Multi-energy data integrated management method, system and device and storage medium

Abstract

The invention discloses a multi-energy data integrated management method, system and device and a storage medium, and relates to the technical field of data management, and the method comprises the steps: obtaining the energy data of each preset energy system in real time; preprocessing the obtained real-time energy data, and processing the preprocessed real-time energy data to obtain a first data processing result; and transmitting the first data processing result. According to the invention, the energy data of each preset energy system can be obtained in real time and comprehensively managed, the obtained first data processing result can provide accurate data support for management decision-makers, and the problem of information isolated island is avoided.

Description

Background Art

[0002] Renewable energy generation, such as solar power generation and wind power generation, has the advantages of cleanliness, convenience and economy, and can greatly alleviate environmental pollution problems. By collecting and managing solar power generation data and wind power generation data, it can provide data support for management decision makers.

[0003] At present, the collection and management of solar power generation data and wind power generation data are often independent of each other, which makes the data provided to management decision makers inaccurate, easily leads to management decision-making errors, and is prone to information island problems. Summary of the invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the prior art and specifically provide a multi-energy data integrated management method, system, device and storage medium, as follows:

[0005] 1) In the first aspect, the present invention provides a method for comprehensive management of multi-energy data, and the specific technical solution is as follows:

[0006] Get real-time energy data for each preset energy system;

[0007] Preprocessing the acquired real-time energy data, and processing the preprocessed real-time energy data to obtain a first data processing result;

[0008] The first data processing result is transmitted.

[0009] The beneficial effects of the multi-energy data comprehensive management method provided by the present invention are as follows:

[0010] It is able to acquire and comprehensively manage the energy data of each preset energy system in real time. The first data processing results obtained can provide accurate data support for management decision makers and avoid the problem of information islands.

[0011] 2) In the second aspect, the present invention also provides a multi-energy data integrated management system, and the specific technical solution is as follows:

[0012] Includes access layer, edge layer and network layer;

[0013] The access layer is used to: obtain energy data of each preset energy system in real time;

[0014] The edge layer is used to: pre-process the acquired real-time energy data, and process the pre-processed real-time energy data to obtain a first data processing result;

[0015] The network layer is used to transmit the first data processing result.

[0016] 3) In a third aspect, the present invention further provides an electronic device, comprising a processor, the processor being coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so that the electronic device implements any one of the above-mentioned multi-energy data integrated management methods.

[0017] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, any of the above-mentioned multi-energy data integrated management methods is implemented.

[0018] It should be noted that the beneficial effects achieved by the technical solutions of the second to fourth aspects of the present invention and the corresponding possible implementation methods can be found in the above-mentioned technical effects of the first aspect and its corresponding possible implementation methods, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments of the present invention:

[0020] Figure 1 A schematic diagram of a flow chart of a multi-energy data integrated management method according to an embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of the network model for the solar power generation prediction model;

[0022] Figure 3 This is a schematic diagram of the Bi-LSTM network model;

[0023] Figure 4 This is a schematic diagram of the network structure of the CNN-LSTM model that introduces attention;

[0024] Figure 5 It is a schematic diagram of queue processing;

[0025] Figure 6 is a schematic diagram of a sensing terminal network;

[0026] Figure 7 The figure is a schematic diagram of the structure of a comprehensive multi-energy data management system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The principles and features of the present invention are described below. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0028] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0029] like Figure 1 As shown, a multi-energy data integrated management method according to an embodiment of the present invention comprises the following steps:

[0030] S1. Obtain energy data of each preset energy system in real time;

[0031] Among them, multiple preset energy systems include: new energy stations, distributed power sources, systems used by load aggregators, virtual power plants, solar power generation systems, wind power generation systems, power grid distribution systems, systems used by electric vehicles, new power grid transmission systems, energy storage systems and electrical equipment (such as electric vehicles, smart buildings, non-industrial air conditioners, large industrial users, etc.), etc., which can be set according to actual conditions to achieve panoramic perception of the operating status of each link of source-grid-load-storage.

[0032] It adopts a variety of communication protocols and technologies, including Modbus, OPC, DLT645, MQTT, etc., supports a variety of data collection methods, and can realize the automatic collection, real-time transmission, storage and monitoring of energy data. It can also parse the received energy data and parse different types of energy data into energy data that conforms to the required data format. The required data format can be set according to the actual situation. It can also compress and optimize the collected energy data to reduce the bandwidth and storage cost of data transmission.

[0033] S2. Preprocessing the acquired real-time energy data, and processing the preprocessed real-time energy data to obtain a first data processing result;

[0034] Among them, the acquired real-time energy data is preprocessed, specifically data integration, data cleaning, data normalization, etc., to filter out unnecessary data and noise.

[0035] Among them, the processed real-time energy data can be processed through energy data analysis indicators. The energy data analysis indicators include energy consumption, energy efficiency, energy cost, energy quality and other aspects, aiming to provide a comprehensive energy data analysis reference, specifically:

[0036] 1) Energy consumption is one of the core parameters of enterprise energy data analysis. Energy consumption refers to the total amount of energy used by an enterprise in the production and operation process. Through the analysis of energy consumption, enterprises can understand the use and rules of energy, and then formulate effective energy-saving plans and optimization measures to improve energy utilization efficiency.

[0037] 2) Energy efficiency is an important parameter to measure the effect of an enterprise's energy use. Energy efficiency refers to the ratio of the energy used by an enterprise in the production and operation process to the products or services produced by the enterprise, and also includes the efficiency of various links such as the production, transmission and conversion of multiple types of energy and multiple grades of energy. Through the analysis of energy efficiency, enterprises can understand the effect of energy use and bottlenecks, and then formulate effective optimization measures to improve energy efficiency and economic benefits.

[0038] 3) Energy cost is another key parameter of enterprise energy management. Energy cost refers to the cost of energy consumed by enterprises in the production and operation process. Through the analysis of energy cost, enterprises can understand the composition and changing rules of energy cost, and then formulate effective energy-saving and cost-reduction measures to improve the economic benefits of enterprises.

[0039] 4) Energy quality is one of the important indicators in enterprise energy management. Energy quality refers to the quality and stability of the energy used by the enterprise. Through the analysis of energy quality, the enterprise can understand the stability and reliability of energy, and then formulate effective safeguards to ensure the normal operation of the enterprise's production and operation.

[0040] Among them, processing the pre-processed real-time energy data also includes: using distributed computing and artificial intelligence technology to achieve rapid data processing and analysis, and being able to extract useful information and features from the pre-processed real-time energy data, that is, the first data processing result, which can reduce network transmission pressure and improve data processing efficiency.

[0041] The useful information and features extracted include: the actual total solar energy and its fluctuations within a preset time period, the actual total wind energy and its fluctuations, the actual total external fuel, the actual total external electricity, the actual heat load, the actual cooling load, the actual electric load, the predicted value of wind power generation, the predicted value of solar power generation, the predicted value of electricity demand and the predicted value of energy demand.

