption metering devices and power generation metering devices to collect power generation data per unit time in real time. The data collection time interval is set to fifteen minutes. At any given moment, the energy production data for F1 is a series of power generation values ​​arranged in chronological order. Similarly, the energy production data for F2, S1, S2, and H1 are also each a series of power generation values ​​arranged in chronological order.

[0013] For energy consumption data, corresponding monitoring equipment was installed at each energy consumption node. Residential areas R1 and R2 are equipped with electricity meters to record electricity consumption per unit time in real time; industrial parks I1 and I2 are equipped with total electricity meters to record electricity consumption per unit time in real time; and commercial center B1 is equipped with electricity meters to record electricity consumption per unit time in real time. The data collection interval is also fifteen minutes. At any given moment, the energy consumption data for R1 is a series of electricity consumption values ​​arranged in chronological order, and the energy consumption data for R2, I1, I2, and B1 are also each a series of electricity consumption values ​​arranged in chronological order.

[0014] For energy transmission data, each transmission path is equipped with a power sensor and a loss monitoring device to collect transmission power and loss data per unit time, with a collection interval of fifteen minutes. At the aforementioned time, the energy transmission data for L1 is a series of transmission power and loss rate values ​​arranged in chronological order, and the energy transmission data for L2 to L10 are also each a series of transmission power and loss rate values ​​arranged in chronological order.

[0015] Next, the historical energy dispatch record set is obtained. This set is stored in the system's database and contains energy dispatch strategies and execution results data over a past period. For example, it includes dispatch strategies formulated over the past six months for different time periods based on energy supply and demand conditions, such as transmitting a portion of F1's power generation to I1 via L2, and transmitting a portion of S1's power generation to B1 via L5, as well as execution results data such as changes in energy supply and demand at each node and load changes along the transmission paths after these dispatch strategies were implemented.

[0016] Step S120: Perform correlation analysis on the real-time energy monitoring data set and the historical energy dispatch record set to generate energy supply and demand correlation features and network load features. The energy supply and demand correlation features are used to characterize the dynamic matching relationship between energy production data and energy consumption data, and the network load features are used to characterize the load status of each transmission path in the energy transmission network.

[0017] In this embodiment, after obtaining the real-time energy monitoring data set and the historical energy dispatch record set, step S120 is performed. First, the energy production data, energy consumption data, and energy transmission data in the real-time energy monitoring data set are correlated with the dispatch strategies and execution result data in the historical energy dispatch record set. By analyzing the inherent relationship between the two, the dynamic matching relationship between energy production and consumption is determined, forming energy supply and demand correlation characteristics; simultaneously, the load situation of the transmission path is analyzed to generate network load characteristics.

[0018] Step S121: Perform timestamp alignment processing on the energy production data, energy consumption data and energy transmission data in the real-time energy monitoring data set to obtain a time-series energy data sequence.

[0019] In this embodiment, since the various types of data in the real-time energy monitoring dataset are collected by different monitoring devices, timestamp inconsistencies may exist. For example, the timestamp of F1's energy production data may differ slightly from the timestamp of R1's energy consumption data. Therefore, in step S121, timestamp alignment processing is required for these data. Specifically, using the system's unified time base as the standard, the timestamps of all data are adjusted to the same time point or time interval. For example, the timestamps of all data are uniformly adjusted to the start time of each fifteen-minute interval, so that energy production data, energy consumption data, and energy transmission data within the same time interval can correspond one-to-one, forming a time-series energy data sequence.

[0020] Step S122: Extract the time series features of energy production data and energy consumption data of each energy node in the time series energy data sequence, and calculate the deviation rate parameter of energy production data and energy consumption data within the same time window.

[0021] In this embodiment, after obtaining the time-series energy data sequence, step S122 is performed. First, the time-series characteristics of energy production data and energy consumption data of each energy node are extracted, and then the deviation rate parameter between the two within the same time window is calculated.

[0022] Step S1221: Perform sliding window partitioning on the time-series energy data sequence to obtain multiple window data units with time continuity. Each window data unit contains energy production data and energy consumption data within a preset time length.

[0023] In this embodiment, the preset time length is set to two hours, and the sliding step size is one hour. The time-series energy data sequence is divided into sliding windows. The first window data unit contains energy production and consumption data for the first two hours starting from the start time; the second window data unit contains energy production and consumption data for the first two hours starting one hour after the start time; and so on, resulting in multiple window data units. For example, the first window contains energy production data for F1, F2, S1, S2, and H1 and energy consumption data for R1, R2, I1, I2, and B1 from 0:00 to 2:00; the second window contains the corresponding data from 1:00 to 3:00, and so on.

[0024] Step S1222: Perform trend fitting processing on the energy production data in each window data unit to generate a production trend curve. The slope of the production trend curve represents the rate of change of the energy production data.

[0025] In this embodiment, for the energy production data in each window data unit, such as the energy production data of F1 from 0:00 to 2:00, a linear fitting method is used for trend fitting. By analyzing the energy production data within this window, a straight line that best fits the trend of these data changes is found, i.e., the production trend curve. The slope of this curve reflects the rate of change of F1 energy production data within this window time; a positive slope indicates that power generation is increasing, and a negative slope indicates that power generation is decreasing. The absolute value of the slope indicates the speed of change. Similarly, trend fitting is performed on the energy production data of F2, S1, S2, and H1 in each window data unit to generate their respective production trend curves.

[0026] Step S1223: Perform periodic component extraction processing on the energy consumption data in each window data unit to generate consumption periodic features. The consumption periodic features include the time interval parameter of the consumption peak occurrence and the peak duration parameter.

[0027] In this embodiment, for the energy consumption data in each window data unit, such as the energy consumption data of R1 from 0:00 to 2:00, periodic component extraction processing is performed. By analyzing the energy consumption data within this window, the periodic variation portion is identified. For example, within this window, the energy consumption of R1 may show two peaks. The time interval between these two peaks and the duration of each peak are recorded. These parameters together constitute the consumption periodic characteristics of R1 within this window. Similarly, periodic component extraction is performed on the energy consumption data of R2, I1, I2, and B1 in each window data unit to generate their respective consumption periodic characteristics.

[0028] Step S1224: Align the production trend curve with the consumption cycle feature on the time axis, and calculate the absolute difference between energy production data and energy consumption data at the same time coordinate point.

[0029] In this embodiment, the production trend curve of the energy production node in each window data unit is aligned with the consumption cycle feature of the corresponding energy consumption node on the time axis to ensure that their time coordinates are consistent. For example, the production trend curve of F1 from 0:00 to 2:00 is aligned with the consumption cycle feature of R1 in the same window on the time axis, so that at each specific time point, such as 0:30, 1:00, etc., there can be corresponding production data on the production trend curve and consumption data in the consumption cycle feature. Then, the absolute difference between the energy production data and the energy consumption data at these same time coordinate points is calculated, that is, the absolute value of production data minus consumption data.

[0030] Step S1225: Calculate the deviation rate parameter based on the absolute difference and the energy consumption data at the corresponding time coordinate point. The deviation rate parameter is the ratio of the absolute difference to the energy consumption data, which is used to characterize the degree of energy supply and demand imbalance.

[0031] In this embodiment, for the absolute difference and energy consumption data at the same time coordinate point, such as the absolute difference between the energy production data of F1 and the energy consumption data of R1 at 0:30 being A, and the energy consumption data of R1 at that time point being B, the deviation rate parameter is the ratio of A to B. The larger this ratio, the more severe the energy supply and demand imbalance between F1 and R1 at that time point; conversely, the smaller the ratio, the more balanced the supply and demand. The above calculation is performed on all time coordinate points in each window data unit to obtain the corresponding deviation rate parameter.

