Time sequence simulation method for new energy cloud energy storage operation mechanism

Abstract

The invention relates to the technical field of power systems, in particular to a time sequence simulation method for a new energy cloud energy storage operation mechanism. Comprising the steps of collecting output data of the new energy power system, and cleaning, standardizing and formatting the collected data. Analyzing user behavior characteristics, and classifying the collected power consumption behaviors of the users; constructing a multi-temporal generative network model, and learning and pre-predicting prediction data of a future power system state from historical data; constructing a dynamic bandwidth allocation strategy according to the predicted data and real-time data of the predicted future power system state, and making a charging and discharging plan of the energy storage equipment; identifying by using the generated real-time data and the charging and discharging plan, and classifying the operation state of the power system; a multi-dimensional time sequence simulation model is constructed. Various factors influencing the operation of the cloud energy storage are considered, more accurate simulation prediction is provided, and the regulation and response capability of the power grid to the fluctuating energy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a timing simulation method for a new energy cloud energy storage operation mechanism. Background Art

[0002] New energy power systems, such as wind power and solar power generation, are becoming increasingly important as they provide a clean and renewable energy source. However, the volatility and unpredictability of these energy sources pose challenges to the stable operation of the power grid. Existing technologies generally lack effective strategies to manage the power supply security performance of these new energy power systems, especially during peak and trough periods of demand.

[0003] For example, the Chinese patent application number: 2021110623418 discloses a method and system for establishing a timing simulation model of a cloud energy storage operation mechanism, wherein the method includes: according to the demand side load conditions, the demand side users are divided into five categories of users, and different power supply combination modes are provided for each category of users; the real-time electricity price of the user during electricity use is obtained, and the power supply relationship between the user and the cloud energy storage is obtained according to the real-time electricity price and the cloud energy storage timing charging and discharging strategy; according to the power supply combination mode, the cloud energy storage timing charging and discharging strategy and the power supply relationship, a timing simulation model of the cloud energy storage operation mechanism is constructed. The present invention conducts targeted discussions in combination with different types of users, arranges a suitable combination of multiple energy storage devices, and reduces electricity costs while ensuring power balance; considers important influencing factors such as user electricity prices and load to design a cloud energy storage timing simulation model; clarifies the charging and discharging conditions of multiple types of cloud energy storage under different working conditions, and obtains the optimal charging and discharging strategy of cloud energy storage equipment.

[0004] The main defects of the above technology are:

[0005] Insufficient analysis of user electricity usage behavior: Existing technologies do not fully consider the differences in electricity usage behavior of different types of users, and lack in-depth analysis of user behavior characteristics and targeted power supply strategies.

[0006] The combination of energy storage devices is single: the advantages of multiple energy storage devices are not fully utilized, and there is a lack of optimization strategies for the combination of multiple devices.

[0007] Lack of real-time electricity price response: Existing technologies fail to dynamically adjust charging and discharging strategies based on real-time electricity price changes.

[0008] Limitations of the simulation model: A time series simulation model that comprehensively considers multiple dimensions such as energy storage equipment, user behavior, and power system status may not have been established. Summary of the invention

[0009] In view of the above deficiencies in the prior art, the technical problem to be solved by the present invention is: to provide a timing simulation method and system for the cloud energy storage operation mechanism of the new energy power system, comprehensively considering factors such as user behavior and energy storage equipment combination, and providing a more comprehensive, flexible and efficient cloud energy storage operation mechanism, which significantly improves the stability of the power system.

[0010] In order to achieve the above objectives, the present invention adopts the following technical solutions:

[0011] A time series simulation method for a new energy cloud energy storage operation mechanism includes the following:

[0012] S1. Use smart meters to monitor users’ electricity usage behavior data in real time, and use meteorological data API to obtain weather information related to electricity usage behavior. The electricity usage behavior data includes:

[0013] Peak and valley power consumption periods: distinguish the specific times of peak and valley power consumption;

[0014] Average daily electricity consumption: the total electricity consumption of users per day;

[0015] Seasonal electricity usage patterns: users’ electricity usage habits in different seasons;

[0016] Type and quantity of electrical equipment used: type and quantity of electrical appliances used by the user.

[0017] S2. User behavior feature analysis, using a hierarchical clustering algorithm to classify the user electricity usage behaviors in S1 and identify the characteristics and electricity usage patterns of different user groups;

[0018] S3. Construct a multi-phase generation network model. According to the results of the user behavior feature analysis in S2, a multi-phase generation network model is constructed. The multi-phase generation network model is used to learn from historical data and predict the future power system state;

[0019] S4. Build a dynamic bandwidth allocation strategy based on the predicted data of the future power system state predicted by S3 and the real-time data obtained by S1, and formulate a charging and discharging plan for the energy storage equipment;

[0020] S5. Classification and recognition of time series signals: using classification and recognition methods to identify the real-time data obtained in S1 and the charging and discharging plan generated in S4, integrating historical operation data, and classifying the operation status of the power system;

[0021] S6. Construction of a time series simulation model for the cloud energy storage operation mechanism. Combining the results of S1 to S5, a time series simulation model including multiple dimensions such as energy storage equipment, user behavior, and power system status is constructed;

[0022] The multi-dimensional time series simulation model established by S1 to S6 can comprehensively reflect the operating status of the cloud energy storage system, provide a simulation environment for testing and optimizing the cloud energy storage operation strategy, and achieve stable operation of the power system and efficient use of energy.