[0042] The process of obtaining the predicted value of solar power generation is as follows:

[0043] Construct a solar power generation prediction model, such as Figure 2As shown in the figure, the solar power generation prediction model is trained based on the standard back propagation algorithm, and the trained solar power generation prediction model is used to predict the power generation of solar power. In the solar power generation prediction model, the encoder-based attention is applied to the input features to focus on the important features of a specific time; the decoder-based time attention is applied to the hidden state of the LSTM encoder to extract important time states. A linear layer is added to the output to predict the power generation of solar power. The entire model is trained based on the standard back propagation algorithm, as follows:

[0044] Photovoltaic power generation is affected by solar irradiance, temperature, wind speed, relative humidity and air pressure at different levels. Therefore, the Pearson product moment correlation coefficient method (PPMCC) is used to calculate the correlation coefficient between each weather variable and photovoltaic power as ξ:

[0045]

[0046] The statistics of different weather variables are used as inputs, and the Pearson correlations of these inputs with the target solar power are calculated. Although these weather variables have different correlations with the PV power, all weather variables will have an impact on the PV power in a specific case. For the convenience of data representation, let is the n-dimensional data feature input at time t, let is a series of k in L time step windows th Input feature data, then a series of n input data on L time step windows can be expressed as:

[0047] The input data includes historical solar energy data (photovoltaic power) and weather data (data of each weather variable), and the output values ​​in the past are set as (y1, y2, y3......y L ) and input values ​​(x1,x2,x3......x L ), the complete model for predicting the photovoltaic power of solar energy can be expressed by the following function F:

[0048] (y L+1 ,y L+2 ,y L+3 ......y L+l )=F(y1,y2,y3......y L ,x1,x2,x3......x L ) (2)

[0049] Among them, (y L+1 ,y L+2 ,y L+3 ......y L+l) is the solar power output to be predicted for l time steps. Both the encoder and decoder based on input and time attention are applied to the LSTM unit, which is used to read the dependencies of the sequence. It allows the processing of long structures during dataset training and creates predictions based on previous data in order to provide a long-term memory. In the solar power generation prediction model, there are internal self-loops for storing and analyzing information. The LSTM unit in the cell layer is shown. There are five important elements in the computational graph of the LSTM unit (input gate, forget gate, output gate, unit and state output. Gate operations such as reading, writing and erasing are performed on the unit storage state. The following formula describes the mathematical representation of the LSTM unit:

[0050]

[0051] Taken together, an LSTM unit can be represented as: t =f(h t-1 ,x t ).

[0052] In order to get the input sequence x k Extract relevant input features from the encoder and apply input attention to the encoder. Input attention can be calculated using x k And the previous hidden state and cell state of the encoder LSTM are applied, as shown in formula (4) and formula (5):

[0053]

[0054] z ε , W ε and U ε are the parameters to be trained. is an attention weight that represents the importance of the kth input feature at time t. To keep the sum of all attention weights equal to 1, a softmax activation is applied to This attention mechanism gives more weight to important features instead of treating all inputs equally. These attention weights can be used to extract new input sequences through formula (4). This new input is fed to update the encoder's LSTM hidden state, as shown in formula (6) and formula (7):

[0055]

[0056] The decoder model is designed to extract the important temporal hidden states and make the final output prediction. The results of the traditional encoder-decoder deteriorate as the length of the input sequence increases. Therefore, after the attention encoder, a temporal attention-based decoder is applied to select the relevant hidden states of the encoder from all time steps. The attention weight of each encoder hidden state can be calculated by using the previous hidden state of the decoder LSTM d t-1 and its cell state c t ' -1 , as shown in formula (8) and formula (9):

[0057]

[0058] where z d ,W d ,U d is the weight to be trained, and β is the importance of the hidden state of the i-th encoder at time t. Since the hidden state h of each encoder i has been mapped to the temporal component of the input, so the attention mechanism uses formula (8) to convert the context vector v t Calculated as encoder (h1,h2,h3......h L ). This vector is different at each time step. The vector is then concatenated with the target value using formula (9). Then, the decoder’s hidden state d is used t-1 and the newly connected value The new hidden state d of the decoder t It can be obtained based on formula (10) to formula (12) using the LSTM nonlinear function of the decoder:

[0059]

[0060] in, and are the weights and biases of the mapping concatenated.

[0061] The final output follows the decoder which consists of a linear layer for the predicted solar energy. The last layer advances the prediction time. The complete model can be represented by the following expression:

[0062]

[0063] in, is the solar power generation to be predicted, W y and is the weight to be trained. In the present invention, L is taken as 8 to predict the solar power generation one day ahead with a resolution of three hours. The entire model has been trained using a standard back propagation algorithm with an objective function defined by the mean square error (MSE):

[0064]

[0065] Where N is the number of training samples, Y t It is the actual value of solar power generation.

[0066] The present invention uses Bayesian optimization for parameter tuning through past statistical data as input data. Traditionally, manual or grid and random search techniques are used to tune hyperparameters, which is very time-consuming and relies on human expertise, while in grid search, the efficiency decreases as the number of hyperparameters increases. In random search, combinations of random parameters are sampled according to the statistical distribution given by the user, which may not be able to discover the best point in the search. Bayesian optimization takes into account past evaluations to select hyperparameters for the next evaluation. By choosing its parameters in an informed manner, it can focus more on areas in the parameter space where more promising parameters can be verified. It has three main parts: the search space from which parameters can be extracted, the objective function, and the agent. A probabilistic model of the objective function is constructed and used to select the most promising hyperparameters to evaluate the true objective function.

[0067] The present invention uses root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient to evaluate the model performance.

[0068] A method for calculating the weights of influencing factors is added to the model, and then the model is trained before prediction.

[0069] Through data training, the prediction accuracy of the model can reach 90%. The root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient for evaluating the model performance are all significantly better than those using ordinary LSTM and RNN. Taking into account the running time, computing power and accuracy, the model has good results.

[0070] The predicted value of wind power generation can be obtained by the following method, specifically:

[0071] Considering that wind power generation is greatly affected by external environmental factors, different environmental factors have different effects. For example, factors such as wind speed, temperature, and humidity have different effects on power generation. The clustering method is used to find the more closely related points, and the Pearson product-moment correlation coefficient method is used to calculate the relationship between each weather variable and wind power generation. The statistical data of different weather characteristics are used as input to calculate the Pearson correlation between these inputs and the target wind power generation. It is calculated that wind speed shows a strong correlation, temperature and humidity show a correlation, while light intensity and air pressure have little correlation. In order to better match different weight ratios, an attention mechanism is needed to focus on features that are more relevant at a specific time.