[0032] Step S123: Construct an energy supply and demand correlation matrix based on the deviation rate parameter. The row vectors of the energy supply and demand correlation matrix represent the production capacity characteristics of energy production nodes, the column vectors represent the consumption demand characteristics of energy consumption nodes, and the matrix element values ​​represent the supply and demand matching degree between the corresponding production nodes and consumption nodes.

[0033] In this embodiment, after obtaining the deviation rate parameter, step S123 is performed to construct an energy supply and demand correlation matrix.

[0034] Step S1231: Number the energy production nodes and energy consumption nodes in the energy Internet of Things system respectively to generate a production node index set and a consumption node index set.

[0035] In this embodiment, the energy production nodes F1, F2, S1, S2, and H1 are numbered 1, 2, 3, 4, and 5 respectively, generating a production node index set of {1,2,3,4,5}; the energy consumption nodes R1, R2, I1, I2, and B1 are numbered 1, 2, 3, 4, and 5 respectively, generating a consumption node index set of {1,2,3,4,5}.

[0036] Step S1232: Construct an initial association matrix framework using the production node index set as the row index and the consumption node index set as the column index. The number of rows in the initial association matrix framework is the same as the number of production nodes, and the number of columns is the same as the number of consumption nodes.

[0037] In this embodiment, an initial association matrix framework of 5 rows and 5 columns is constructed based on the production node index set and the consumption node index set. The row indexes correspond to production nodes 1 to 5, and the column indexes correspond to consumption nodes 1 to 5.

[0038] Step S1233: For each combination of production node and consumption node, extract the deviation rate parameter sequence within the corresponding time window, and calculate the mean and variance of the deviation rate parameter sequence.

[0039] In this embodiment, for the combination of production node 1 (F1) and consumption node 1 (R1), the deviation rate parameters in each window data unit are extracted to form a deviation rate parameter sequence. Then, the mean of this sequence is calculated, which is the sum of all deviation rate parameters divided by the number of parameters; the variance of this sequence is calculated, which is the sum of the squares of the differences between each deviation rate parameter and the mean divided by the number of parameters. The same processing is performed on other combinations of production nodes and consumption nodes, such as production node 1 and consumption node 2, production node 2 and consumption node 1, etc., to obtain the mean and variance of their respective deviation rate parameter sequences.

[0040] Step S1234: Normalize the mean and variance to generate standardized deviation feature values, wherein the range of the standardized deviation feature values ​​is a preset interval.

[0041] In this embodiment, the preset interval is set to 0 to 1. For the mean and variance of each combination of production and consumption nodes, a min-max normalization method is used. Specifically, for the mean, the minimum value among all combined means is subtracted, and then divided by the difference between the maximum and minimum values ​​among all combined means to obtain the normalized mean. The variance is normalized using the same method to obtain the normalized variance. Then, the normalized mean and variance are combined to generate a standardized deviation eigenvalue, which falls within the interval of 0 to 1.

[0042] Step S1235: Fill the standardized deviation eigenvalues ​​into the corresponding row and column index positions in the initial correlation matrix framework to obtain the energy supply and demand correlation matrix. The smaller the value in the matrix, the higher the supply and demand matching degree between the corresponding production node and consumption node.

[0043] In this embodiment, the standardized deviation eigenvalues ​​of production node 1 and consumption node 1 are filled into the first row and first column of the initial correlation matrix framework. Similarly, the standardized deviation eigenvalues ​​of production node 1 and consumption node 2 are filled into the first row and second column, and so on, until all positions are filled, resulting in the energy supply and demand correlation matrix. For example, a value of 0.2 in the first row and first column indicates a high degree of supply and demand matching between production node 1 and consumption node 1; a value of 0.7 in the first row and second column indicates a low degree of supply and demand matching between production node 1 and consumption node 2.

[0044] Step S124: Perform association rule mining on the historical energy scheduling strategies and execution result data in the historical energy scheduling record set to extract key load factors affecting energy transmission efficiency. The key load factors include transmission path length parameters, path loss rate parameters, and node connectivity parameters.

[0045] In this embodiment, an association rule mining algorithm is used to process the historical energy dispatch record set. By analyzing the relationship between the selection of transmission paths and the transmission efficiency in the execution results of historical dispatch strategies, key load factors affecting energy transmission efficiency are identified. For example, the analysis reveals that longer transmission paths tend to have lower transmission efficiency; higher path loss rates also lead to lower transmission efficiency; and higher node connectivity (i.e., the more transmission paths a node is connected to) can significantly impact transmission efficiency. Therefore, transmission path length, path loss rate, and node connectivity are identified as key load factors.

[0046] Step S125: Perform feature mapping processing on the energy transmission data in the real-time energy monitoring data set according to the key load factors to generate a network load feature vector characterizing the load status of each transmission path. The dimension of the network load feature vector corresponds to the number of transmission paths in the energy transmission network.

[0047] In this embodiment, after determining the key load factors, feature mapping processing is performed on the energy transmission data in the real-time energy monitoring dataset.

[0048] Step S1251: Perform topology analysis on each transmission path in the energy transmission network to determine the node sequence and connection relationship contained in each transmission path.

[0049] In this embodiment, the topology of transmission paths L1 to L10 is analyzed. For example, L1 connects F1 and R1, and its node sequence is F1-R1, meaning that F1 and R1 are directly connected through L1; L2 connects F1 and I1, and its node sequence is F1-I1, meaning that F1 and I1 are directly connected through L2, and so on, to determine the node sequence and connection relationship of each transmission path.

[0050] Step S1252: Calculate the physical distance feature of each transmission path based on the transmission path length parameter in the key load factors. The physical distance feature is the sum of the distances between each adjacent node in the path.

[0051] In this embodiment, for each transmission path, such as L1, the distance between F1 and R1 is measured, and this distance is the physical distance feature of L1. For a transmission path containing multiple adjacent nodes (if any), the sum of the distances between each adjacent node is calculated as its physical distance feature. For example, if the node sequence of a transmission path is ABC, then its physical distance feature is the sum of the distance from A to B and the distance from B to C.

[0052] Step S1253: Based on the path loss rate parameter in the key load factors and the energy transmission data in the real-time energy monitoring data set, calculate the actual loss rate of each transmission path. The actual loss rate is the ratio of the difference between the energy input of the starting node and the energy received by the ending node of the transmission path to the energy input.

[0053] In this embodiment, for each transmission path, such as L1, the energy input of its starting node F1 and the energy received by its ending node R1 are obtained from real-time energy monitoring data. The difference between the two is calculated, and then the difference is divided by the energy input of the starting node to obtain the actual loss rate of L1. The same calculation is performed for other transmission paths L2 to L10 to obtain their respective actual loss rates.

[0054] Step S1254: Based on the node connectivity parameter in the key load factors, count the number of connections of each node in each transmission path, and calculate the average node connectivity as the path connection complexity feature.

[0055] In this embodiment, for each transmission path, such as L1, it contains nodes F1 and R1. The number of transmission paths connected to F1 is counted; assuming F1 connects to L1 and L2, then the number of connections for F1 is 2. Similarly, the number of transmission paths connected to R1 is counted; assuming R1 connects to L1 and L7, then the number of connections for R1 is 2. The average of these two connection counts, which is 2, is used as the path connection complexity feature of L1. The same process is performed on other transmission paths to obtain their respective path connection complexity features.

[0056] Step S1255: Perform feature fusion processing on the physical distance feature, actual loss rate and path connection complexity feature to generate a path load feature vector containing three dimensions.

