Power grid provincial and regional information interaction method based on improved bat algorithm

Abstract

The invention discloses a power grid provincial and regional information interaction method based on an improved bat algorithm. The method comprises the following steps: constructing a corresponding interaction architecture of power grid provincial and regional demands; on the basis of building an interactive framework corresponding to provincial and regional demands of a power grid, an information interaction service scene needs to be determined, and in a typical demand response information interaction scene, terminals or systems of different demand responses use the same information exchange service during information exchange, and collected data are sent to an information interaction model; a province and region demand corresponding information interaction model is established; analyzing the data by adopting an SVM (Support Vector Machine), and carrying out anomaly detection on data anomalies; and an improved bat algorithm is adopted to realize parameter optimization of the SVM. Basic support is provided for improving the interaction quality of provincial and regional information, so that a globally optimal solution is better found, and the abnormal condition of power grid data can be accurately detected.

Description

Technical Field

[0001] The invention relates to the field of information interaction, in particular to a power grid land-saving information interaction method based on an improved bat algorithm. Background Art

[0002] The main function of demand response is to adjust the user load, which can effectively absorb new energy, optimize resource allocation, and reduce peak loads and fill valleys. In order to better meet the demand, my country is vigorously building a power demand response market and introducing adjustable control resources in the power balance to effectively reflect the role of reducing peak loads and filling valleys, and expand the scope of implementation of demand response.

[0003] The increasing number of smart power terminals connected to the power communication network and participating in demand response interactions will generate a large amount of demand response business data flow and cause network congestion and delay, making the communication transmission quality worse. If the emergency control business in the DR business is not responded to in time, it may affect the safe operation of the power grid.

[0004] Therefore, how to reduce end-to-end transmission delay and delay jitter, reduce the packet loss rate during data transmission, and effectively solve the link congestion problem during demand response communication has become the current focus. The core issue of provincial and local information interaction is how to ensure the quality of information transmission to meet user needs; in the process of demand response information interaction, the reliability of information transmission should be guaranteed to the greatest extent.

[0005] Since the classification results of SVM are determined by its parameters, and traditional manual parameter selection is usually not the optimal parameter, the classification effect of SVM is difficult to meet the needs of practical applications. Summary of the invention

[0006] The purpose of the present invention is to solve the above problems existing in the prior art and to provide a power grid land-saving information interaction method based on an improved bat algorithm.

[0007] The technical solution of the present invention is: a method for interacting power grid land-saving information based on an improved bat algorithm, comprising:

[0008] Step 1: Build a grid provincial demand response interactive framework, which includes power suppliers, power regulators, end users, and aggregators. The power regulator is mainly responsible for supervising and controlling the response of power demand, the aggregator integrates the DR resources of users, and the power supplier is responsible for guidance and management;

[0009] Step 2: On the basis of building the interactive architecture corresponding to the provincial and local demands of the power grid, it is necessary to determine the information interaction business scenario. In a typical demand response information interaction scenario, different demand response terminals or systems use the same information exchange service when exchanging information, and provide the collected data to the information interaction model; therefore, a provincial and local demand corresponding information interaction model is established;

[0010] Step 3: In the established provincial and local demand corresponding information interaction model, real-time data is included. Based on security and real-time considerations, the data is intelligently processed in real time to find abnormal information. SVM is used to analyze the data and perform anomaly detection on data anomalies.

[0011] Step 4 uses the improved bat algorithm to optimize the parameters of SVM.

[0012] Further, step 1 includes:

[0013] Step 1.1 First, the power supplier determines the response requirements of the power system, thereby determining the relevant response plan; after determining the response plan, the response plan is provided to the aggregator; the aggregator obtains the response plan from the power supplier, and the power supplier manages the planning process;

[0014] Step 1.2 After receiving the plan, the aggregator formulates a corresponding DR plan and publishes the plan to users to stimulate users to participate in demand response; users set the conditions for participating in demand response according to the benefits they can obtain in the plan and selectively participate in demand response; after receiving the DR signal, users choose DR services and sign contracts based on the actual power supply and available response time;

[0015] Step 1.3 Finally, the aggregator and service provider manage the contract and evaluate and settle the user's participation in demand response.