[0072] The wind power generation is predicted by the trained Bi-LSTM, such as Figure 3 As shown, the forward layer of the Bi-LSTM updates the parameters in the unit just like the LSTM network. The unidirectional LSTM model uses previous information to predict subsequent information, and the bidirectional LSTM (Bi-LSTM) improves prediction accuracy by combining the forward and backward information of the input sequence based on the LSTM. For the output at a certain time, the forward LSTM layer has the information of the input sequence at that time and before that time, and the backward LSTM has the input sequence information at that time and after. The output vectors of the two LSTM layers can be processed by addition, averaging, or concatenation. The forward LSTM in the horizontal direction and the vertical direction are calculated simultaneously for each time step t.

[0073] Hidden Vector and the reverse LSTM hidden vector Specifically, it is calculated by formula (15) to formula (17):

[0074]

[0075] Among them, LSTM(·) represents the LSTM function, and are the weight of the forward LSTM and the weight of the reverse LSTM, respectively, and the bias of the output layer is represented by . The attention model is used to represent the correlation between features and output results. The attention model assigns different attention weights through feature vectors to distinguish the importance of features and improve the accuracy of prediction. The present invention uses a feedforward attention model. The target attention weight v t Calculated by the hidden state output by the Bi-LSTM network:

[0076] v t =tanh(h t ) (18)

[0077] The generated attention weights are assigned to the corresponding hidden layer states h t, so that the attention weights generated by the model can work. The attention weights are calculated by formula (19), and the attention weighted average is calculated by formula (20):

[0078]

[0079] Use one-dimensional CNN to process sequence data. Compared with traditional neural networks, it introduces convolutional layers and pooling layers. The input data is compressed and important features are extracted through convolution operations. The goal of the maximum pooling layer is to reduce the size of the feature map to grasp the features and reduce the computational cost.

[0080] Through data training, the prediction results of the model can reach 95% accuracy on this data set, and the root mean square error (RMSE) for evaluating model performance has obvious advantages over ordinary LSTM and RNN. Taking into account the running time, computing power and accuracy, the model has good results.

[0081] The predicted energy demand is obtained in the following way:

[0082] like Figure 4 As shown in the figure, the CNN-LSTM model is used to predict energy demand while introducing attention. The model first uses the CNN module to extract effective feature information from the historical energy demand data sequence. Two convolutional pooling modules are used to extract and enhance the features of the input data, and then the output results are connected through one-dimensional expansion and a fully connected layer to obtain abstract high-level features. The LSTM module is used to further learn timing information. The two-layer LSTM network can fully extract and learn timing information. A loss layer is added between the two layers of the LSTM network. During the training process, the forgetting network randomly inactivates neurons according to the set probability, which can effectively avoid the phenomenon of overfitting. The attention mechanism is introduced to assign different weight probabilities to the features further extracted by the LSTM module to enhance the influence of important information on the results. Finally, the results are output through the output layer. The attention model assigns different attention weights through feature vectors to distinguish the importance of features and improve the accuracy of predictions. Target attention weight v t The hidden state h output by the LSTM network is calculated using formula (18).

[0083] The generated attention weights are assigned to the corresponding hidden layer states h t , so that the attention weights generated by the model can work.

[0084] Through data training, the prediction results of the model can reach 93% accuracy on this data set, and the root mean square error (RMSE) for evaluating model performance has obvious advantages over ordinary LSTM and RNN. Taking into account the running time, computing power and accuracy, the model has good results.

[0085] S3. Transmit the first data processing result.

[0086] It can support multiple network protocols, such as TCP / IP, HTTP, MQTT, etc., to ensure efficient transmission and real-time response of data (such as the first data processing result, etc.), and can also encrypt the transmitted data (such as the first data processing result, etc.) to ensure security and confidentiality. It can also perform quality control on the data to ensure the reliability and validity of the data (such as the first data processing result, etc.).

[0087] Optionally, in the above technical solution, the real-time energy data will be prioritized, and the data packets with different business requirements will have different priorities. A strict priority strategy will be implemented. The queue processing process is as follows: Figure 5 As shown in the figure, the specific performance is as follows: the energy data transmitted from the upper layer will be composed of multiple data frames. The data frames enter the transmission queue at a certain rate, receive services in turn and then leave the system. Therefore, the transmission queue is modeled as an M / G / 1 queue. Among them, M represents the input process, the arrival of the data frame obeys Poisson distribution, G represents the service time, which obeys uniform distribution, and 1 means that only one data frame can be processed at a time.

[0088] In this model, high-priority packets are strictly processed first, which may cause low-priority packets to wait for a long time. Therefore, a packet waiting time evaluation index, information age, is introduced as an indicator to measure the freshness of information. It refers to the waiting time from the latest information entering the waiting queue to the completion of processing. At a point in time, the entire queue state can be represented as a tuple S (M, H, V, A), where M refers to the number of packets in the queue at a certain moment. Let the matrix composed of the data waiting time in each priority queue be H, where h i，j , represents the current waiting time of the jth unprocessed data at priority i. According to the definition of information age, if the data is not processed, then the information I in the priority queue i，j The freshness of the information "ages" in the next time slot:

[0089] h i , j (t) = h i , j (t-1)+u(t) (21)

[0090] The priority weight matrix is ​​V, and a new processing evaluation coefficient A is set, where the new weight sends the maximum value out as quickly as possible. In order to ensure the priority and the minimum overall information age, it is necessary to select the most appropriate priority weight matrix according to time and formulate a suitable aging strategy. At the same time, the constraint strategy also needs to be met:

[0091] M i ×size <Cache size (twenty two)

[0092] The final solution is min∑A i Solve to obtain the minimum information age.

[0093] There are many factors that affect the age of information. In terms of service demand, there are factors such as data transmission delay, energy consumption, success rate and throughput; in terms of network status, there are factors such as channel resources, processing speed, cache resources and historical records. By measuring these factors in multiple dimensions and working together, we can normalize the influencing factors and jointly decide on the node processing strategy, forwarding strategy and content caching decision.

[0094] (X1, X2, X3...., X n ) represents the factors that affect the evaluation coefficient, x1, x2, x3..., x n Represents the specific value of the uncertain factor. According to the weight model theory, the following objective function can be defined:

[0095]

[0096] Among them, (w1, w2, w3...., w n ) is the weight of each variable in the overall target, U i (x i ) is the effect function, and the Pearson correlation coefficient can be used as the effect function. At the same time, in order to adapt to the constant changes in network status, dynamic variables need to be introduced to define the following objective function:

[0097]

[0098] Among them, i (x i ) is a dynamically adjusted parameter.

[0099] Looking at the overall model, artificial intelligence learning is performed based on historical data, and the artificial intelligence model is used to estimate the amount of data in each priority queue at different times. Then, based on the arrival quantity and priority control requirements, the parameters are dynamically adjusted together to minimize the overall information age.

[0100] The entire scheduling method makes reasonable arrangements for high-priority and low-priority data, assisting the normal operation of the entire network.

[0101] Optionally, in the above technical solution, it also includes:

[0102] S4. Process all energy data to obtain a second data processing result;

[0103] Among them, processing all energy data specifically refers to: analyzing and processing all energy data, including data mining, modeling, prediction, etc., providing data analysis and intelligent decision support, and the second data processing results include: data mining results, established data models and prediction results.