[0023] Furthermore, in S2, the hierarchical clustering algorithm includes:

[0024] S21. Select features that are helpful in distinguishing different user behaviors from the user electricity consumption behaviors collected in S1, such as peak and valley electricity consumption periods, average daily electricity consumption, seasonal electricity consumption patterns, etc.;

[0025] S22. Determine the number of clusters, and determine the number of clusters by silhouette coefficient analysis;

[0026] S23. Perform hierarchical clustering and use Euclidean distance to measure the similarity between samples. According to the selected distance metric, each user is regarded as an initial cluster, and then the most similar cluster pairs are gradually merged. The Euclidean distance formula is:

[0027]

[0028] In the formula: x and y are two samples, n is the number of features;

[0029] S24. Clustering result evaluation: use the silhouette coefficient, Davies-Bouldin index and other indicators to evaluate the clustering effect and ensure that the clustering results are meaningful;

[0030] The silhouette coefficient formula is: a is the average distance between a sample and the nearest sample in the same cluster, and b is the average distance between a sample and the nearest sample in another cluster;

[0031] The Davies-Bouldin index formula is: S i and S j is the sum of the internal distances of clusters i and j, d(C i ,C j ) is the distance between clusters i and j.

[0032] S25. Interpretation and application of results. Interpret the clustering results, identify the characteristics and power consumption patterns of each cluster, arrange the characteristics and power consumption patterns of each cluster in chronological order, and the characteristics at each time point should be recorded. Apply the clustering results to formulate targeted power supply strategies.

[0033] Furthermore, in S3, the multi-temporal generation network model includes the following types of layers:

[0034] Input layer: The input layer receives time series data, which is the data in S2 after user behavior feature analysis;

[0035] Circular layer: GRU is used as a circulatory layer to process time series data. The circulatory layer is used to process sequence data and capture dynamic features and long-term dependencies in the time series. The final output of the circulatory layer is a feature sequence.

[0036] Fully connected layer: Use the fully connected layer to integrate the output of the recurrent layer and map the output of the recurrent layer into a dense high-dimensional space for classification or regression tasks and learning high-level feature representation of the data;

[0037] Generating layer: The generating layer generates new data points or sequences based on the output of the fully connected layer.

[0038] Furthermore, the GRU is a popular recurrent neural network (RNN) variant, which is particularly suitable for processing and predicting time series data. The GRU includes:

[0039] Update gate: determines how much information is learned from the current input x t Flow into the hidden state h t ;

[0040] Reset gate: decides how to handle the hidden state h of the previous time step t-1 , allowing the network to reset or remember its previous state;

[0041] Candidate hidden state: is a candidate for the hidden state of the current time step, without considering the previous hidden state;

[0042] Final hidden state: is the final hidden state of the current time step, updated by combining the update gate, reset gate, and candidate hidden state, where:

[0043] Update Gate:

[0044] z t =σ(W z ·[h t-1 ,x t ]+b z );

[0045] By updating the gate formula, we know that: t is based on the hidden state h of the previous time step t-1 and the current input x t The weight matrix W of the gate is updated through the weight matrix z Perform a linear transformation and then add the bias term to update the bias term b of the gate z , and finally the output value z of the update gate is obtained by nonlinear transformation through the sinmoid function σ tWill determine the candidate hidden state at the current time step and the hidden state h at the previous time step t-1 How much information will be updated, that is, how much new information will be passed to the next time step;

[0046] Reset the gate:

[0047] r t =σ(W r ·[h t-1 ,x t ]+b r );

[0048] By resetting the gate formula, we know that: t is based on the hidden state h of the previous time step t-1 and the input data x at the current time step t By resetting the gate weight matrix W r , and then add the reset gate bias term b r , and finally obtained by nonlinear transformation through the sinmoid function σ; reset the output value r of the gate t When calculating the candidate hidden state at the current time step When the hidden state h of the previous time step is t-1 How much weight should be given if r t Close to 1,h t-1 In calculation will be retained if r t Close to 0, h t-1 will be ignored, allowing the network to reset its state.

[0049] Candidate hidden states:

[0050]

[0051] The candidate hidden state formula shows that the candidate hidden state is based on the hidden state h of the previous time step. t-1 , output r through the reset gate of the current time step t Adjust, add the current input x t The influence of, and the weight matrices W and W x The linear transformation of the bias term b is finally obtained by nonlinear transformation through the tanh function. Together with the update gate, it determines the final hidden state at the current time step.

[0052] Final hidden state:

[0053]

[0054] Represents candidate hidden states According to the output value z of the update gate t The complement of (1-z t ) is weighted contribution to the final hidden state;

[0055] z t ·h t-1 represents the hidden state h of the previous time step t-1 According to the output value z of the update gate t Update the gate output value z t ;

[0056] The final hidden state formula shows that: the final hidden state h of the current time step is obtained t , which contains both the candidate hidden states through the current time step The information of the previous time step is also retained. t-1 information;

[0057] In the formula, z t is the update gate output of the current time step, σ is the sinmoid activation function, whose output value ranges from 0 to 1 and is usually used in the gating mechanism, W z Update the gate weight matrix, W z It is used to linearly transform the input data and the hidden state of the previous time step, h t-1 is the hidden state of the previous time step, b z is the bias term of the update gate, x t is the input data of the current time step; r t is the reset gate output of the current time step, W r Reset the weight matrix of the gate, which is used to linearly transform the input data and the hidden state of the previous time step, h t-1 is the hidden state of the previous time step, b r is the bias term for the reset gate;

[0058] is the candidate hidden state of the current time step; tanh is a nonlinear activation function used to introduce nonlinear characteristics, and the output value range is between -1 and 1; W is the weight matrix used to linearly transform the input data and the previous hidden state; b is the bias term, b adds a constant offset to the candidate hidden state; W x For the current input x t The weight matrix, W x Used to process input data for the current time step.