[0016] Furthermore, the specific process of step 2 to establish the information interaction model corresponding to provincial and local needs is:

[0017] Step 2.1 In the demand response system, the general demand response service management system or demand response aggregator is the upper node UN, and the demand response end user or small aggregator is the lower node DN; in the information interaction model, UN is equivalent to the server, and DN is equivalent to the client; N performs operations on n DNs as a whole, sends signals to each DN unit, and the DN unit executes them;

[0018] The demand response service provider generally acts as an upper node UN to generate demand response signals; the aggregator, as the middle layer, can act as both an upper node UN and a lower node DN, and is divided into large and small aggregators; when the demand response service provider interacts with the aggregator, the aggregator acts as a lower node DN; when the aggregator interacts with the end user, the aggregator acts as an upper node UN; large aggregators can directly participate in demand response, while small aggregators need to aggregate loads before participating in demand response; the end user acts as a lower node DN;

[0019] Step 2.2 establishes a periodic data flow model F, F = (L, B, E, D), where L is the length of the DR periodic data message, B is the end-to-end delay of the message, E is the arrival time interval of the data flow, and D is the maximum delay for the service to reach the destination node. The value of the end-to-end delay must be less than or equal to the maximum delay for the service to reach the destination node.

[0020] Further, step 3 includes:

[0021] Step 3.1 Set the n-dimensional input x in the acquired raw data i With y j , and {x i ,y j}, i, j = 1, 2, ..., n; then the SVM is defined as:

[0022]

[0023] in is the feature mapping function of SVM, that is, the regression function of the sample, which can be characterized as

[0024]

[0025] Among them, ||ω|| 2 is a complexity parameter, which is used to characterize the complexity of F(·); C is a penalty factor, which is used to express the penalty level of SVM for classification errors and improve the probability of finding the global optimal solution; is the empirical risk value, which is used to characterize the error between the predicted result and the actual result. If ε represents a slack variable, which is used to eliminate the regional error of the actual value, it conforms to the following formula:

[0026] |yF(x)| ε =max{0,|yF(x)|-ε}

[0027] therefore, Can be rewritten as

[0028]

[0029] Step 3.2, combined, we can get:

[0030]

[0031] Substituting the Lagrange multiplier method into the above formula, the dual optimization problem can be solved:

[0032]

[0033] Among them, K(x i ,y j ) represents the kernel function of SVM; RBF kernel function is selected as the kernel function of SVM, as shown below:

[0034] K(x i ,y j ) = exp(-γ||x i -y j || 2 Step 3.3 Set the solution format to (a, a*) and substitute:

[0035]

[0036] in, Substitute it into the above formula F(x),

[0037]

[0038] Achieve binary classification of data samples, where F(x) is either +1 or -1. The sample type can be determined by the classification result, thereby detecting abnormal data.

[0039] Further, step 4 includes the following steps:

[0040] Step 4.1 Set the basic parameters of the bat algorithm, including the population size S p , the dimension of the population D, the upper limit of iteration I max , impulse loudness A(0), emissivity R(0), A(0) decay coefficient A f , the enhancement coefficient R of R(0) f And the pulse frequency {f min , f max};

[0041] Step 4.2: Initialize the bat population, calculate and evaluate the initial positions of individual bats in the population, calculate the fitness value of the individual and find the existing optimal solution;

[0042] Step 4.3 updates the bat's transmission pulse frequency, flight speed and spatial position and optimizes them;

[0043] Step 4.4: According to the conditions obtained from step 4.1 to step 4.3, determine (rand <Ai )∩(f(x i ) <f(x * )) is true, if so, then update the loudness pulse emission rate according to the following formula

[0044]

[0045]

[0046] Among them, f(x i ) represents the fitness of the current bat individual, f(x * ) represents the optimal fitness of space;

[0047] Step 4.5 is based on whether the iteration limit is reached. If so, the optimal fitness value is output; otherwise, jump to step 6 to continue to search for the best result; and then detect the abnormal data collected by the interactive information system;

[0048] Step 4.6 updates the flight speed and position of the individuals in the population according to the following formula. The update calculation formula is as follows:

[0049] f i =f min +β(f max -f min );

[0050]

[0051]

[0052]

[0053] In the formula, f i 、f min 、f max represent the current, minimum and maximum pulse transmission frequencies respectively; β represents a random number in the range of (0,1), ω represents the rate weight parameter, and v i represents the real-time rate of the current individual, x i represents the current position of the i-th individual, x * represents the optimal solution in space, and t represents the current number of iterations;

[0054] Step 4.7 If the current bat individual pulse emission rate satisfies the random number, the individual will perform local optimization according to the following formula:

[0055] The beneficial effects of the present invention are:

[0056] The physical architecture model and data flow model of demand response under smart grid are established, the demand response types and services are analyzed and the demand response services are divided into 4 levels, and several typical demand response information interaction scenarios are summarized. This provides basic support for the optimization strategy to improve the transmission quality of provincial and local information interaction and solve the problems of link congestion, packet loss and large delay in the provincial and local information interaction process.