[0104] Among them, the data mining results include: data analyzed according to energy data analysis indicators, the established data models include: solar power generation prediction model, Bi-LSTM network model and attention-introducing CNN-LSTM model, etc. The prediction results may include the prediction data in the first data processing result. It should be noted that the process of analyzing and processing all energy data can adopt the data processing process of real-time energy data. The difference is that when obtaining the first data processing result, real-time energy data is used, and when obtaining the data processing result, all energy data are used.

[0105] Among them, the prediction results in the second data processing results also include: the predicted power load. Power load prediction is an important basis for power system decision-making, because predicting active and reactive loads in the future can help power system decision makers make better decisions. The present invention proposes a reactive load prediction method based on the combination of K-means clustering and random forest. The method first starts from the data characteristics of the load, uses K-means clustering to obtain different load categories, and then establishes a random forest prediction model in each load category. When predicting, the load data is first divided into different categories, and then the corresponding random forest model is used for prediction. Finally, the prediction results of all categories are merged to obtain the final reactive load prediction result. The specific process is as follows:

[0106] 1) Data collection and preprocessing:

[0107] Collect reactive power data sets of different loads in the regional power system over 8 days. Load data is collected approximately every 2 minutes, with a total of 10 sampling points.

[0108] The data and labels are made into a data set for subsequent normalization processing.

[0109] Normalize the load data and divide it into a training set and a test set. The purpose of normalizing the load data is to reduce the value of the load data to a certain range, which helps to reduce the magnitude gap within the data, and helps to reduce the complexity of the subsequent deep learning process, speed up the execution of the program, and better adapt to the application scenarios of massive data analysis and processing. The normalization method is as follows:

[0110] Before clustering, all load data are mapped to [0, 1] using min-max normalization, which is implemented by formula (25):

[0111]

[0112] Among them, x max is the minimum value of a load history data, and x min It is the maximum value of a certain load history data.

[0113] 2) Demand side load clustering analysis:

[0114] In order to make the load forecasting results more accurate, the K-means clustering algorithm is first used to divide the load into k categories, and the loads with similar historical data are grouped into one category.

[0115] Use the K-means clustering algorithm to divide multiple loads into several typical types. K-means clustering selects the original cluster center and the number of clusters so that the Euclidean distance from each feature vector to the selected cluster center is the smallest. The steps of the K-means clustering analysis algorithm are as follows:

[0116] ① Select k points in the data set as the initial cluster center points;

[0117] ② Calculate the Euclidean distance from other points to the k center points and assign them to the cluster represented by the nearest cluster center. The calculation formula of Euclidean distance is:

[0118]

[0119] Where N is the size of the sample vector.

[0120] ③ Calculate the average of all samples classified into each category and update the center of each category.

[0121] ④ Perform iterative updates according to the above steps ① to ③ until the square error criterion function stabilizes at the minimum value. The square error criterion calculation formula is:

[0122]

[0123] Among them, k is the number of clusters to be clustered, n is the number of samples in the class, is the average value of the class samples.

[0124] After the data is processed by the above algorithm, the above data set can be divided into k types of load data with different categories, different degrees and different loads. The loads with the same historical data are classified into the same category, that is: Based on the historical load data, this paper constructs four features for each load type: load label, load at the same time of the previous day, load at the same time of the previous 12 hours, and average load of the previous 24 hours.

[0125] The data after clustering can be distributed in a "similar" manner, which can help the subsequent feature processing to have a good distribution and accelerate the fitting speed of the random forest algorithm model training and the accuracy of the algorithm model.

[0126] 3) Use random sample training samples and some features to form a random subspace training decision tree. Random forest is a comprehensive learning method composed of multiple decision trees. It can effectively solve the problem of fast processing of big data. At present, the amount of SCADA data in the power grid is increasing, and there is an urgent need for artificial intelligence algorithms such as random forests to help power companies analyze and process big data in power grid operation. Therefore, the random forest algorithm has good application prospects in the power reactive load forecasting in the field of big data.

[0127] Random forest is composed of its weak model classification and regression tree (CART) combined with bagging algorithm. CART can be used for regression and classification, and the corresponding random forest can also be used for classification and regression analysis.

[0128] CART uses binary division to recursively divide each feature, thereby dividing the feature space into finite units and determining the predicted probability distribution on these units; CART uses the Gini coefficient to select features. Assuming that a given data set has categories, and the number of categories is, the Gini coefficient of the data set is:

[0129]

[0130] If the sample set D is divided into two parts D1 and D2 according to a certain attribute value a of feature A, the Gini coefficient under the feature condition is:

[0131]

[0132] The constructed data set is sent to the random forest algorithm for training, and the model is used to meet the demand for load forecasting within a certain period of time. Specifically:

[0133] ①Algorithm initialization:

[0134] RF is a set of regression decision trees {h(X,θ k ),k=1,...,N tree}, where k represents a random vector, independent and identically distributed to the Kth decision tree. An algorithm scale (for regression) is used to obtain the final predictions of the random forest.

[0135] For the input vector X, it contains at most J different categories and is set to the correct classification category. For the input vector X and the output vector Y, the edge function is defined as follows:

[0136] F(X,Y)=a k I(h(X,θ k )=Y)-max j≠Y a k I(h(X,θ k )=j) (30)

[0137] Where j represents a category in J, I(·) represents the metric function, and a k represents the average value, and k=1,...,n.

[0138] But since the number of decision trees tends to infinity:

[0139]

[0140] a vk is an average function of the setting, θ k is the decision variable representing the Kth decision tree, E θ Represents the expected function. From this, we can infer that the generalization error of RF is:

[0141]

[0142] Among them, YE θ h(Xθ) represents the regression function of the random forest. The average generalization error is obtained by adding an expectation function to formula (32):

[0143]

[0144] The random variables θ and θ′ are independent of each other in different decision trees. Assume that for θ, E Y =E X h(Xθ), it can be proved that:

[0145]

[0146] in, Represents the weight coefficient of the residual Yh(Xθ) and Yh(Xθ′).

[0147] Finally, the mean absolute percentage error (MAPE) and root mean square error (RMSE) are taken as evaluation indicators to evaluate the accuracy of the reactive load prediction model.

[0148] ②Algorithm steps:

[0149] Step 1: Extract cluster features, load labels: There are multiple different loads in each type of load. Therefore, I label these loads in the order of 1, 2, and 3. Find the load data at the same time of the previous day in the load history data and assign it to the feature y at that moment 24 Find the load data at the same time in the previous 12 hours in the load history data and assign it to the characteristic y at this time 12 . Average load of the previous 24 hours: As shown in formula (35), the average load of the previous 24 hours is assigned to a at this moment.

[0150]

[0151] Step 2: Using the bagging idea, K different sample data sets are extracted from the original data set, and the sample size of each data set is the same as that of the original data set.

[0152] Step 3: Build a regression decision tree for each sample to generate a decision tree. During the generation process, select the optimal variable for node splitting until all decision trees stop growing to form a random forest.