[0059] h t is the final hidden state, Represents candidate hidden states According to the output value z of the update gate t The complement of (1-z t ) is weighted contribution to the final hidden state;

[0060] z t ·h t-1 represents the hidden state h of the previous time step t-1 According to the output value z of the update gate t Update the output value z of the gate t .

[0061] Furthermore, the calculation formula of the fully connected layer is:

[0062] p=f(W fc y+b fc );

[0063] y=F(H)=[h1,h2…h T ];

[0064] H={h1,h2…h T};

[0065] In the formula: p is the output of the fully connected layer, which is usually used as the feature representation of classification or regression tasks; H is the output sequence of the recurrent layer; T is the length of the sequence; each h t is the final hidden state of the current time step t; y is the flattened vector of the recurrent layer output sequence H; W fc is the weight matrix of the fully connected layer, W fc The dimension of b depends on the dimension of the input vector y and the dimension of the output of the fully connected layer; fc is the bias vector of the fully connected layer, b fc Dimension and W fc The output dimension of is the same as that of , and f is the activation function.

[0066] Furthermore, the formula for converting the output of the fully connected layer into the required data points or sequences is:

[0067]

[0068] p is the output of the fully connected layer, is the output of the generation layer, i.e., the predicted data point or sequence, W g is the weight matrix of the generating layer, b g is the bias term of the generated layer.

[0069] Furthermore, in S4, a method for constructing a dynamic bandwidth allocation strategy based on the predicted data of the future power system state and the real-time data to formulate a charging and discharging plan for the energy storage device is as follows:

[0070] S41. Generative layer output predicted by the multi-temporal generative network model in S3 Combined with real-time data r , to update the estimate of the power system state, the estimation formula is:

[0071]

[0072] p e is the updated estimate of the power system state, is the output of the generation layer, p r is real-time data, ω is a weight coefficient between 0 and 1, which is adjusted according to the credibility of the predicted data and the real-time data;

[0073] S42. Analyze power demand in different time periods and determine the difference in resource demand during peak and off-peak periods;

[0074] S43. According to the results of resource demand analysis, formulate a charging and discharging plan for energy storage equipment. The formula is:

[0075]

[0076] In the formula: S(t) is the charge and discharge plan, which determines whether to charge or discharge or remain unchanged according to the power system status; S c It is a charging plan, which is executed when the power demand is lower than the off-peak period; S d It is a discharge plan, which is executed when the power demand is higher than the peak period; v is the electricity demand during off-peak hours, P P is the peak electricity demand, p e An estimate of the updated power system state;

[0077] S44. Convert the charging and discharging plan into bandwidth allocation, that is, adjust the charging and discharging rate of the energy storage device to match the fluctuation of power demand. The conversion formula is:

[0078] B a =B b +(S c -S d )·δ s ;

[0079] In the formula, B a is the allocated bandwidth, B b is the basic bandwidth, ΔB is the bandwidth adjustment, δ s is the adjustment coefficient determined according to the charge and discharge plan, δ s Between 0 and 1;

[0080] S45. Implement specific charging and discharging operations according to the formulated charging and discharging plan for energy storage equipment and convert the charging and discharging plan into bandwidth allocation to ensure the stable operation of the power system.

[0081] Furthermore, in S5, the specific method for classifying the power system operating status is:

[0082] S51. The real-time monitored power system data is integrated with the pre-established charging and discharging plan to form a comprehensive data set, and features that are helpful for classification are extracted from the comprehensive data set. The comprehensive data set is represented as:

[0083] F = {x pc ,x ts ,x c / dr};

[0084] x pc is the power change, reflecting the fluctuation of power demand; x ts is a timestamp, which is used to consider the impact of time factors on power demand; x c / dr is the charge and discharge rate, the current charge and discharge status of the energy storage device;

[0085] S52. Using historical data and known power system status labels, train a classification model, apply the trained classification model to real-time data, and classify the system status;

[0086] S53. The results of S52 classification are evaluated using indicators such as accuracy and recall, where the calculation formulas for accuracy and recall are:

[0087]

[0088]

[0089] In the formula: A is the accuracy evaluation index; R is the recall evaluation index; P c is the number of correct classifications; K is the total number of classifications; G is the total number of samples with the actual state being this state. G refers to the total number of samples with the actual label being the positive class in the entire data set, including those correctly predicted by the model and those predicted incorrectly;

[0090] S54. Convert the classification results into status feedback of the power system to provide decision support for the cloud energy storage system and adjust the charging and discharging plan according to the classification results to adapt to changes in power demand.

[0091] Furthermore, the time series simulation model constructed by S6 including multiple dimensions such as energy storage equipment, user behavior, and power system status is:

[0092] M t+1 =L(M t ,Pn ,D t ,U t ,W t ,S t ,B t );

[0093] L is the state update function, which calculates the state of the next time step based on the current state and input variables;

[0094] M t is the state of the power system at time t, including power supply, demand, energy storage level, etc.; P n Time t is the output of the new energy power system; D t Time t load demand, that is, the power consumption of the power system, U t is the user behavior feature at time t identified in S2; W t is the weather condition at time t; S t is the charge and discharge plan at time t, B t is the bandwidth allocation at time t.

[0095] In summary, due to the adoption of the above technical solution, the beneficial technical effects of the invention are:

[0096] The method for establishing the time series simulation model of the cloud energy storage operation mechanism of the new energy power system adopts a user-customized power supply strategy, which is provided through detailed user classification and power supply mode, increasing the adaptability to user needs and the flexibility of the power supply strategy. At the same time, by optimizing the charging and discharging strategy and the combination of energy storage equipment, the stability and reliability of the power supply are improved. The optimization strategy helps to better integrate and utilize new energy such as wind power and photovoltaics.

[0097] The method for establishing a timing simulation model of the cloud energy storage operation mechanism of the new energy power system improves the utilization efficiency of the energy storage equipment and the regulation capacity of the power grid through the combination of multiple devices.