[0057] The data of the power grid is collected and analyzed to optimize power consumption, detect power consumption status and fault status, and thus ensure the safe operation of the power grid. On the user side, the data of power users can be collected and analyzed, so that users can interact with the power grid and achieve the purpose of information interaction. On the basis of considering the two-way interaction between the power grid and users, the interaction of power consumption information between the power grid and users can be realized, thereby ensuring the safety of power supply and being conducive to energy saving and consumption reduction. At the same time, a periodic data flow model and a corresponding information interaction model of provincial and local demand are established. Based on the established information interaction model, the corresponding types of demand and services are analyzed respectively, and the obtained data stream is analyzed to provide basic support for improving the interaction quality of provincial and local information. Based on security and real-time considerations, the improved bat algorithm is used to realize the parameter optimization of SVM. The improvement process is mainly to reconstruct the local optimization formula of the standard individual in the BA population. As the number of iterations gradually increases, the forward speed of the bat will gradually slow down. And the forward speed is dynamically adjusted according to the existing individual flight status, so as to better find the global optimal solution.

[0058] When detecting anomalies in the power consumption data of power grid users under the framework of the real-time information interaction system of the power grid, the improved bat algorithm is used. Compared with other methods, the proposed method can more accurately detect anomalies in power grid data. It can be suitable for intelligent data processing of real-time interactive power grid information system and has good practical engineering application capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flow chart of the present invention;

[0060] Figure 2 It is a schematic diagram of the interactive architecture corresponding to the land-saving demand of the power grid constructed by the present invention;

[0061] Figure 3 The land-saving demand corresponding information interaction model of the present invention

[0062] Figure 4 2 is a comparison diagram of the embodiments of the present invention. DETAILED DESCRIPTION

[0063] A method for power grid land-saving information interaction based on an improved bat algorithm, such as Figure 1 As shown, the process is as follows:

[0064] Step 1: Build an interactive framework corresponding to the grid land saving needs, such as Figure 2 As shown, the interactive architecture includes power suppliers, power regulators, end users, and aggregators, among which the power regulators are mainly responsible for supervising and controlling the response of power demand, the aggregators integrate the DR resources of users, and the power suppliers are responsible for guidance and management;

[0065] Step 1.1 First, the power supplier determines the response requirements of the power system, thereby determining the relevant response plan; after determining the response plan, the response plan is provided to the aggregator; the aggregator obtains the response plan from the power supplier, and the power supplier manages the planning process;

[0066] Step 1.2 After receiving the plan, the aggregator formulates a corresponding DR plan and publishes the plan to users to stimulate users to participate in demand response; users set the conditions for participating in demand response according to the benefits they can obtain in the plan and selectively participate in demand response. After receiving the DR signal, users choose DR services and sign contracts based on the actual power supply and available response time;

[0067] Step 1.3 Finally, the aggregator and service provider manage the contract and evaluate and settle the user's participation in demand response.

[0068] Step 2: Based on the interactive architecture of the power grid provincial and local demand, it is necessary to determine the information interaction business scenario. In the typical demand response information interaction scenario, different demand response terminals or systems use the same information exchange service when exchanging information. The information model is the exchange entity in the information exchange service, and the collected data is given to the information interaction model. Therefore, the provincial and local demand corresponding information interaction model is established;

[0069] The specific process of establishing the information interaction model corresponding to provincial and local needs is as follows: Figure 3 As shown:

[0070] Step 2.1 In the demand response system, the general demand response service management system or demand response aggregator is the upper node UN, and the demand response end user or small aggregator is the lower node DN. In the information interaction model, UN is equivalent to the server, and DN is equivalent to the client.

[0071] Among them, the demand response service provider generally acts as an upper node UN to generate demand response signals. Aggregators, as the middle layer, can act as both upper nodes UN and lower nodes DN, and are divided into large and small aggregators. When the demand response service provider interacts with the aggregator. The aggregator acts as a lower node DN. When the aggregator interacts with the end user, the aggregator acts as an upper node UN. Large aggregators can directly participate in demand response, while small aggregators need to aggregate loads before participating in demand response. End users act as lower nodes DN;

[0072] Step 2.2 establishes a periodic data flow model F, F = (L, B, E, D), where L is the length of the DR periodic data message, B is the end-to-end delay of the message, E is the arrival time interval of the data flow, and D is the maximum delay for the service to reach the destination node. The value of the end-to-end delay must be less than or equal to the maximum delay for the service to reach the destination node.

[0073] Step 3: In the established provincial and local demand corresponding information interaction model, real-time data is included. Based on security and real-time considerations, the data needs to be processed intelligently in real time to effectively discover abnormal information. Therefore, SVM is used to analyze the data and perform anomaly detection on data anomalies.