[0153] Step 4: Input the test data into the random forest model and use the average of the decision tree prediction results as the prediction value

[0154] Step 5: Get the prediction results and perform error calculation to evaluate the model performance.

[0155] Compared with a single prediction model, this method can better consider the volatility and nonlinear characteristics of load data, and classify and process data through clustering methods, thereby improving the prediction accuracy. In addition, the random forest algorithm can not only handle the complex relationship between multiple input features, but also automatically identify important features, further improving the prediction accuracy. This method has achieved good results in practical applications and can be effectively applied to reactive load forecasting.

[0156] Optionally, the quality of the heat energy output from the source side is divided, and the heat grade requirements of different loads are clarified according to the quality of the heat energy demand of each energy-consuming unit, so as to construct a ladder utilization relationship of energy and improve the utilization rate of heat energy. On the other hand, it is also necessary to comprehensively consider the quantity and price relationship between heat energy supply and demand, dynamically optimize and solve, and achieve maximum economic benefits. The method includes: obtaining the heat energy demand data on the load side and the heat energy output data on the source side and constructing a data set; using the K-Means clustering algorithm to perform cluster analysis on the monitoring data to obtain the classification level of supply and demand heat energy; matching the output heat energy with the load through an intelligent optimization algorithm; and finally completing the system scheduling work in combination with the energy ladder utilization structure.

[0157] The dispatching strategy of the power dispatching system includes a series of key operational steps, the purpose of which is to maximize the efficiency of energy resource utilization and economic benefits while ensuring the reliable operation of the power grid. It includes the following:

[0158] 1) Determine the storage status of the urban energy storage system. If the energy storage system is insufficient, it means that other energy sources need to be used first to meet the electricity demand. In this case, the dispatch system will give priority to activating the gas cogeneration module and calling for external power.

[0159] 2) When energy storage is insufficient, the dispatch system will first rely on gas cogeneration and purchased electricity to ensure the continuity and stability of power supply.

[0160] 3) If the energy storage system has sufficient power, the dispatch system will use a more complex algorithm for comprehensive dispatch. This dispatch strategy will take into account a variety of factors, including traditional energy costs, new energy costs, electricity market prices, etc. The goal is to optimize the dispatch plan and achieve maximum economic benefits.

[0161] 4) At the same time, the dispatching system will also determine the distribution method of the output electricity in the park. Part of the electricity will be sold to the power grid for profit, and the other part will be supplied to users in the park, which not only obtains economic benefits but also promotes the consumption and utilization of renewable energy. In addition, the dispatching system can also dynamically adjust the charging and discharging time according to the electricity price, realize valley charging and peak discharging, and reduce the operating cost of the system. The dispatching system will also consider the heat load demand, optimize the cogeneration of electricity and heat, and improve the efficiency of energy utilization.

[0162] Through real-time monitoring and intelligent dispatch analysis of production and demand, the power dispatch system can provide dispatch strategies based on power generation resources to ensure the safe and stable operation of the power grid and maximize the economic benefits of the park, while maximizing the role of clean energy to meet the requirements of environmental protection and sustainable development.

[0163] The second data processing result also includes: an optimization operation strategy determined by the multi-energy coupling collaborative optimization operation method, specifically:

[0164] The electricity generated by each energy subsystem is used to charge its own energy storage system and meet various energy needs of the park. When the park's own energy storage system is insufficient, the energy subsystem considers the park's own production capacity and relies on purchasing energy from external power grids and gas pipelines to meet its own needs. Through IES, the coordinated optimization of electricity, heat, gas, and cold can be achieved. While meeting the operating constraints and IES requirements of each subsystem, the ultimate benefit of the entire cogeneration system, including the cascade utilization of multi-grade thermal energy, is optimized to carry out multi-energy flow coordination of the system.

[0165] In order to determine whether the energy subsystems in IES are coupled to each other through energy distribution, the energy demand parameter ▽ is defined. Assuming that the external energy supply system can provide IES with sufficient electricity, gas, heat, cold and other energy sources, the energy distribution coefficient of the external energy network does not have a negative value, that is, ▽>0, for the energy flow between any energy subsystem i and energy subsystem j. For each energy subsystem, the energy demand parameter is shown in formula (36):

[0166]

[0167] The interaction between energy subsystems i and j can be expressed by the coefficients a i,j 、b i,j , c i,j and d i,j Indicates. represents the flow of power demand of the subsystems in the whole system, They represent the demand flows of heat, cold and gas respectively, so ▽ expresses the total amount of demand in the whole system, where each X can be specifically divided into two parts: civil and commercial. At the same time, the matrix N = [N1, N2, N3, N4] is defined, and the matrix form is as follows, where the superscript a represents the parameter of the matrix, 1≤i≥4 and rounded, and the matrix M as a whole represents the subsystem flow distribution parameter matrix in the whole system:

[0168]

[0169] Substituting the matrix N into formula (36) we obtain:

[0170] M=[C1 C2 C3 C4] T (38)

[0171]

[0172] Among them, i∈{e,g,h,k} represents different subsystems such as cooling, heating, electricity, and gas, and the constraints are set as Even if the energy flow of each subsystem cannot be 0, it can be seen that C in the formula represents the energy produced by the entire system, which is solar energy, wind energy, transferred electrical energy, and thermal energy, respectively, while Q, W, E, and R represent their respective production or scheduling costs, so Cost represents the cost of the energy supplied by the entire system.

[0173] Most of the previous optimization algorithms use the particle swarm optimization (PSO) algorithm proposed by Kennedy and Eberhart after studying the foraging behavior of bird groups to solve the optimization problem in the objective function. Based on the previous algorithm, an improved swarm intelligence algorithm is used to solve the objective function optimization problem of the model in this paper. Specifically:

[0174] In order to realize the reasonable multi-energy flow representation between multiple subsystems in IES, the unknown variables M = [C1 C2 C3C4] T , the optimal energy distribution of IES can be achieved. The optimal economic effect problem of IES is expressed by a linear programming model as follows:

[0175] max[ops],k=2,…,(N+1) (40)

[0176] Among them, ops represents the optimal economic effect value of the entire system, that is:

[0177] ops=▽-Cost (41)

[0178] For N subsystems in IES, the dimension of the state variable of each subsystem is represented by d, and n subsystems are represented by particle swarms. Then the position of the i-th particle can be represented by x i =(x1,x2,...,x id ), the velocity of the ith particle can be expressed as v i =(v1,v2,...,v id ), obviously, each particle position representing the energy subsystem is a solution of IES, and the size of ops (i.e., fitness function) is used to judge the quality of the solution. The steps are as follows:

[0179] Start solving:

[0180] N energy subsystems are represented by the same number of particle groups. The position and velocity of the particles corresponding to each energy subsystem are x i and v i , i∈[1],[n], use formula (40) to calculate the adaptive value f(x i ), according to f(x i ), calculate the global optimal value f of the particle swarm gb and the particle position x gb =(x gb1 ,x gb2 ,x gbd ) and the optimal value f experienced by each particle db and the particle position x db =(x db1 ,x db2 ,x dbd ), update the velocity and position of each particle respectively:

[0181] v ij (t+1)=v ij j(t)+c1r 1j (x dbj -xij (t))+c2r 2j (x gbj -x ij (t)) (42)

[0182] Where i∈[1],[n],j∈[1],[n],t is the number of updates, c1 and c2 are constants greater than 0, representing the learning factor, r 1j and r 2j are independent random numbers and follow a uniform distribution in [0, 1]:

[0183] x ij (t+1)=x ij (t)+v ij (t+1) (43)

[0184] Determine whether formula (41) converges to the minimum value. If it converges to the minimum value, the final solution is obtained; otherwise, jump to the "start solving" step to continue solving the optimal solution of the model.