[0098] The method for establishing the timing simulation model of the cloud energy storage operation mechanism of the new energy power system comprehensively considers the various factors affecting the operation of cloud energy storage, provides more accurate simulation predictions, and improves the grid's regulation and response capabilities to fluctuating energy through the cloud energy storage timing simulation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 The flowchart is a timing simulation method for a new energy cloud energy storage operation mechanism. DETAILED DESCRIPTION

[0100] In order to make the purpose, technical solution and advantages of the invention more clear, the invention is further described in detail below in combination with the embodiments. It should be understood that the specific embodiments described here are only used to explain the invention and are not used to limit the invention.

[0101] like Figure 1 As shown, a time series simulation method of a new energy cloud energy storage operation mechanism includes the following:

[0102] S1. Use smart meters to monitor users’ electricity usage behavior data in real time, and use meteorological data API to obtain weather information related to electricity usage behavior. The electricity usage behavior data includes:

[0103] Peak and valley power consumption periods: distinguish the specific times of peak and valley power consumption;

[0104] Average daily electricity consumption: the total electricity consumption of users per day;

[0105] Seasonal electricity usage patterns: users’ electricity usage habits in different seasons;

[0106] Type and quantity of electrical equipment used: type and quantity of electrical appliances used by the user.

[0107] S1. Use smart meters to monitor users’ electricity usage behavior data in real time, and use meteorological data API to obtain weather information related to electricity usage behavior. The electricity usage behavior data includes:

[0108] Peak and valley power consumption periods: distinguish the specific times of peak and valley power consumption;

[0109] Average daily electricity consumption: the total electricity consumption of users per day;

[0110] Seasonal electricity usage patterns: users’ electricity usage habits in different seasons;

[0111] Type and quantity of electrical equipment used: type and quantity of electrical appliances used by the user.

[0112] S2. User behavior feature analysis, using a hierarchical clustering algorithm to classify the user electricity usage behaviors in S1 and identify the characteristics and electricity usage patterns of different user groups;

[0113] Through cluster analysis, user groups with similar electricity usage behaviors can be identified, providing accurate user behavior characteristics for subsequent power supply strategy formulation, which helps to achieve more personalized and efficient energy distribution.

[0114] In S2, the hierarchical clustering algorithm includes the following:

[0115] S21. Select features that are helpful in distinguishing different user behaviors from the electricity consumption behavior data obtained in S1, such as peak and valley electricity consumption periods, average daily electricity consumption, seasonal electricity consumption patterns, etc.;

[0116] S22. Determine the number of clusters, use a hierarchical clustering algorithm, calculate the silhouette coefficient under different cluster numbers (κ values), draw a relationship graph between the silhouette coefficient and the number of clusters (κ values), and select the κ value when the silhouette coefficient is the largest as the optimal number of clusters;

[0117] S23. Perform hierarchical clustering and use Euclidean distance to measure the similarity between samples. The smaller the distance, the more similar the electricity consumption behaviors between samples. According to the selected distance metric, each user is regarded as an initial cluster, and then the most similar cluster pairs are gradually merged. In the user electricity consumption behavior analysis, each sample represents the electricity consumption record of a user within a certain time range, which can be the electricity consumption per hour.

[0118] The Euclidean distance formula is:

[0119]

[0120] In the formula: x and y are two samples, n is the number of features;

[0121] S24. Clustering result evaluation: use the silhouette coefficient, Davies-Bouldin index and other indicators to evaluate the clustering effect and ensure that the clustering results are meaningful;

[0122] The silhouette coefficient formula is: a is the average distance between a sample and the nearest sample in the same cluster, b is the average distance between a sample and the nearest samples in other clusters, and a high silhouette coefficient indicates a high similarity of samples within a cluster;

[0123] The Davies-Bouldin index formula is: S i and S j is the sum of the internal distances of clusters i and j, d(C i ,C j ) is the distance between clusters i and j. A low Davies-Bouldin index indicates a high degree of separation between clusters.

[0124] S25. Interpretation and application of results. Interpret the clustering results and identify the characteristics and power consumption patterns of each cluster. The characteristics and power consumption patterns of each cluster are arranged in chronological order, and the characteristics at each time point should be recorded. The clustering results should be used to formulate targeted power supply strategies.

[0125] Analyze the characteristics of each cluster, such as peak electricity consumption hours, average daily electricity consumption, etc., and formulate differentiated power supply strategies for different user groups based on the different clustering results. For example, for user groups with high electricity consumption during peak hours, demand response incentives can be provided to encourage increased electricity consumption during off-peak hours. For users with significant seasonal changes in electricity consumption, seasonal electricity price discounts can be provided.

[0126] S3. Construct a multi-phase generation network model. According to the results of the user behavior feature analysis in S2, a multi-phase generation network model is constructed. The multi-phase generation network model is used to learn from historical data and predict the future power system state;

[0127] The multi-phase generation network can capture the timing characteristics of changes in the power system state and provide more accurate prediction results, thereby helping the cloud energy storage system to make more reasonable charging and discharging plans in advance and ensure that there are appropriate resources to respond when electricity demand peaks or troughs.

[0128] In the S3, the multi-temporal generative network model contains the following types of layers:

[0129] Input layer: The input layer receives time series data, which is the data in S2 that is arranged in chronological order after analyzing user behavior characteristics;

[0130] Circular layer: GRU is used as a circulatory layer to process time series data. The circulatory layer is used to process sequence data and capture dynamic features and long-term dependencies in the time series. The final output of the circulatory layer is a feature sequence.