[0057] In this embodiment, for each transmission path, its physical distance characteristics, actual loss rate, and path connection complexity characteristics are combined in sequence to form a three-dimensional path load feature vector. For example, if the physical distance characteristic of L1 is D1, the actual loss rate is L1 loss, and the path connection complexity characteristic is C1, then the path load feature vector of L1 is [D1, L1 loss, C1]. Similarly, if the physical distance characteristic of L2 is D2, the actual loss rate is L2 loss, and the path connection complexity characteristic is C2, then the path load feature vector of L2 is [D2, L2 loss, C2], and so on. L3 to L10 each generate their own corresponding three-dimensional path load feature vectors.

[0058] Step S1256: Arrange the path load feature vectors of all transmission paths in a preset order to form a network load feature vector that represents the load status of the entire energy transmission network.

[0059] In this embodiment, the preset order is the transmission path numbering from L1 to L10. The path load feature vectors of L1, L2, ... up to L10 are arranged sequentially to form a 30-dimensional network load feature vector. Each three consecutive dimensions in this vector correspond to the physical distance feature, actual loss rate, and path connection complexity feature of a transmission path, respectively, fully reflecting the load status of each transmission path in the entire energy transmission network.

[0060] Step S130: Call the pre-built energy scheduling optimization model to perform joint optimization processing on the energy supply and demand correlation characteristics and the network load characteristics to generate a preliminary energy scheduling strategy set, which includes multiple candidate scheduling schemes and corresponding scheduling priority parameters.

[0061] In this embodiment, after obtaining the energy supply and demand correlation characteristics and network load characteristics, a pre-built energy scheduling optimization model is invoked for processing. This model is trained based on historical data and can comprehensively consider the energy supply and demand relationship and network load status to generate a reasonable scheduling strategy. By inputting the energy supply and demand correlation characteristics and network load characteristics into the model, and through internal calculations and analysis, a preliminary energy scheduling strategy set containing multiple candidate scheduling schemes and their respective scheduling priority parameters is finally generated.

[0062] Step S131: Input the energy supply and demand correlation features and the network load features into the feature preprocessing layer of the energy scheduling optimization model to obtain a standardized feature vector.

[0063] In this embodiment, the feature preprocessing layer of the energy dispatch optimization model first processes the input energy supply and demand correlation features and network load features. The energy supply and demand correlation feature is a 5x5 matrix, and the network load feature is a 30-dimensional vector. The feature preprocessing layer first expands the energy supply and demand correlation matrix into a 25-dimensional vector, and then concatenates it with the 30-dimensional network load feature vector to form a 55-dimensional initial feature vector. Next, the initial feature vector is standardized by subtracting the mean of each dimension and dividing by the standard deviation of each dimension, so that the feature values ​​of each dimension conform to a distribution with a mean of 0 and a standard deviation of 1, finally obtaining a standardized feature vector.

[0064] Step S132: The energy supply and demand correlation features in the standardized feature vector are processed by the supply and demand prediction module of the energy dispatch optimization model to generate an energy supply and demand prediction curve for a future preset time period.

[0065] In this embodiment, the first twenty-five dimensions of the standardized feature vector correspond to energy supply and demand correlation features. These twenty-five features are input into the supply and demand forecasting module, which performs time-series forecasting. The preset time period is set to the next twenty-four hours. The supply and demand forecasting module contains multiple processing units. First, it performs time-series decomposition on the input features, extracting trend components, periodic components, and random components. Then, it uses linear regression to extend the forecast for the trend components, repeats the forecast based on historical cycle patterns for the periodic components, and estimates the random components based on historical fluctuation ranges. Finally, it synthesizes the forecast results of these three parts to generate an energy supply and demand forecast curve between each energy production node and energy consumption node for the next twenty-four hours. The horizontal axis of the curve represents time, and the vertical axis represents the energy supply and demand.

[0066] Step S1321: Extract sub-vectors corresponding to energy supply and demand correlation features from the standardized feature vector, and determine the energy production-related feature components and energy consumption-related feature components contained therein.

[0067] In this embodiment, sub-vectors corresponding to energy supply and demand correlation features are extracted from the first twenty-five dimensions of the standardized feature vector. These sub-vectors are then analyzed and divided into energy production-related feature components and energy consumption-related feature components based on the source and meaning of the features. For example, features related to the production capacity of production nodes 1 to 5 are classified as energy production-related feature components, while features related to the consumption demand of consumption nodes 1 to 5 are classified as energy consumption-related feature components. Specifically, the column features corresponding to production node 1 and production node 2 are classified as energy production-related feature components; the row features corresponding to consumption node 1 and consumption node 2 are classified as energy consumption-related feature components.

[0068] Step S1322: Input the energy production-related feature components into the production prediction sub-network of the supply and demand prediction module, extract time-series features through the long short-term memory network layer, and generate a production trend prediction sequence.

[0069] In this embodiment, energy production-related feature components are input into a production prediction sub-network. This sub-network contains a Long Short-Term Memory (LSTM) network layer with multiple memory units. First, the energy production-related feature components are sequentially input into each memory unit of the LSM network layer according to time sequence. Each memory unit filters and updates the information based on the currently input features and the state information from the previous time step using a gating mechanism, retaining important temporal features and discarding unimportant information. After processing through multiple time steps, the LSM network layer outputs a vector containing temporal features. This vector is then processed by a fully connected layer and converted into a production trend prediction sequence corresponding to a preset future time period. This production trend prediction sequence reflects the changing trends of energy production at various future times.

[0070] Step S1323: Input the energy consumption-related feature components into the consumption prediction sub-network of the supply and demand prediction module, and capture periodic features through the gated recurrent unit layer to generate a consumption trend prediction sequence.

[0071] In this embodiment, energy consumption-related feature components are input into a consumption prediction subnetwork. This subnetwork contains a gated recurrent unit layer, which consists of multiple gated recurrent units. Energy consumption-related feature components enter the gated recurrent unit layer sequentially, and each gated recurrent unit controls the flow and retention of information through update and reset gates. The update gate determines whether to include the currently input feature information in the unit state, and the reset gate determines whether to ignore past unit states. Through this mechanism, the gated recurrent unit layer can effectively capture the periodic patterns in energy consumption features. After processing, the features output by the gated recurrent unit layer are transformed by a fully connected layer to generate a consumption trend prediction sequence for a future preset time period. This consumption trend prediction sequence reflects the periodic changing trend of energy consumption.

[0072] Step S1324: Perform time axis alignment processing on the production trend prediction sequence and the consumption trend prediction sequence, and perform curve fitting processing on the aligned production trend prediction sequence and consumption trend prediction sequence respectively to generate energy production prediction curve and energy consumption prediction curve for a future preset time period.

[0073] In this embodiment, the time interval between the production trend prediction sequence and the consumption trend prediction sequence is fifteen minutes, consistent with the historical data collection interval. The two sequences are aligned along the time axis to ensure that each time point has a corresponding production and consumption prediction value. Then, a polynomial fitting method is used to curve fit the production trend prediction sequence. An appropriate polynomial order is selected so that the fitted curve closely approximates each data point in the prediction sequence, generating an energy production prediction curve. Similarly, a polynomial fitting method is used to generate an energy consumption prediction curve for the consumption trend prediction sequence. For example, if the data points of the production trend prediction sequence exhibit a quadratic curve trend, a quadratic polynomial is selected for fitting to obtain the corresponding energy production prediction curve.

[0074] Step S1325: Plot the energy production forecast curve and the energy consumption forecast curve in the same coordinate system to obtain an energy supply and demand forecast curve that includes production trends and consumption trends. The intersection of the energy supply and demand forecast curves represents the supply and demand equilibrium point.

[0075] In this embodiment, the energy production forecast curve and the energy consumption forecast curve are plotted on the same coordinate system with time as the horizontal axis and energy quantity as the vertical axis. By observing the trends of the two curves, the relationship between energy production and consumption over a predetermined future time period can be clearly seen. When the two curves intersect, the corresponding time point and energy quantity are the supply-demand equilibrium point, indicating that energy production and consumption have reached a state of equilibrium at that moment. For example, if the energy production forecast curve intersects with the energy consumption forecast curve eight hours in the future, it means that energy supply and demand will reach equilibrium in the next eight hours.