[0185] Also includes: storing all energy data.

[0186] S5. Transmit the second data processing result.

[0187] Optionally, in the above technical solution, it also includes:

[0188] S010, construct a perception terminal network including terminal nodes, central nodes and routing nodes, such as Figure 6 As shown, the functions of the terminal node, central node and routing node are as follows:

[0189] 1) Terminal node: directly responsible for collecting energy data and sending the collected energy data to the routing device node or coordinator node. The terminal node can be directly connected to the coordinator node or connected to the coordinator node through the routing node. The terminal node is specifically a terminal device.

[0190] 2) Routing node: responsible for forwarding data packets, performing data routing and routing maintenance, allowing routing nodes to join the perception terminal network and assist its child nodes in communication; routing nodes are relays between terminal nodes and central nodes, and they relay communications between terminal nodes and central nodes. Routing nodes specifically refer to routing devices.

[0191] 3) Central node: The central node is the convergence point of information of each node in the perception terminal network. It is the core node of the perception terminal network and is responsible for establishing, maintaining and managing the perception terminal network, and realizing data transmission between each node and the host computer. The central node has strong communication, processing and transmission capabilities, and can send data to the remote control terminal.

[0192] S011. Using a distributed address allocation algorithm to allocate a routing address to each routing node and each terminal node in the perception terminal network, wherein the distributed address allocation algorithm is described as follows:

[0193] Given the maximum depth of the sensing terminal network and the maximum number of devices (routing devices and terminal devices) (the sum of the maximum number of routing devices and the maximum number of terminal devices), the number of sub-segment addresses SA that can be allocated to each device can be calculated using formula (44):

[0194]

[0195] Among them, Cm represents: the maximum number of all devices, Rm represents: the maximum number of sub-devices (routing devices and terminal devices) allowed to be connected by the routing device, d′ represents: the network depth of the routing device or terminal device itself, and Lm represents: the maximum depth of the perceived terminal network.

[0196] If the SA value of a device is 0, it has no routing capability and is a terminal device. The routing address is assigned using formula (45). If the SA value of a device is greater than 0, it has routing capability and is a routing device. The routing address is assigned using formula (46):

[0197] An1=Ak1+1+Cskip(d″)*(n1-1) (45)

[0198] An2=Ak2+Cskip(d″′)*Rm*(n2-1) (46)

[0199] Formula (45) can be called the terminal device routing address allocation formula, wherein An1 represents: the routing address allocated to the terminal device to which the routing address is currently to be allocated, Ak1 represents: the routing address of the terminal device to which the routing address is currently to be allocated, which has been previously allocated, Cskip(d″) represents: the offset calculated based on the current network depth d″, specifically representing the allocation interval of the terminal device's routing address, n1 represents: the number of devices (including terminal devices and routing devices) to which routing addresses are currently allocated, and the terminal device address allocation formula represents: the routing address of the terminal device to which the routing address is currently to be allocated: is the routing address of the terminal device to which the routing address is currently to be allocated, which has been previously allocated, which has been allocated, plus Cskip(d″)*(n1-1), and the offset Cskip(d″) changes with the change of the network depth. Specifically, the offset Cskip(d″) determines the allocation interval of the terminal device's routing address, and (n1-1) represents: the number of devices (including terminal devices and routing devices) to which routing addresses are currently allocated minus 1, which is used to adjust the increment of routing addresses.

[0200] Formula (46) can be called the routing address allocation formula for routing devices, wherein An2 represents: the routing address allocated to the routing device to which the routing address is currently to be allocated, Ak2 represents: the routing address of the routing device to which the routing address is currently to be allocated, which has been previously allocated, Cskip(d″′) represents: the offset calculated based on the current network depth d″′, specifically representing the allocation interval of the routing address of the routing device, n2 represents: the number of devices (including terminal devices and routing devices) to which routing addresses are currently allocated, and the routing device address allocation formula represents: the routing address of the routing device to which the routing address is currently to be allocated: is the routing address of the routing device to which the routing address is currently to be allocated, which has been previously allocated, which has been allocated, plus Cskip(d″′)*Rm*(n2-1), and the offset Cskip(d″′) changes with the change of the network depth. Specifically, Cskip(d″′) determines the allocation interval of the routing address of the routing device, and (n2-1) is the number of devices (including terminal devices and routing devices) to which routing addresses are currently allocated minus 1, which is used to adjust the increment of the routing address.

[0201] In general, the terminal device routing address allocation formula is used for the allocation of routing addresses of terminal devices. It will increase each time it is allocated, and the routing address interval between each two adjacent terminal devices with assigned routing addresses is controlled by SAC(d″). The routing address allocation formula for routing devices is used for the allocation of routing addresses of routing devices. It will increase each time it is allocated, and the routing address interval between each two adjacent terminal devices with assigned routing addresses depends on the network depth, routing capacity (reflected in Rm) and the number of devices with assigned routing addresses.

[0202] When calculating Cskip(d″′), in addition to considering d″′, the type of the routing device is also considered (different types of routing devices have different Rm).

[0203] Assume that a routing device wants to send a data packet to the target routing address D′. The network depth in the network address of this routing device is d″″. The routing device will first determine whether the target device of the target address is its child device. If so, it should satisfy formula (47):

[0204] A<D′<A+Cskip(d″″-1) (47)

[0205] Wherein, A is a positive integer.

[0206] If the target device is a child device of the routing device, the next hop routing address N′ is:

[0207]

[0208] If the target device is not a child device of the routing device, the routing device sends the data packet to the parent node of the routing device.

[0209] In S1, energy data of each preset energy system is obtained in real time, including:

[0210] S10, collecting energy data of each preset energy system in real time through each terminal node, and sending it to the central node through the corresponding routing node;

[0211] S11. Receive energy data of each preset energy system sent by the central node.