[0131] Fully connected layer: Use the fully connected layer to integrate the output of the recurrent layer and map the output of the recurrent layer into a dense high-dimensional space for classification or regression tasks and learning high-level feature representation of the data;

[0132] Generating layer: The generating layer generates new data points or sequences based on the output of the fully connected layer.

[0133] The GRU is a popular recurrent neural network (RNN) variant, which is particularly suitable for processing and predicting time series data. The GRU includes:

[0134] Update gate: determines how much information is learned from the current input x t Flow into the hidden state h t ;

[0135] Reset gate: decides how to handle the hidden state h of the previous time step t-1 , allowing the network to reset or remember its previous state;

[0136] Candidate hidden state: is a candidate for the hidden state of the current time step, without considering the previous hidden state;

[0137] Final hidden state: is the final hidden state of the current time step, updated by combining the update gate, reset gate, and candidate hidden state, where:

[0138] Update Gate:

[0139] zt =σ(W z ·[h t-1 ,x t ]+b z );

[0140] By updating the gate formula, we know that: t is based on the hidden state h of the previous time step t-1 and the current input x t The weight matrix W of the gate is updated through the weight matrix z Perform a linear transformation and then add the bias term to update the bias term b of the gate z , and finally the output value z of the update gate is obtained by nonlinear transformation through the sinmoid function σ t Will determine the candidate hidden state at the current time step and the hidden state h at the previous time step t-1 How much information will be updated, that is, how much new information will be passed to the next time step;

[0141] Reset the gate:

[0142] r t =σ(W r ·[h t-1 ,x t ]+b r );

[0143] By resetting the gate formula, we know that: t is based on the hidden state h of the previous time step t-1 and the input data x at the current time step t By resetting the gate weight matrix W r , and then add the reset gate bias term b r , and finally obtained by nonlinear transformation through the sinmoid function σ; reset the output value r of the gate t When calculating the candidate hidden state at the current time step When the hidden state h of the previous time step is t-1 How much weight should be given if r t Close to 1,h t-1 In calculation will be retained if r t Close to 0, h t-1 will be ignored, allowing the network to reset its state.

[0144] Candidate hidden states:

[0145]

[0146] The candidate hidden state formula shows that the candidate hidden state is based on the hidden state h of the previous time step. t-1 , output r through the reset gate of the current time step t Adjust, add the current input x t The influence of, and the weight matrices W and W x The linear transformation of the bias term b is finally obtained by nonlinear transformation through the tanh function. Together with the update gate, it determines the final hidden state at the current time step.

[0147] Final hidden state:

[0148]

[0149] Represents candidate hidden states According to the output value z of the update gate t The complement of (1-z t ) is weighted contribution to the final hidden state;

[0150] z t ·h t-1 represents the hidden state h of the previous time step t-1 According to the output value z of the update gate t Update the gate output value z t ;

[0151] The final hidden state formula shows that: the final hidden state h of the current time step is obtained t , which contains both the candidate hidden states through the current time step The information of the previous time step is also retained. t-1 information.

[0152] In the formula, z t is the update gate output of the current time step, σ is the sinmoid activation function, whose output value ranges from 0 to 1 and is usually used in the gating mechanism, W z Update the gate weight matrix, W z It is used to linearly transform the input data and the hidden state of the previous time step, h t-1 is the hidden state of the previous time step, b z is the bias term of the update gate, x t is the input data of the current time step; r t is the reset gate output of the current time step, W r Reset the weight matrix of the gate, which is used to linearly transform the input data and the hidden state of the previous time step, h t-1 is the hidden state of the previous time step, b ris the bias term for the reset gate;

[0153] is the candidate hidden state of the current time step; tanh is a nonlinear activation function used to introduce nonlinear characteristics, and the output value range is between -1 and 1; W is the weight matrix used to linearly transform the input data and the previous hidden state; b is the bias term, b adds a constant offset to the candidate hidden state; W x For the current input x t The weight matrix, W x Used to process input data for the current time step.

[0154] h t is the final hidden state, Represents candidate hidden states According to the output value z of the update gate t The complement of (1-z t ) is weighted contribution to the final hidden state;

[0155] z t ·h t-1 represents the hidden state h of the previous time step t-1 According to the output value z of the update gate t Update the output value z of the gate t .

[0156] The calculation formula of the fully connected layer is:

[0157] p=f(W fc y+b fc );

[0158] y=F(H)=[h1,h2…h T ];

[0159] H={h1,h2…h T};

[0160] In the formula: p is the output of the fully connected layer, which is usually used as the feature representation of classification or regression tasks; H is the output sequence of the recurrent layer; T is the length of the sequence; each h t is the final hidden state of the current time step t; y is the flattened vector of the recurrent layer output sequence H; W fc is the weight matrix of the fully connected layer, W fc The dimension of b depends on the dimension of the input vector y and the dimension of the output of the fully connected layer; fc is the bias vector of the fully connected layer, b fc Dimension and W fc The output dimension of is the same as that of , and f is the activation function.

[0161] The formula for the generation layer to convert the output of the fully connected layer into the required data points or sequences is:

[0162]

[0163] p is the output of the fully connected layer, is the output of the generation layer, i.e., the predicted data point or sequence, W g is the weight matrix of the generating layer, b g is the bias term of the generated layer.