[0076] Step S133: The load assessment module of the energy scheduling optimization model performs load level assessment on the network load features in the standardized feature vector to generate load level labels for each transmission path.

[0077] In this embodiment, the last thirty dimensions of the standardized feature vector correspond to network load features. These thirty features are input into a load assessment module, which contains multiple fully connected layers and a softmax output layer. First, the network load features undergo linear transformation and activation function processing in the first fully connected layer, changing the feature dimensions from thirty to twenty. Next, the twenty-dimensional features are input into a second fully connected layer, further transforming them into ten-dimensional features. Finally, the ten-dimensional features are input into the softmax output layer, which calculates the probability that each transmission path belongs to a different load level, categorizing load levels into light, medium, and heavy load. Based on the principle of maximizing probability, a corresponding load level label is assigned to each transmission path; for example, L1 is labeled light load, L2 is labeled heavy load, and so on.

[0078] Step S134: Input the energy supply and demand forecast curve and the load level label into the strategy generation module of the energy dispatch optimization model, and generate multiple candidate dispatch schemes based on the preset dispatch rule base.

[0079] In this embodiment, the energy supply and demand forecast curve includes the changes in energy supply and demand for each node over the next 24 hours, and the load level label reflects the current load status of each transmission path. These two are input into the strategy generation module, which stores a preset scheduling rule library. The scheduling rule library contains various rules, such as: when the predicted consumption of an energy-consuming node exceeds the predicted production of its corresponding energy-producing node, energy needs to be allocated from other producing nodes; when the load level of a transmission path is overloaded, the transmission volume of that path needs to be reduced or switched to another path, etc. The strategy generation module identifies node pairs with large supply-demand gaps based on the energy supply and demand forecast curve, combines this with the load status of each path in the load level label, and calls corresponding rules from the rule library to generate various candidate scheduling schemes. For example, for an energy gap in R1 in a future period, a scheme can be generated to transmit some energy from F2 to R1 via L3, and simultaneously transmit some energy from S2 to R1 via L7; or a scheme can be generated to transmit energy only from F2 to R1 via L3, etc.

[0080] Step S135: Prioritize the multiple candidate scheduling schemes and calculate the comprehensive evaluation index of each candidate scheduling scheme. The comprehensive evaluation index includes supply and demand matching score, load balancing score and transmission efficiency score.

[0081] In this embodiment, there are ten candidate scheduling schemes. Three scores are calculated for each scheme, and then the scores are combined to obtain a comprehensive evaluation index.

[0082] Step S1351: For each candidate scheduling scheme, extract the energy allocation parameters and transmission path selection parameters contained therein.

[0083] In this embodiment, each candidate scheduling scheme specifies the energy allocation from each energy production node to each energy consumption node, as well as the transmission path used. For example, in a candidate scheme, the energy allocation parameters are the amount of energy allocated from F1 to I1, the amount of energy allocated from S1 to B1, etc.; the transmission path selection parameters are L2 from F1 to I1, L5 from S1 to B1, etc. These specific parameters are extracted from each scheme and used as the basis for subsequent scoring calculations.

[0084] Step S1352: Calculate the supply-demand matching score based on the energy allocation parameters and the supply-demand matching degree parameter in the energy supply-demand correlation characteristics. The supply-demand matching score is positively correlated with the magnitude of the supply-demand matching degree parameter.

[0085] In this embodiment, the supply-demand matching degree parameter in the energy supply-demand correlation feature is the element value in the energy supply-demand correlation matrix; the smaller the value, the higher the matching degree. For each candidate scheduling scheme, its energy allocation parameter is compared with the corresponding supply-demand matching degree parameter. The sum of the products of the allocation parameter and the matching degree parameter is calculated, and then the sum is normalized to fall within the range of 0 to 100, thus obtaining the supply-demand matching degree score. For example, if a scheme allocates more energy to node pairs with high matching degree (small parameter value), the sum of the products is smaller, resulting in a higher score after normalization; conversely, the score is lower.

[0086] Step S1353: Calculate the load balancing score based on the transmission path selection parameters and the load level label in the network load characteristics. The load balancing score is negatively correlated with the variance of the load level of each transmission path.

[0087] In this embodiment, the transmission path selection parameters determine the transmission paths used by each scheme. Based on the load level labels in the network load characteristics, the load level values ​​of these paths are obtained (1 for light load, 2 for medium load, and 3 for heavy load). The variance of these load level values ​​is calculated; the smaller the variance, the more balanced the load distribution across paths. The variance is normalized, and the load balance score is obtained by subtracting the normalized variance value from 100. For example, if a scheme uses paths with load level values ​​of 1, 2, and 1, the variance is small, and the corresponding load balance score is high; if the path load level values ​​are 1, 3, and 3, the variance is large, and the score is low.

[0088] Step S1354: Calculate the transmission efficiency score based on the path length and path loss rate parameters in the transmission path selection parameters. The transmission efficiency score is negatively correlated with the product of the path length and the path loss rate.

[0089] In this embodiment, the transmission path selection parameters include the length and loss rate of the path used. For each candidate scheduling scheme, the product of the length and loss rate of each path is calculated, and the products of all paths are summed to obtain the total product value. The total product value is then normalized, and the transmission efficiency score is obtained by subtracting the normalized total product value from 100. For example, a smaller total product value for the paths used by a scheme indicates that the combined impact of loss and distance during transmission is smaller, resulting in a higher transmission efficiency score; conversely, a larger total product value results in a lower score.

[0090] Step S1355: Perform standardized weighted summation on the supply-demand matching score, load balancing score, and transmission efficiency score to generate a comprehensive evaluation index. The weight parameters in the weighted summation are obtained through training with historical scheduling data.

[0091] In this embodiment, the supply-demand matching score, load balancing score, and transmission efficiency score are first standardized to the range of 0 to 1 by dividing each by 100. The weight parameters are determined through training with historical scheduling data, with the supply-demand matching score having a weight of 0.4, the load balancing score having a weight of 0.3, and the transmission efficiency score having a weight of 0.3. The comprehensive evaluation index is calculated as: supply-demand matching score × 0.4 + load balancing score × 0.3 + transmission efficiency score × 0.3. For example, if a candidate solution has three scores of 80, 70, and 90, after standardization they are 0.8, 0.7, and 0.9, and the comprehensive evaluation index is 0.8.

[0092] Step S1356: Compare the comprehensive evaluation index with the preset scoring threshold, and select candidate scheduling schemes with comprehensive evaluation indices greater than the scoring threshold as effective scheduling schemes.

[0093] In this embodiment, the preset scoring threshold is 0.6. The comprehensive evaluation index of each candidate scheduling scheme is compared with 0.6. If the index is greater than 0.6, the scheme is a valid scheduling scheme; otherwise, it is an invalid scheduling scheme. For example, the scheme with a comprehensive evaluation index of 0.8 is a valid scheduling scheme, while the scheme with a comprehensive evaluation index of 0.5 is an invalid scheduling scheme and is excluded from the initial energy scheduling strategy set.

[0094] Step S136: Sort the candidate scheduling schemes according to the magnitude of the comprehensive evaluation index to generate a preliminary energy scheduling strategy set containing scheduling priority parameters, wherein the scheduling priority parameters are positively correlated with the magnitude of the comprehensive evaluation index.

[0095] In this embodiment, the selected effective scheduling schemes are sorted in descending order of their comprehensive evaluation indicators. The scheme with the highest comprehensive evaluation indicator is ranked first and has the highest priority; the scheme with the second highest indicator is ranked second and has the next highest priority, and so on. Each scheme is assigned a corresponding scheduling priority parameter, the value of which corresponds to its ranking position. For example, the parameter for the first scheme is 1, the parameter for the second scheme is 2, and so on. The smaller the parameter value, the higher the priority. These ranked schemes and their priority parameters are combined to form a preliminary energy scheduling strategy set.