[0212] Optionally, in the above technical solution, it also includes:

[0213] S012. Determine the optimal position of each terminal node and each routing node in the perception terminal network through a decision access algorithm, and the decision access algorithm is explained as follows:

[0214] S0121. Initialize the particle swarm: Assume that there are N particles in the D-dimensional space, and assign a random initial position and initial velocity to each particle:

[0215] Initial position of the particle: x i =(x i1 , x i2 , …x iD ), the initial velocity of the particle: v i =(v i1 , v i2 ,…v iD ), the position of the particle in each dimension must be restricted to the interval: [X min，d , X max，d ], the particle speed in each dimension is limited to the interval: [-V max，d , V max，d ].

[0216] Among them, each terminal node and each routing node in the perception terminal network is a particle, and there are N particles in total.

[0217] S0122. Calculate the fitness value of each particle according to the fitness function: Introduce the fitness function. In the network selection of heterogeneous networks, fitness can be understood as the user's satisfaction with the services provided by the network. Different business types have different characteristics and service quality requirements. In order to express the requirements of different businesses for network attributes, the fitness function is used to uniformly calculate the network attributes. Network attributes can be divided into benefit attributes and cost attributes. Benefit attributes refer to attributes with larger utility values, such as bandwidth. Cost attributes refer to attributes with smaller utility values, such as latency, jitter, and packet loss rate.

[0218] To find the best individual fitness value, the formula for calculating the best position that a particle has ever experienced is as follows:

[0219] pbest i =(p i1 , p i2 , …p iD ) (49)

[0220] For each particle, the fitness value of its current position is compared with the fitness value corresponding to its historical best position (pbest). If the fitness value of the current position is higher, the current position is used to update the historical best position.

[0221] S0123. Find the optimal fitness value of the group:

[0222] The best position solution for all particles is calculated by formula (50):

[0223] gbest i =(g i1 , g i2 , …g iD ) (50)

[0224] For each particle, the fitness value of the current position is compared with the fitness value corresponding to its global best position (gbest). If the fitness value of the current position is higher, the global best position is updated with the current position.

[0225] Update the speed and position of each particle. Specifically, the particle speed update formula is:

[0226]

[0227] The particle position update formula is:

[0228]

[0229] represents the d-th dimension component of the velocity vector of particle i at the k′th iteration; represents the d-th dimension of the particle position vector at the k′th iteration; c1 and c2 represent acceleration constants, which adjust the maximum learning step size; r1 and r2 represent two random parameters, with a value range of [0, 1] to increase the randomness of the search; w represents the inertia weight, a non-negative number, which is used to adjust the search range of the solution space. The inertia weight (w) represents the influence of the particle's previous velocity on the current velocity. If w is large, the particle has a strong global optimization ability and a weak local optimization ability. Conversely, the particle has a strong local optimization ability and a weak global optimization ability. In other words, if w is too large, it is easy to miss the optimal solution; if w is too small, the algorithm converges slowly or is prone to fall into a local optimal solution. When the problem solution space is large, in order to achieve a balance between search speed and search accuracy, the usual practice is to make the algorithm have a higher global search ability in the early stage to obtain a suitable seed, and a higher local search ability in the later stage to improve convergence accuracy. Therefore, there is the following dynamic inertia weight formula:

[0230]

[0231] run indicates the current number of iterations; run max Represents the maximum total number of iterations of the algorithm; as the number of iterations increases, w decreases continuously, so that the algorithm has a stronger global convergence ability in the early stage and a stronger local convergence ability in the late stage.

[0232] Formula (53) shows that the speed of a particle is related to its previous speed, pbest, and gbest. Therefore, formula (53) is divided into three parts: the first part is the previous speed of the particle; the second part is the "cognitive" part, which represents the particle's own thinking, which is the distance between the current position of particle i and its best position; the third part is the "social part", which represents the information sharing and cooperation between particles, which is the distance between the previous position of particle i and the best position of the group. In general, the speed of a particle is the weighted vector sum of the speeds in the three directions.

[0233] S0124, determine whether the algorithm is finished: if the end condition is not met, return to S0122, if the end condition is met (the preset number of iterations is reached or the fitness value no longer changes significantly), the algorithm ends, and the global best position (gbest) is the global optimal solution.

[0234] The present invention helps to realize efficient, safe and reliable transmission and processing of data, and also provides flexible expansion and integration capabilities.

[0235] Optionally, the summarized data can be displayed visually in the form of charts, reports, etc., so that users can more easily understand the data situation and trends and support users in data exploration and decision-making.

[0236] The platform system docking and terminal collection are used to aggregate multi-source data to achieve source-grid-load-storage coordination. The platform needs to access the power grid data of WAMS, SCADA and other systems to achieve real-time access and management of massive data on the power supply side and energy storage side, and apply big data stream processing technology to achieve preliminary data cleaning. The platform needs to support multiple types of communication protocols (including but not limited to 376.1, 101, 104, etc.), and support real-time collection and monitoring of load-side data from terminals such as charging piles, smart sockets, electric heating equipment, and non-invasive equipment.

[0237] Data access status monitoring: This function can monitor the access status of various data sources in real time, including access speed, access success rate, access error rate and other indicators. Through intuitive chart display, users can quickly understand the data access status.

[0238] Data quality monitoring: This function can monitor the quality of the accessed data, including data integrity, accuracy, consistency, etc., timely discover and handle data quality issues, and ensure the accuracy and reliability of the data.

[0239] Data real-time monitoring: This function can monitor the real-time access of data in real time, including indicators such as data update frequency and delay time, to ensure that data is updated in a timely manner and meet the real-time needs of users.

[0240] Data anomaly monitoring: This function can monitor abnormal situations in the data access process, such as data source connection interruption, data format error, etc., and promptly discover and solve these abnormal problems to ensure the reliability of data access.

[0241] Data analysis: This function can analyze the monitored data, identify abnormal problems and bottlenecks in data access, and provide solutions to help enterprises improve the efficiency and reliability of data access.

[0242] Optionally, the first data processing result and the second data processing result may be displayed in a multi-dimensional visual manner, specifically:

[0243] ① Visualization of power supply equipment: HT uses a self-developed 2D and 3D visualization engine to intelligently transform the building power supply system and form a self-management system, so that every detail can be interacted and fed back in real time. It includes sensors, smart IoT GPRS / 4G / NB-LOT, and big data system platform equipment information, and realizes a 2D configuration effect that can be viewed anytime and anywhere even on mobile phones. By collecting relevant national, provincial, and municipal power consumption limit standards, comprehensive data is summarized, and real-time warnings are issued for equipment in an over-limited state.

[0244] ② Visualization of power generation data: Supports lightweight photovoltaic power generation and wind power generation monitoring scenarios. Combining the principles and models of photovoltaic and wind power generation, the solar panels (assemblies), controllers, inverters and wind turbines are presented in the form of logic diagrams to help managers sort out the power generation data context. This allows for output energy balance adjustment, comprehensive analysis, power supply reliability, and improved emergency response efficiency for power outages on rainy days.

[0245] ③ Energy storage system modeling: Energy storage is easily affected by multiple factors such as weather, temperature, and humidity during actual operation, resulting in intermittent photovoltaic and wind power and random fluctuations in grid power. Energy storage systems in multiple environmental scenarios are built through satellite images and on-site photography. Energy storage equipment is integrated and connected with environmental data and TN-S power supply system data to provide continuous and stable power supply to the load, thereby improving system efficiency and power supply reliability.