[0164] S4. Build a dynamic bandwidth allocation strategy based on the predicted data of the future power system state predicted by S3 and the real-time data obtained by S1, and formulate a charging and discharging plan for the energy storage equipment;

[0165] A method for building a dynamic bandwidth allocation strategy based on the predicted data and real-time data of the predicted future power system state and formulating a charging and discharging plan for energy storage equipment:

[0166] S41. Generative layer output predicted by the multi-temporal generative network model in S3 Combined with real-time data r , to update the estimate of the power system state, the estimation formula is:

[0167]

[0168] p e is the updated estimate of the power system state, is the output of the generation layer, p r is real-time data, ω is a weight coefficient between 0 and 1, which is adjusted according to the credibility of the predicted data and the real-time data;

[0169] S42. Analyze power demand in different time periods and determine the difference in resource demand during peak and off-peak periods;

[0170] S43. According to the results of resource demand analysis, formulate a charging and discharging plan for energy storage equipment. The formula is:

[0171]

[0172] In the formula: S(t) is the charge and discharge plan, which determines whether to charge or discharge or remain unchanged according to the power system status; S c It is a charging plan, which is executed when the power demand is lower than the off-peak period; S d It is a discharge plan, which is executed when the power demand is higher than the peak period; v P is the electricity demand during off-peak hours. P is the peak electricity demand, p e An estimate of the updated power system state;

[0173] S44. Convert the charging and discharging plan into bandwidth allocation, that is, adjust the charging and discharging rate of the energy storage device to match the fluctuation of power demand. The conversion formula is:

[0174] B a =B b +(S c -S d )·δ s ;

[0175] In the formula, B a is the allocated bandwidth, B b is the basic bandwidth, ΔB is the bandwidth adjustment, δ s is the adjustment coefficient determined according to the charge and discharge plan, δ s Between 0 and 1;

[0176] S45. Implement specific charging and discharging operations according to the formulated charging and discharging plan for energy storage equipment and convert the charging and discharging plan into bandwidth allocation to ensure the stable operation of the power system.

[0177] S5. Classification and recognition of time series signals: using classification and recognition methods to identify the real-time data obtained in S1 and the charging and discharging plan generated in S4, integrating historical operation data, and classifying the operation status of the power system;

[0178] S51. The real-time monitored power system data is integrated with the pre-established charging and discharging plan to form a comprehensive data set, and features that are helpful for classification are extracted from the comprehensive data set. The comprehensive data set is represented as:

[0179] F = {x pc ,x ts ,x c / dr};

[0180] x pc is the power change, reflecting the fluctuation of power demand; x ts is a timestamp, which is used to consider the impact of time factors on power demand; x c / dr is the charge and discharge rate, the current charge and discharge status of the energy storage device;

[0181] S52. Using historical data and known power system status labels, train a classification model, apply the trained classification model to real-time data, and classify the system status;

[0182] S53. The results of S52 classification are evaluated using indicators such as accuracy and recall, where the calculation formulas for accuracy and recall are:

[0183]

[0184]

[0185] In the formula: A is the accuracy evaluation index; R is the recall evaluation index; P c is the number of correct classifications; K is the total number of classifications; G is the total number of samples with the actual state being this state. G refers to the total number of samples with the actual label being the positive class in the entire data set, including those correctly predicted by the model and those predicted incorrectly;

[0186] S54. Convert the classification results into power system status feedback, provide decision support for the cloud energy storage system, and adjust the charging and discharging plan according to the classification results to adapt to changes in power demand;

[0187] By classifying and identifying the S51 to S54 timing signals, the current state of the power system can be quickly and accurately determined, providing timely status feedback and decision support for the operation of the cloud energy storage system.

[0188] S6. Construction of a time series simulation model for the cloud energy storage operation mechanism. Combining the results of S1 to S5, a time series simulation model including multiple dimensions such as energy storage equipment, user behavior, and power system status is constructed. The specific model is expressed as:

[0189] M t+1 =L(M t ,P n ,D t ,U t ,W t ,S t ,B t );

[0190] L is the state update function, which calculates the state of the next time step based on the current state and input variables;

[0191] M t is the state of the power system at time t, including power supply, demand, energy storage level, etc.; P n Time t is the output of the new energy power system; D t Time t load demand, that is, the power consumption of the power system, U t is the user behavior feature at time t identified in S2; W t is the weather condition at time t; S t is the charge and discharge plan at time t, B t is the bandwidth allocation at time t.

[0192] The multi-dimensional time series simulation model established by S1 to S6 can comprehensively reflect the operating status of the cloud energy storage system, provide a simulation environment for testing and optimizing the cloud energy storage operation strategy, and achieve stable operation of the power system and efficient use of energy.

[0193] This time series simulation model can fully reflect the operating status of the cloud energy storage system, provide a simulation environment for testing and optimizing cloud energy storage operation strategies, and help achieve stable operation of the power system and efficient use of energy.

[0194] The above description is a preferred embodiment of the invention and is not intended to limit the invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the invention should be included in the protection scope of the invention.

Claims

1. A time series simulation method for a new energy cloud energy storage operation mechanism, comprising the following: S1. Use smart meters to monitor users’ electricity usage behavior data in real time, and use meteorological data API to obtain weather information related to electricity usage behavior. The electricity usage behavior data includes: Peak and valley power consumption periods: distinguish the specific times of peak and valley power consumption; Average daily electricity consumption: the total electricity consumption of users per day; Seasonal electricity usage patterns: users’ electricity usage habits in different seasons; Type and quantity of electrical equipment: the type and quantity of electrical appliances used by the user; S2. User behavior feature analysis, using a hierarchical clustering algorithm to classify the user electricity usage behaviors in S1 and identify the characteristics and electricity usage patterns of different user groups; S3. Construct a multi-phase generation network model. According to the results of the user behavior feature analysis in S2, a multi-phase generation network model is constructed. The multi-phase generation network model is used to learn from historical data and predict the future power system state; S4. Build a dynamic bandwidth allocation strategy based on the predicted data of the future power system state predicted by S3 and the real-time data obtained by S1, and formulate a charging and discharging plan for the energy storage equipment; S5. Classification and recognition of time series signals: using classification and recognition methods to identify the real-time data obtained in S1 and the charging and discharging plan generated in S4, integrating historical operation data, and classifying the operation status of the power system; S6. Construction of a time series simulation model for the cloud energy storage operation mechanism. Combining the results of S1 to S5, a time series simulation model including multiple dimensions such as energy storage equipment, user behavior, and power system status is constructed; The multi-dimensional time series simulation model established by S1 to S6 can comprehensively reflect the operating status of the cloud energy storage system, provide a simulation environment for testing and optimizing the cloud energy storage operation strategy, and achieve stable operation of the power system and efficient use of energy.