[0096] Step S140: Based on the network load characteristics, dynamically adjust the preliminary energy scheduling strategy set to obtain the target energy scheduling instruction, which includes the energy allocation parameters and transmission path selection parameters of each energy node.

[0097] In this embodiment, although the schemes in the initial energy scheduling strategy set have been prioritized, some transmission paths may still be overloaded. These schemes are dynamically adjusted based on network load characteristics to ensure that the load on all transmission paths is within a reasonable range, ultimately resulting in an executable target energy scheduling instruction.

[0098] Step S141: Extract the candidate scheduling scheme with the highest scheduling priority from the preliminary energy scheduling strategy set as the benchmark scheduling scheme.

[0099] In this embodiment, the initial energy scheduling strategy set is sorted from highest to lowest priority, and the candidate scheduling scheme ranked first is the scheme with the highest scheduling priority. This scheme is extracted as the baseline scheduling scheme. For example, if the highest priority scheme in the initial set is scheme A, then scheme A is used as the baseline scheduling scheme.

[0100] Step S142: Analyze the energy allocation parameters and transmission path selection parameters in the baseline scheduling scheme to determine the initial allocation amount and initial transmission path of each energy node.

[0101] In this embodiment, the baseline scheduling scheme is analyzed to clarify the energy allocation parameters and transmission path selection parameters. The energy allocation parameters include the amount of energy allocated by each energy-producing node to each energy-consuming node; for example, F1 allocates energy A1 to I1, and S1 allocates energy A2 to B1, etc. The transmission path selection parameters include the transmission path corresponding to each energy allocation; for example, the transmission from F1 to I1 uses L2, and the transmission from S1 to B1 uses L5, etc. Based on these parameters, the initial allocation amount and initial transmission path for each energy node are determined.

[0102] Step S143: Based on the path load feature vector in the network load characteristics, identify the target path in the initial transmission path whose load level exceeds a preset threshold.

[0103] In this embodiment, the path load feature vector in the network load characteristics contains the load information of each transmission path. Combined with the load level label generated by the load assessment module, the load status of the initial transmission path is determined. The preset threshold is set to medium load, that is, a path with a load level of "heavy load" is considered to exceed the threshold. For example, if the L2 load level in the initial transmission path is "heavy load" and exceeds the preset threshold, then L2 is identified as the target path.

[0104] Step S144: Perform path replacement processing on the target path, and select alternative paths with load levels lower than a preset threshold from the energy transmission network. The alternative paths have the same start and end nodes as the target path.

[0105] In this embodiment, for the target path L2, its starting node is F1 and its ending node is I1. All transmission paths with starting node F1 and ending node I1 are searched from the topology of the energy transmission network. Assume that besides L2, there is another transmission path L11, whose node sequence is F1 through intermediate node M1 to I1, i.e., F1-M1-I1. Next, the load level data of L11 is obtained. This load level data is calculated based on the path load characteristics corresponding to the previously generated network load feature vector. Assume the preset threshold is 0.6, and the load level of L11 is 0.4, which is lower than the preset threshold; therefore, L11 meets the conditions for a candidate path.

[0106] For example, step S1441: Perform topology modeling on the energy transmission network to construct a network topology graph containing all nodes and transmission paths, where nodes represent energy nodes, edges represent transmission paths, and the weight of the edges represents the path load level.

[0107] In this embodiment, all nodes in the energy transmission network (including energy production nodes F1, F2, S1, S2, H1, energy consumption nodes R1, R2, I1, I2, B1, and intermediate node M1, etc.) and transmission paths (L1 to L11, etc.) are abstractly modeled. In the constructed network topology graph, each node is represented by a circle symbol, edges are represented by line segments connecting nodes, and the weight of each edge is labeled with the corresponding path load level. For example, the weight of the edge corresponding to L2 is 0.7 (higher than the preset threshold of 0.6), and the weight of the edge corresponding to L11 is 0.4 (lower than the preset threshold of 0.6).

[0108] Step S1442: Using the starting node of the target path as the source node and the ending node as the target node, execute the shortest path search algorithm in the network topology graph to find all possible paths from the source node to the target node.

[0109] In this embodiment, F1 is the source node and I1 is the target node. A breadth-first search algorithm is executed in the network topology graph to find all possible paths. First, starting from F1, the nodes directly connected to it are traversed, revealing nodes M1 and I1 (via the first segment of L11 and L2, respectively). For node M1, the nodes directly connected to it are traversed again, revealing I1 (via the second segment of L11), thus obtaining the path F1-M1-I1 (i.e., L11). For the directly connected I1, the path F1-I1 (i.e., L2) is obtained. Therefore, all possible paths from F1 to I1 are L2 and L11.

[0110] Step S1443: Calculate the sum of load levels for each possible path, where the sum of load levels is the sum of the load levels of each transmission path in the path.

[0111] In this embodiment, for path L2, since it is a transmission path directly connecting F1 and I1, its total load level is its own load level of 0.7. For path L11, it consists of two transmission paths, namely the path from F1 to M1 and the path from M1 to I1. Assuming that the load levels of these two paths are 0.2 and 0.2 respectively, then the total load level of L11 is the sum of 0.2 and 0.2, which is 0.4.

[0112] Step S1444: Select possible paths whose total load level is lower than a preset threshold as a set of candidate paths.

[0113] In this embodiment, the preset threshold is 0.6. The total load level of path L2 is 0.7, which is higher than the preset threshold, so it is excluded; the total load level of path L11 is 0.4, which is lower than the preset threshold, so it is selected into the candidate path set. At this time, the candidate path set only contains L11.

[0114] Step S1445: Evaluate the transmission efficiency of the candidate paths in the candidate path set, and calculate the transmission efficiency parameter of each candidate path. The transmission efficiency parameter is negatively correlated with the product of the path length and the path loss rate.

[0115] In this embodiment, for the candidate path L11, its path length and path loss rate are first obtained. It is assumed that the total length of L11 is the sum of the lengths from path F1 to M1 and from M1 to I1, and the path loss rate is the average of the path loss rates of the two segments. Then, the product of the path length and the path loss rate is calculated, assuming this product is P11. The transmission efficiency parameter is calculated by dividing a fixed baseline value by this product. Since a larger product results in a smaller transmission efficiency parameter, the transmission efficiency parameter is negatively correlated with this product. Therefore, the calculated transmission efficiency parameter for L11 is E11.

[0116] Step S1446: Sort the set of candidate paths according to the transmission efficiency parameter, and select the candidate path with the highest transmission efficiency parameter as the replacement path.

[0117] In this embodiment, only L11 is available in the candidate path set, so L11 is directly selected as the replacement path. If there are multiple candidate paths, such as another path L12 with a transmission efficiency parameter of E12, and E11 is greater than E12, then L11 is selected as the replacement path.

[0118] Step S1447: Assign the energy transmission task corresponding to the target path to the replacement path and update the transmission path selection parameters.

[0119] In this embodiment, the energy transmission task from F1 to I1, originally handled by the target path L2, is now assigned to the alternative path L11. Simultaneously, the portion of the transmission path selection parameters involving L2 is updated to L11, ensuring that the transmission path recorded by the system is the currently used alternative path.

[0120] Step S145: Based on the load level and transmission efficiency parameters of the candidate paths, adjust the energy allocation parameters of the corresponding energy nodes, repeat the path replacement and parameter adjustment steps until the load level of all transmission paths is lower than the preset threshold, and generate the adjusted energy scheduling scheme.