[0246] ④ Visualization of energy storage information: The scene is displayed in the style of a line chart, highlighting the energy storage equipment model. Monitor the energy storage equipment and real-time power running on each floor, combined with the real-time update of background data, to form a knowledge base and release the stored energy when needed.

[0247] ⑤ Demonstration of distribution current flow: The power supply system is usually built at the same time as the main building, and there are problems such as aging equipment, many safety hazards, and low intelligence. The system uses a building power supply system as the background to simulate the application scenario of distribution equipment in the building. At the beginning of the page, it is displayed in a wireframe style, and the flow of lines is used to express the entire process of current transmission from the underground distribution room to each floor, and then to each device.

[0248] ⑥ Demonstration of water storage control process: Water storage technology combines the excess valley power of the nighttime power grid with the sensible heat of water to store cold, and uses the stored low-temperature chilled water to provide cooling for air conditioning during the peak hours of electricity consumption during the day. The system integrates background chart data, builds a text-based cold storage process flow, and demonstrates the monitoring and maintenance of visual system data, saving equipment investment and achieving "a small horse pulling a big cart".

[0249] ⑦Demonstration of cooling flow: The red dynamic lines show the functional lines of the water storage cooling system, integrating the power line carrier technology to build a dynamic functional live. It realizes the emergency function, improves the reliability of the air conditioning system, and shortens the start-up time to reach the required temperature. In the actual "valley control peak use", the air conditioning cooling saves a lot of air conditioning electricity bills and reduces the production cost of the factory.

[0250] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0251] like Figure 7 As shown, a multi-energy data integrated management system 200 according to an embodiment of the present invention includes an access layer 201, an edge layer 202 and a network layer 203;

[0252] The access layer 201 is used to: obtain energy data of each preset energy system in real time;

[0253] The edge layer 202 is used to: pre-process the acquired real-time energy data, and process the pre-processed real-time energy data to obtain a first data processing result;

[0254] The network layer 203 is used to transmit the first data processing result.

[0255] Further, it also includes the platform layer;

[0256] The platform layer is used to: process all energy data to obtain the second data processing result;

[0257] The network layer 203 is further used to transmit the second data processing result.

[0258] Furthermore, it also includes a perception terminal network including terminal nodes, central nodes and routing nodes;

[0259] The access layer 201 is also used for:

[0260] Using a distributed address allocation algorithm to allocate routing addresses to each routing node and each terminal node in the sensing terminal network;

[0261] The access layer 201 is also specifically used for:

[0262] The energy data of each preset energy system is collected in real time through each terminal node and sent to the central node through the corresponding routing node;

[0263] Receive energy data of each preset energy system sent by the central node.

[0264] Furthermore, the access layer 201 is also used for:

[0265] The optimal location of each terminal node and each routing node in the sensing terminal network is determined by a decision access algorithm.

[0266] It should be noted that the beneficial effects of the multi-energy data integrated management system provided by the above embodiment are the same as the beneficial effects of the multi-energy data integrated management method described above, which will not be repeated here. In addition, when the system provided by the above embodiment realizes its functions, it only takes the division of the above functional modules as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided by the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0267] Among them, a multi-energy data integrated management system 200 of the present invention can be a computer program (including program code) running in a computer device. For example, a multi-energy data integrated management system 200 of the present invention is an application software that can be used to execute corresponding steps in a multi-energy data integrated management method of the present invention.

[0268] In some embodiments, a multi-energy data integrated management system 200 of the present invention can be implemented by a combination of software and hardware. As an example, a multi-energy data integrated management system 200 of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute a multi-energy data integrated management method of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs) or other electronic components.

[0269] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The name of a module does not limit the module itself in some cases.

[0270] An electronic device according to an embodiment of the present invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the above-mentioned multi-energy data integrated management methods is implemented. That is to say, an electronic device according to an embodiment of the present invention may include but is not limited to: a processor and a memory; a memory for storing a computer program; a processor for executing the multi-energy data integrated management method shown in any embodiment of the present invention by calling a computer program.

[0271] Among them, the electronic device can also be a terminal device, and the terminal device can be any device that can install applications, including at least one of a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart car device.

[0272] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, and when the computer program is executed by a processor, any of the above-mentioned multi-energy data integrated management methods is implemented.

[0273] The computer-readable storage medium provided by the embodiment of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or component.

[0274] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0275] Those skilled in the art know that the present invention can be implemented as a system, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.

[0276] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A comprehensive management method for multi-energy data, characterized in that: include: Get real-time energy data for each preset energy system; Preprocessing the acquired real-time energy data, and processing the preprocessed real-time energy data to obtain a first data processing result; The first data processing result is transmitted.

2. A multi-energy data integrated management method according to claim 1, characterized in that: Also includes: Processing all energy data to obtain a second data processing result; The second data processing result is transmitted.

3. A multi-energy data integrated management method according to claim 1 or 2, characterized in that: Also includes: Construct a perception terminal network including terminal nodes, central nodes and routing nodes; Allocating a routing address to each routing node and each terminal node in the sensing terminal network using a distributed address allocation algorithm; Get real-time energy data for each preset energy system, including: The energy data of each preset energy system is collected in real time through each terminal node, and sent to the central node through the corresponding routing node; Receive energy data of each preset energy system sent by the central node.

4. A multi-energy data integrated management method according to claim 3, characterized in that: Also includes: The optimal location of each terminal node and each routing node in the sensing terminal network is determined by a decision access algorithm.

5. A comprehensive management system for multi-energy data, characterized in that: Includes access layer, edge layer and network layer; The access layer is used to: obtain energy data of each preset energy system in real time; The edge layer is used to: preprocess the acquired real-time energy data, and process the preprocessed real-time energy data to obtain a first data processing result; The network layer is used to transmit the first data processing result.

6. A multi-energy data integrated management system according to claim 5, characterized in that: It also includes the platform layer; The platform layer is used to: process all energy data to obtain a second data processing result; The network layer is also used to transmit the second data processing result.

7. A multi-energy data integrated management system according to claim 5 or 6, characterized in that: Also included is a sensing terminal network including terminal nodes, central nodes and routing nodes; The access layer is also used for: Allocating a routing address to each routing node and each terminal node in the sensing terminal network using a distributed address allocation algorithm; The access layer is also specifically used for: The energy data of each preset energy system is collected in real time through each terminal node, and sent to the central node through the corresponding routing node; Receive energy data of each preset energy system sent by the central node.

8. A multi-energy data integrated management system according to claim 7, characterized in that: The access layer is also used for: The optimal location of each terminal node and each routing node in the sensing terminal network is determined by a decision access algorithm.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, a multi-energy data integrated management method as claimed in any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for integrated management of multi-energy data according to any one of claims 1 to 4 is implemented.

Images