2. According to claim 1, a timing simulation method for a new energy cloud energy storage operation mechanism is characterized in that: In S1, the hierarchical clustering algorithm includes: S21. Select features that are helpful in distinguishing different user behaviors from the user electricity consumption behaviors collected in S1, such as peak and valley electricity consumption periods, average daily electricity consumption, and seasonal electricity consumption patterns; S22. Determine the number of clusters by analyzing the silhouette coefficient; S23. Perform hierarchical clustering and use Euclidean distance to measure the similarity between samples. According to the selected distance metric, each user is regarded as an initial cluster, and then the most similar cluster pairs are gradually merged. The Euclidean distance formula is: In the formula: x and y are two samples, n is the number of features; S24. Clustering result evaluation: use the silhouette coefficient and Davies-Bouldin index to evaluate the clustering effect and ensure that the clustering results are meaningful; The silhouette coefficient formula is: a is the average distance between a sample and the nearest sample in the same cluster, and b is the average distance between a sample and the nearest sample in another cluster; The Davies-Bouldin index formula is: S i and S j is the sum of the internal distances between clusters i and j, d(C i ,C j ) is the distance between clusters i and j. S25. Interpretation and application of results. Interpret the clustering results, identify the characteristics and power consumption patterns of each cluster, arrange the characteristics and power consumption patterns of each cluster in chronological order, and the characteristics at each time point should be recorded. Apply the clustering results to formulate targeted power supply strategies.

3. According to claim 1, a timing simulation method for a new energy cloud energy storage operation mechanism is characterized in that: In the S3, the multi-temporal generative network model contains the following types of layers: Input layer: The input layer receives time series data, which is the data in S2 after user behavior feature analysis; Circular layer: GRU is used as a circulatory layer to process time series data. The circulatory layer is used to process sequence data and capture dynamic features and long-term dependencies in the time series. The final output of the circulatory layer is a feature sequence. Fully connected layer: Use the fully connected layer to integrate the output of the recurrent layer and map the output of the recurrent layer into a dense high-dimensional space for classification or regression tasks and learning high-level feature representation of the data; Generating layer: The generating layer generates new data points or sequences based on the output of the fully connected layer.

4. The timing simulation method of a new energy cloud energy storage operation mechanism according to claim 3 is characterized in that: The GRU is a popular recurrent neural network (RNN) variant, which is particularly suitable for processing and predicting time series data. The GRU includes: Update gate: determines how much information is learned from the current input x t Flow into the hidden state h t ; Reset gate: decides how to handle the hidden state h of the previous time step t-1 , allowing the network to reset or remember its previous state; Candidate hidden state: is a candidate for the hidden state of the current time step, without considering the previous hidden state; Final hidden state: is the final hidden state of the current time step, updated by combining the update gate, reset gate, and candidate hidden state, where: Update Gate: z t =σ(W z ·[h t-1 ,x t ]+b z ); By updating the gate formula, we know that: t is based on the hidden state h of the previous time step t-1 and the current input x t The weight matrix W of the gate is updated through the weight matrix z Perform a linear transformation and then add the bias term to update the bias term b of the gate z , and finally the output value z of the update gate is obtained by nonlinear transformation through the sinmoid function σ t Will determine the candidate hidden state at the current time step and the hidden state h at the previous time step t-1 How much information will be updated, that is, how much new information will be passed to the next time step; Reset the gate: r t =σ(W r ·[h t-1 ,x t ]+b r ); By resetting the gate formula, we know that: t is based on the hidden state h of the previous time step t-1 and the input data x at the current time step t By resetting the gate weight matrix W r , and then add the reset gate bias term b r , and finally obtained by nonlinear transformation through the sinmoid function σ; reset the output value r of the gate t When calculating the candidate hidden state at the current time step When the hidden state h of the previous time step is t-1 How much weight should be given if r t Close to 1,h t-1 In calculation will be retained if r t Close to 0, h t-1 will be ignored, allowing the network to reset its state. Candidate hidden states: The candidate hidden state formula shows that the candidate hidden state is based on the hidden state h of the previous time step. t-1 , output r through the reset gate of the current time step t Adjust, add the current input x t The influence of, and the weight matrices W and W x The linear transformation of the bias term b is finally obtained by nonlinear transformation through the tanh function. Together with the update gate, it determines the final hidden state at the current time step. Final hidden state: Represents candidate hidden states According to the output value z of the update gate t The complement of (1-z t ) is weighted contribution to the final hidden state; z t ·h t-1 represents the hidden state h of the previous time step t-1 According to the output value z of the update gate t Update the gate output value z t ; The final hidden state formula shows that: the final hidden state h of the current time step is obtained t , which contains both the candidate hidden states through the current time step The information of the previous time step is also retained. t-1 information; In the formula, z t is the update gate output of the current time step, σ is the sinmoid activation function, whose output value ranges from 0 to 1 and is usually used in the gating mechanism, W z Update the gate weight matrix, W z It is used to linearly transform the input data and the hidden state of the previous time step, h t-1 is the hidden state of the previous time step, b z is the bias term of the update gate, x t is the input data of the current time step; r t is the reset gate output of the current time step, W r Reset the weight matrix of the gate, which is used to linearly transform the input data and the hidden state of the previous time step, h t-1 is the hidden state of the previous time step, b r is the bias term for the reset gate; is the candidate hidden state of the current time step; tanh is a nonlinear activation function used to introduce nonlinear characteristics, and the output value range is between -1 and 1; W is the weight matrix used to linearly transform the input data and the previous hidden state; b is the bias term, b adds a constant offset to the candidate hidden state; W x For the current input x t The weight matrix, W x Used to process input data for the current time step. h t is the final hidden state, Represents candidate hidden states According to the output value z of the update gate t The complement of (1-z t ) is weighted contribution to the final hidden state; z t ·h t-1 represents the hidden state h of the previous time step t-1 According to the output value z of the update gate t Update the gate output value z t .