[0121] In this embodiment, the load level of the alternative path L11 is 0.4, and the transmission efficiency parameter is E11. Since the energy transmission task of L2 is allocated to L11, the energy allocation parameter of F1 needs to be adjusted. The energy originally allocated to L2 for transmission to I1 is now transmitted through L11. Therefore, it is necessary to ensure that the energy allocated by F1 to L11 is consistent with the energy previously allocated to L2 to meet the energy demand of I1. After the adjustment, the load levels of all transmission paths are rechecked. At this time, the load level of L2 will decrease because it no longer undertakes transmission tasks, assuming it drops to 0.3. The load level of L11 may increase due to the addition of transmission tasks, assuming it rises to 0.5, both below the preset threshold of 0.6. At this point, the load levels of all transmission paths meet the requirements, generating an adjusted energy scheduling scheme. This adjusted energy scheduling scheme clearly defines the energy allocation for each energy node and the corresponding transmission path.

[0122] Step S146: Format the energy allocation parameters and transmission path selection parameters in the adjusted energy dispatch scheme to generate the target energy dispatch instruction.

[0123] In this embodiment, the adjusted energy dispatching scheme includes energy allocation parameters such as the amount of energy allocated by F1 to L11 for transmission to I1, and the amount of energy allocated by F1 to L1 for transmission to R1. Transmission path selection parameters include selecting L11 from F1 to I1 and selecting L1 from F1 to R1. These parameters are formatted according to a system-preset instruction format, organizing the energy allocation parameters and transmission path selection parameters into structured data, such as XML, which includes fields such as node identifier, energy allocation amount, and transmission path identifier, ultimately generating the target energy dispatching instruction.

[0124] Step S150: Send the target energy dispatch instruction to the energy control terminal in the energy Internet of Things system, and update the target energy dispatch instruction and the corresponding execution result data to the historical energy dispatch record set.

[0125] In this embodiment, the generated target energy dispatching command is sent to various energy control terminals, such as the power generation control terminal of F1, the power consumption control terminal of I1, and the transmission control terminal of L11, through the communication module of the energy Internet of Things system. Upon receiving the command, the energy control terminal executes the corresponding operations according to the requirements of the command, such as F1 transmitting energy to L11 according to the specified energy allocation, and L11 performing energy transmission regulation according to the command. During the command execution process, the system collects execution result data in real time, including the actual energy production and consumption of each node and the actual load changes of the transmission path. After the command execution is completed, the target energy dispatching command and the corresponding execution result data are stored in the system's database, updating the historical energy dispatching record set.

[0126] Furthermore, the method also includes: Step S210: Construct an energy dispatch optimization model, which includes a feature preprocessing layer, a supply and demand forecasting module, a load assessment module, and a strategy generation module.

[0127] In this embodiment, a feature preprocessing layer is first constructed. This layer adopts a fully connected neural network structure, including an input layer, hidden layers, and an output layer. The number of neurons in the input layer is consistent with the total dimension of the energy supply and demand correlation features and the network load features. The hidden layer contains multiple neurons used to perform nonlinear transformations on the input features. The output layer outputs a standardized feature vector. Next, a supply and demand prediction module is constructed. This module includes a production prediction sub-network and a consumption prediction sub-network. The production prediction sub-network adopts a Long Short-Term Memory (LSTM) network, including an input layer, an LSTM layer, and an output layer. The LSTM layer is used to capture the temporal features of energy production data. The consumption prediction sub-network adopts a gated recurrent unit (GRU) network, including an input layer, a GRU layer, and an output layer. The GRU layer is used to capture the periodic features of energy consumption data. Then, a load assessment module is constructed. This module adopts a multilayer perceptron structure. The input is the network load features. After processing by multiple hidden layers, the load level label of each transmission path is output. Finally, a strategy generation module is constructed, which combines decision trees and reinforcement learning algorithms. The decision trees are used to generate preliminary scheduling rules based on historical data, and the reinforcement learning algorithm is used to optimize the scheduling strategy during the actual scheduling process. The modules are connected through data interfaces to form a complete energy scheduling optimization model.

[0128] Step S211: Collect historical energy data, which includes historical energy production data, historical energy consumption data, historical energy transmission data, and corresponding historical scheduling strategies and execution result data.

[0129] In this embodiment, historical energy data from the past year is extracted. Historical energy production data includes the power generation of F1, F2, S1, S2, and H1 in different time periods; historical energy consumption data includes the power consumption of R1, R2, I1, I2, and B1 in different time periods; historical energy transmission data includes the transmission power and loss rate of each transmission path; historical scheduling strategies include various energy allocation and transmission path selection schemes formulated in the past; and execution result data includes the energy supply and demand balance and transmission efficiency after the execution of the corresponding scheduling strategy.

[0130] Step S212: Preprocess the historical energy data, including data cleaning, missing value filling and outlier handling, to obtain preprocessed historical energy data.

[0131] In this embodiment, during the data cleaning process, duplicate data records are removed, such as energy data collected repeatedly at the same time point and node. For missing values, interpolation is used to fill them in. For example, if the power generation data for a certain time period F1 is missing, linear interpolation is performed based on the power generation data before and after that time period to obtain the filled power generation data. For outliers, the standard deviation of the data is calculated, and data that deviates from the average by more than three times the standard deviation is identified as outliers and replaced with the average of the normal data before and after the outlier.

[0132] Step S213: Divide the preprocessed historical energy data into a training set, a validation set, and a test set. The training set is used for model training, the validation set is used to adjust model parameters, and the test set is used to evaluate model performance.

[0133] In this embodiment, the preprocessed historical energy data is divided in a 7:2:1 ratio, with 70% of the data used as the training set, 20% as the validation set, and 10% as the test set. For example, the data from the first 8.4 months of the past year is used as the training set, the next 2.4 months as the validation set, and the last 1.2 months as the test set.

[0134] Step S214: Set the initial parameters of the model, including the learning rate, number of iterations, number of hidden layer neurons, etc. for each module.

[0135] In this embodiment, the learning rate of the feature preprocessing layer is set to a small value, and the number of neurons in the hidden layer is determined according to the dimension of the input features. In the supply and demand prediction module, the learning rate of the long short-term memory network and the gated recurrent unit network is set to a value suitable for training time-series data, and the number of iterations is set to a large value to ensure that the model is fully trained. The initial parameters of the load assessment module and the policy generation module are set with reference to the empirical values ​​in the relevant fields.

[0136] Step S215: Train the energy dispatch optimization model using the training set, and adjust the model parameters through the backpropagation algorithm to continuously reduce the prediction error of the model.

[0137] In this embodiment, the energy supply and demand correlation features and network load features from the training set are input into the energy scheduling optimization model. The energy scheduling optimization model outputs the predicted energy supply and demand curve, load level label, and scheduling strategy. The prediction results are compared with the actual results in the training set, the prediction error is calculated, and the error is propagated from the output layer to the input layer using the backpropagation algorithm. The weights and biases of neurons in each layer are adjusted, and this process is repeated until the prediction error of the energy scheduling optimization model reaches a preset minimum value.

[0138] Step S216: Validate the trained model using a validation set, and adjust the model's hyperparameters, such as the learning rate and number of iterations, based on the validation results.

[0139] In this embodiment, data from the validation set is input into the trained energy dispatch optimization model, and the predictive performance of the energy dispatch optimization model is observed. If the prediction error of the energy dispatch optimization model on the validation set is large, it indicates that the energy dispatch optimization model may be overfitting or underfitting. At this time, the learning rate is adjusted, such as increasing the learning rate to speed up model convergence or decreasing the learning rate to improve model accuracy. At the same time, the number of iterations is adjusted until the performance of the energy dispatch optimization model on the validation set reaches its optimum.