5. According to claim 3, a timing simulation method for a new energy cloud energy storage operation mechanism is characterized in that: The calculation formula of the fully connected layer is: p=f(W fc ·y+b fc ); y=F(H)=[h1,h2…h T ]; H={h1,h2…h T }; In the formula: p is the output of the fully connected layer, which is usually used as the feature representation of classification or regression tasks; H is the output sequence of the recurrent layer; T is the length of the sequence; each h t is the final hidden state at the current time step t; y is the flattened vector of the recurrent layer output sequence H; W fc is the weight matrix of the fully connected layer, W fc The dimension of b depends on the dimension of the input vector y and the dimension of the output of the fully connected layer; fc is the bias vector of the fully connected layer, b fc Dimension and W fc The output dimension of is the same as that of , and f is the activation function.

6. The timing simulation method of a new energy cloud energy storage operation mechanism according to claim 3 is characterized in that: The formula for the generation layer to convert the output of the fully connected layer into the required data points or sequences is: p is the output of the fully connected layer, is the output of the generation layer, i.e., the predicted data point or sequence, W g is the weight matrix of the generating layer, b g is the bias term of the generated layer.

7. The timing simulation method of a new energy cloud energy storage operation mechanism according to claim 1 is characterized in that: In S4, a method for constructing a dynamic bandwidth allocation strategy based on the predicted data of the future power system state and the real-time data to formulate a charging and discharging plan for the energy storage device is as follows: S41. Generative layer output predicted by the multi-temporal generative network model in S3 Combined with real-time data r , to update the estimate of the power system state, the estimation formula is: p e is the updated estimate of the power system state, is the output of the generation layer, p r is real-time data, ω is a weight coefficient between 0 and 1, which is adjusted according to the credibility of the predicted data and the real-time data; S42. Analyze power demand in different time periods and determine the difference in resource demand during peak and off-peak periods; S43. According to the results of resource demand analysis, formulate a charging and discharging plan for energy storage equipment. The formula is: In the formula: S(t) is the charge and discharge plan, which determines whether to charge or discharge or remain unchanged according to the power system status; S c It is a charging plan, which is executed when the power demand is lower than the off-peak period; S d It is a discharge plan, which is executed when the power demand is higher than the peak period; v P is the electricity demand during off-peak hours. P is the peak electricity demand, p e An estimate of the updated power system state; S44. Convert the charging and discharging plan into bandwidth allocation, that is, adjust the charging and discharging rate of the energy storage device to match the fluctuation of power demand. The conversion formula is: B a =B b +(S c -S d )·d s ; In the formula, B a is the allocated bandwidth, B b is the basic bandwidth, ΔB is the bandwidth adjustment, δ s is the adjustment coefficient determined according to the charge and discharge plan, δ s Between 0 and 1; S45. Implement specific charging and discharging operations according to the formulated charging and discharging plan for energy storage equipment and convert the charging and discharging plan into bandwidth allocation to ensure the stable operation of the power system.

8. The timing simulation method of a new energy cloud energy storage operation mechanism according to claim 1 is characterized in that: In S5, the specific method for classifying the power system operating status is: S51. The real-time monitored power system data is integrated with the pre-established charging and discharging plan to form a comprehensive data set, and features that are helpful for classification are extracted from the comprehensive data set. The comprehensive data set is represented as: F={x pc ,x ts ,x c / dr }; x pc is the power change, reflecting the fluctuation of power demand; x ts is a timestamp, which is used to consider the impact of time factors on power demand; x c / dr is the charge and discharge rate, the current charge and discharge status of the energy storage device; S52. Using historical data and known power system status labels, train a classification model, apply the trained classification model to real-time data, and classify the system status; S53. The results of S52 classification are evaluated using indicators such as accuracy and recall, where the calculation formulas for accuracy and recall are: In the formula: A is the accuracy evaluation index; R is the recall evaluation index; P c is the number of correct classifications; K is the total number of classifications; G is the total number of samples with the actual state being this state. G refers to the total number of samples with the actual label being the positive class in the entire data set, including those correctly predicted by the model and those predicted incorrectly; S54. Convert the classification results into status feedback of the power system to provide decision support for the cloud energy storage system and adjust the charging and discharging plan according to the classification results to adapt to changes in power demand.

9. The timing simulation method of a new energy cloud energy storage operation mechanism according to claim 1 is characterized in that: The time series simulation model constructed by S6 including multiple dimensions such as energy storage equipment, user behavior, and power system status is: M t+1 =L(M t ,P n ,D t ,U t ,W t ,S t ,B t ); L is the state update function, which calculates the state of the next time step based on the current state and input variables; M t is the state of the power system at time t, including power supply, demand, energy storage level, etc.; P n Time t is the output of the new energy power system; D t Time t load demand, that is, the power consumption of the power system, U t is the user behavior feature at time t identified in S2; W t is the weather condition at time t; S t is the charge and discharge plan at time t, B t is the bandwidth allocation at time t.

Images