[0140] Step S217: Use the test set to evaluate the performance of the adjusted model, calculate the model's accuracy, recall and other metrics, and ensure that the model meets the preset performance requirements.

[0141] In this embodiment, data from the test set is input into the adjusted energy dispatch optimization model to obtain the prediction results of the energy dispatch optimization model. Indicators such as the degree of agreement between the predicted energy supply and demand curves and the actual curves, the accuracy of load level labels, and the execution effect of dispatch strategies are calculated. If all indicators reach preset thresholds, such as an accuracy exceeding 90%, it indicates that the model performance meets the requirements; otherwise, the process returns to step S215 to retrain the model.

[0142] Step S310: Regularly calibrate the monitoring equipment in the energy Internet of Things system to ensure the accuracy of the collected real-time energy monitoring data.

[0143] In this embodiment, the system's sensors, smart meters, and other monitoring devices are calibrated monthly. For example, the F1 power generation metering device is tested using a standard power supply. The device's measured value is compared with the actual output value of the standard power supply. If the difference exceeds the allowable range, the device's parameters are adjusted until the measured value is within the allowable error range. The power sensors along the transmission path are calibrated using the same method to ensure the accuracy and reliability of the collected transmission power data.

[0144] Step S311: Regularly back up and clean up the historical energy dispatch record set, delete invalid historical data, and retain valuable dispatch records.

[0145] In this embodiment, the historical energy dispatch record set is backed up every three months, and the backup data is stored on a separate storage device to prevent data loss. Simultaneously, invalid historical data, such as incomplete dispatch strategies or data with abnormal and unrecoverable execution results, is cleaned up. Valuable dispatch records are then organized chronologically to facilitate subsequent data analysis and model training.

[0146] Step S312: Update the energy dispatch optimization model regularly and adjust the model parameters based on new historical data and actual operating conditions.

[0147] In this embodiment, the energy dispatch optimization model is updated every six months using newly generated historical data. New historical data is added to the training set, and the model is retrained according to steps S215 to S217, adjusting parameters such as weights and biases. Simultaneously, based on the actual operation of the system, such as changes in the energy transmission network topology or the addition or removal of energy nodes, the model's input features and structure are adjusted accordingly to ensure the model can adapt to the new operating environment and improve the accuracy of dispatch optimization.

[0148] Figure 2 The illustration shows exemplary hardware and software components of an energy dispatch optimization system 100 based on the Internet of Energy (IoT) that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the energy dispatch optimization system 100 based on the Internet of Energy (IoT) and to perform the functions described in this application.

[0149] The energy dispatch optimization system 100 based on the energy Internet of Things can be a general-purpose server or a special-purpose server; both can be used to implement the energy dispatch optimization method based on the energy Internet of Things of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0150] For example, the energy dispatch optimization system 100 based on the Internet of Energy (IoT) may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the energy dispatch optimization system 100 based on the IoT may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The energy dispatch optimization system 100 based on the IoT also includes an I / O interface 150 between the computer and other input / output devices.

[0151] For ease of explanation, only one processor is described in the energy dispatch optimization system 100 based on the energy Internet of Things. However, it should be noted that the energy dispatch optimization system 100 based on the energy Internet of Things in this application may also include multiple processors. Therefore, the steps performed by one processor as described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the energy dispatch optimization system 100 based on the energy Internet of Things performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0152] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the energy scheduling optimization method based on the energy Internet of Things is implemented as described above.

[0153] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. An energy dispatch optimization method based on the Internet of Things for Energy, characterized in that, The method includes: Acquire a set of real-time energy monitoring data and a set of historical energy dispatch records from an energy Internet of Things (IoT) system. The set of real-time energy monitoring data includes energy production data, energy consumption data, and energy transmission data from multiple energy nodes. The set of historical energy dispatch records includes energy dispatch strategies and execution result data for corresponding historical time periods. The real-time energy monitoring data set and the historical energy dispatch record set are subjected to correlation analysis to generate energy supply and demand correlation characteristics and network load characteristics. The energy supply and demand correlation characteristics are used to characterize the dynamic matching relationship between energy production data and energy consumption data, and the network load characteristics are used to characterize the load status of each transmission path in the energy transmission network. The pre-built energy scheduling optimization model is invoked to jointly optimize the energy supply and demand correlation characteristics and the network load characteristics to generate a preliminary energy scheduling strategy set, which includes multiple candidate scheduling schemes and corresponding scheduling priority parameters. Based on the network load characteristics, the preliminary energy scheduling strategy set is dynamically adjusted to obtain the target energy scheduling instruction, which includes the energy allocation parameters and transmission path selection parameters of each energy node. The target energy dispatch instruction is sent to the energy control terminal in the energy Internet of Things system, and the target energy dispatch instruction and the corresponding execution result data are updated to the historical energy dispatch record set. The process of calling a pre-built energy scheduling optimization model to jointly optimize the energy supply and demand correlation characteristics and the network load characteristics generates a preliminary set of energy scheduling strategies, including: The energy supply and demand correlation features and the network load features are input into the feature preprocessing layer of the energy scheduling optimization model to obtain a standardized feature vector. The energy supply and demand correlation features in the standardized feature vector are processed by the supply and demand prediction module of the energy dispatch optimization model to generate an energy supply and demand prediction curve for a future preset time period. The load assessment module of the energy dispatch optimization model performs load level assessment on the network load features in the standardized feature vector to generate load level labels for each transmission path. The energy supply and demand forecast curve and the load level label are input into the strategy generation module of the energy dispatch optimization model, and multiple candidate dispatch schemes are generated based on the preset dispatch rule base. The multiple candidate scheduling schemes are prioritized and sorted, and a comprehensive evaluation index for each candidate scheduling scheme is calculated. The comprehensive evaluation index includes supply and demand matching score, load balancing score and transmission efficiency score. Candidate scheduling schemes are sorted according to the magnitude of the comprehensive evaluation index to generate a preliminary energy scheduling strategy set containing scheduling priority parameters, wherein the scheduling priority parameters are positively correlated with the magnitude of the comprehensive evaluation index. The step of using the supply and demand forecasting module of the energy dispatch optimization model to perform time-series forecasting of the energy supply and demand correlation features in the standardized feature vector to generate an energy supply and demand forecast curve for a future preset time period includes: Extract sub-vectors corresponding to energy supply and demand correlation features from the standardized feature vectors, and determine the energy production-related feature components and energy consumption-related feature components contained therein; The energy production-related feature components are input into the production prediction sub-network of the supply and demand prediction module, and time-series features are extracted through the long short-term memory network layer to generate a production trend prediction sequence. The energy consumption-related feature components are input into the consumption prediction sub-network of the supply and demand prediction module, and the periodic features are captured through the gated recurrent unit layer to generate a consumption trend prediction sequence. The production trend prediction sequence and the consumption trend prediction sequence are aligned on the time axis. The aligned production trend prediction sequence and the consumption trend prediction sequence are then subjected to curve fitting to generate an energy production prediction curve and an energy consumption prediction curve for a future preset time period. The energy production forecast curve and the energy consumption forecast curve are plotted on the same coordinate system to obtain an energy supply and demand forecast curve that includes production trends and consumption trends. The intersection of the energy supply and demand forecast curves represents the supply and demand equilibrium point. The step of dynamically adjusting the preliminary energy scheduling strategy set based on the network load characteristics to obtain the target energy scheduling instruction includes: The candidate scheduling scheme with the highest scheduling priority is extracted from the preliminary energy scheduling strategy set and used as the benchmark scheduling scheme. The energy allocation parameters and transmission path selection parameters in the baseline scheduling scheme are analyzed to determine the initial allocation amount and initial transmission path of each energy node. Based on the path load feature vector in the network load characteristics, identify the target path in the initial transmission path whose load level exceeds a preset threshold. The target path is replaced by selecting alternative paths from the energy transmission network with load levels lower than a preset threshold. The alternative paths have the same start and end nodes as the target path. Based on the load level and transmission efficiency parameters of the alternative paths, adjust the energy allocation parameters of the corresponding energy nodes, repeat the path replacement and parameter adjustment steps until the load level of all transmission paths is lower than the preset threshold, and generate the adjusted energy scheduling scheme. The energy allocation parameters and transmission path selection parameters in the adjusted energy dispatch scheme are formatted to generate the target energy dispatch instruction. The target path is subjected to path replacement processing, and alternative paths with load levels below a preset threshold are selected from the energy transmission network, including: The energy transmission network is modeled to construct a network topology graph containing all nodes and transmission paths, where nodes represent energy nodes, edges represent transmission paths, and the weight of an edge represents the path load level. Using the starting node of the target path as the source node and the ending node as the target node, the shortest path search algorithm is executed in the network topology graph to find all possible paths from the source node to the target node. Calculate the sum of load levels for each possible path, where the sum of load levels is the sum of the load levels of each transmission path in the path; Filter out possible paths whose total load level is lower than a preset threshold as a set of candidate paths; The transmission efficiency of the candidate paths in the candidate path set is evaluated, and the transmission efficiency parameter of each candidate path is calculated. The transmission efficiency parameter is negatively correlated with the product of the path length and the path loss rate. The candidate path set is sorted according to the transmission efficiency parameter, and the candidate path with the highest transmission efficiency parameter is selected as the replacement path. Assign the energy transmission task corresponding to the target path to the replacement path and update the transmission path selection parameters.

2. The energy dispatch optimization method based on the energy Internet of Things according to claim 1, characterized in that, The process of performing correlation analysis on the real-time energy monitoring data set and the historical energy dispatch record set to generate energy supply and demand correlation characteristics and network load characteristics includes: The energy production data, energy consumption data, and energy transmission data in the real-time energy monitoring dataset are timestamped to obtain a time-series energy data sequence. Extract the time series features of energy production data and energy consumption data of each energy node in the time series energy data sequence, and calculate the deviation rate parameter of energy production data and energy consumption data within the same time window; An energy supply and demand correlation matrix is ​​constructed based on the deviation rate parameter. The row vectors of the energy supply and demand correlation matrix represent the production capacity characteristics of energy production nodes, the column vectors represent the consumption demand characteristics of energy consumption nodes, and the matrix element values ​​represent the supply and demand matching degree between the corresponding production nodes and consumption nodes. The historical energy scheduling strategies and execution results data in the historical energy scheduling record set are processed by association rule mining to extract key load factors that affect energy transmission efficiency. The key load factors include transmission path length parameters, path loss rate parameters, and node connectivity parameters. Based on the key load factors, feature mapping processing is performed on the energy transmission data in the real-time energy monitoring dataset to generate a network load feature vector characterizing the load status of each transmission path. The dimension of the network load feature vector corresponds to the number of transmission paths in the energy transmission network.

3. The energy dispatch optimization method based on the energy Internet of Things according to claim 2, characterized in that, The step of extracting the time-series features of energy production and consumption data of each energy node in the time-series energy data sequence, and calculating the deviation rate parameter of energy production and consumption data within the same time window, includes: The time-series energy data sequence is divided into multiple time-continuous window data units by sliding window partitioning. Each window data unit contains energy production data and energy consumption data within a preset time length. The energy production data in each window data unit is subjected to trend fitting processing to generate a production trend curve. The slope of the production trend curve represents the rate of change of the energy production data. The energy consumption data in each window data unit is processed by extracting periodic components to generate consumption periodic features, which include the time interval parameter of the consumption peak and the peak duration parameter. The production trend curve and the consumption cycle feature are aligned on the time axis, and the absolute difference between energy production data and energy consumption data at the same time coordinate point is calculated. The deviation rate parameter is calculated based on the absolute difference and the energy consumption data at the corresponding time coordinate point. The deviation rate parameter is the ratio of the absolute difference to the energy consumption data, and is used to characterize the degree of energy supply and demand imbalance.

4. The energy dispatch optimization method based on the energy Internet of Things according to claim 2, characterized in that, The construction of the energy supply and demand correlation matrix based on the deviation rate parameter includes: The energy production nodes and energy consumption nodes in the energy Internet of Things system are numbered respectively to generate a production node index set and a consumption node index set. Using the set of production node indexes as row indexes and the set of consumption node indexes as column indexes, an initial association matrix framework is constructed. The number of rows in the initial association matrix framework is the same as the number of production nodes, and the number of columns is the same as the number of consumption nodes. For each combination of production and consumption nodes, extract the deviation rate parameter sequence within the corresponding time window, and calculate the mean and variance of the deviation rate parameter sequence; The mean and variance are normalized to generate standardized deviation feature values, and the range of the standardized deviation feature values ​​is a preset interval. The standardized deviation eigenvalues ​​are filled into the corresponding row and column index positions in the initial correlation matrix framework to obtain the energy supply and demand correlation matrix. The smaller the value in the matrix, the higher the supply and demand matching degree between the corresponding production node and consumption node.

5. The energy dispatch optimization method based on the energy Internet of Things according to claim 2, characterized in that, The step of performing feature mapping processing on the energy transmission data in the real-time energy monitoring dataset based on the key load factors to generate a network load feature vector characterizing the load status of each transmission path includes: The topology of each transmission path in the energy transmission network is analyzed to determine the node sequence and connection relationship of each transmission path. Based on the transmission path length parameter in the key load factors, calculate the physical distance characteristic of each transmission path, where the physical distance characteristic is the sum of the distances between all adjacent nodes in the path; Based on the path loss rate parameter in the key load factors and the energy transmission data in the real-time energy monitoring data set, the actual loss rate of each transmission path is calculated. The actual loss rate is the ratio of the difference between the energy input at the starting node and the energy received at the ending node of the transmission path to the energy input. Based on the node connectivity parameter in the key load factors, the number of connections of each node in each transmission path is counted, and the average node connectivity is calculated as a path connection complexity feature. The physical distance feature, actual loss rate and path connection complexity feature are fused to generate a path load feature vector with three dimensions. The path load feature vectors of all transmission paths are arranged in a preset order to form a network load feature vector that characterizes the load status of the entire energy transmission network.

6. The energy dispatch optimization method based on the energy Internet of Things according to claim 1, characterized in that, The process of prioritizing the multiple candidate scheduling schemes and calculating the comprehensive evaluation index for each candidate scheduling scheme includes: For each candidate scheduling scheme, extract the energy allocation parameters and transmission path selection parameters contained therein; Based on the energy allocation parameters and the supply-demand matching degree parameter in the energy supply-demand correlation characteristics, a supply-demand matching degree score is calculated, and the supply-demand matching degree score is positively correlated with the magnitude of the supply-demand matching parameter. Based on the transmission path selection parameters and the load level label in the network load characteristics, a load balance score is calculated. The load balance score is negatively correlated with the variance of the load level of each transmission path. Based on the path length and path loss rate parameters in the transmission path selection parameters, a transmission efficiency score is calculated. The transmission efficiency score is negatively correlated with the product of the path length and the path loss rate. The supply-demand matching score, load balancing score, and transmission efficiency score are standardized and weighted to generate a comprehensive evaluation index. The weight parameters in the weighted summation process are obtained by training with historical scheduling data. The comprehensive evaluation index is compared with a preset scoring threshold, and candidate scheduling schemes with comprehensive evaluation indices greater than the scoring threshold are selected as effective scheduling schemes.

7. An energy dispatching and optimization system based on the Internet of Things for Energy, characterized in that, The device includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the energy dispatch optimization method based on the energy Internet of Things as described in any one of claims 1-6.

Images