[0074] Step 3.1 Set the n-dimensional input x in the acquired raw data i With y j , and {x i ,y j}, i, j = 1, 2, ..., n. Then SVM is defined as:

[0075]

[0076] in is the feature mapping function of SVM, that is, the regression function of the sample, which can be characterized as

[0077]

[0078] Among them, ||ω|| 2 is a complexity parameter, which is used to characterize the complexity of F(·); C is a penalty factor, which is used to express the penalty level of SVM for classification errors and improve the probability of finding the global optimal solution; is the empirical risk value, which is used to characterize the error between the predicted result and the actual result. If ε represents a slack variable, which is used to eliminate the regional error of the actual value, it conforms to the following formula:

[0079] |yF(x)| ε =max{0,|yF(x)|-ε}

[0080] therefore, Can be rewritten as

[0081]

[0082] Step 3.2, combined, we can get:

[0083]

[0084] Substituting the Lagrange multiplier method into the above formula, the dual optimization problem can be solved:

[0085]

[0086] Among them, K(x i ,y j ) represents the kernel function of SVM. This paper selects RBF kernel function as the kernel function of SVM, which is as follows: K(x i ,y j ) = exp(-γ||x i -y j || 2

[0087] Step 3.3 Set the solution format to (a, a*) and substitute:

[0088]

[0089] in, Substitute it into the above formula F(x),

[0090]

[0091] This achieves a binary classification solution for the data sample, where F(x) is either +1 or -1. The sample type can be determined by the classification result, thereby detecting abnormal data.

[0092] Step 4: Since the classification result of SVM is determined by its parameters, and the traditional manual parameter selection is usually not the optimal parameter, the classification effect of SVM is difficult to meet the needs of practical applications. Therefore, the improved bat algorithm is used to achieve SVM parameter optimization.

[0093] Step 4.1 Set the basic parameters of the bat algorithm, including the population size S p , the dimension of the population D, the upper limit of iteration I max , impulse loudness A(0), emissivity R(0), A(0) decay coefficient A f , the enhancement coefficient R of R(0) f And the pulse frequency {f min , f max};

[0094] Step 4.2: Initialize the bat population, calculate and evaluate the initial positions of individual bats in the population, calculate the fitness value of the individual and find the existing optimal solution;

[0095] Step 4.3 updates the bat's transmission pulse frequency, flight speed and spatial position and optimizes them;

[0096] Step 4.4: According to the conditions obtained from step 4.1 to step 4.3, determine (rand i )∩(f(x i ) <f(x * )) is true, if so, then update the loudness pulse emission rate according to the following formula

[0097]

[0098]

[0099] Among them, f(x i ) represents the fitness of the current bat individual, f(x * ) represents the optimal fitness of space;

[0100] Step 4.5 is based on whether the iteration limit is reached. If so, the optimal fitness value is output; otherwise, jump to step 4.6 and continue to search for the best result; and then detect the abnormal data collected by the interactive information system.

[0101] Step 4.6 updates the flight speed and position of the individuals in the population according to the following formula. The update calculation formula is as follows:

[0102] f i =f min +β(f max -f min );

[0103]

[0104]

[0105]

[0106] In the formula, f i 、f min 、f max Respectively represent the current, minimum and maximum pulse transmission frequencies. β represents a random number in the range of (0,1), ω represents the rate weight parameter, and v i represents the real-time rate of the current individual, x i represents the current position of the i-th individual, x * represents the optimal solution in space, and t represents the current number of iterations; ​

[0107] Step 4.7 If the current bat individual pulse emission rate satisfies the random number, the individual will perform local optimization according to the following formula:

[0108] IBA-SVM is applied to intelligent data processing of power grid to detect abnormal data collected by interactive information system.

[0109] Data selection: The open source data on power grid users' electricity consumption was made public by the Irish Electricity Company in 2012. The electricity consumption data of 1,800 power users within 300 days from 2009 to 2010 were selected as the basic data set, with the ratio of training set to test set being 9:1.

[0110] In order to test the effectiveness of the proposed method, IBA-SVM is compared with BP neural network, extreme learning machine and SVM. The classification results of the test set are shown in Figure 3 shown.

[0111] from Figure 3 It can be seen that when the power consumption data of power grid users is detected under the framework of the real-time interactive power grid information system, the detection accuracy of BPNN is 77.22%, the detection accuracy of ELM is 87.22%, the detection accuracy of SVM is 90.56%, and the accuracy of the proposed method is 94.44%. Compared with other methods, the proposed method can detect the abnormal situation of power grid data more accurately. It can be applied to the intelligent data processing of the real-time interactive power grid information system, so it has good engineering practical application ability.

[0112] In order to further demonstrate the advantages of this method, the above methods are compared with this method in terms of computational efficiency, and the computational time of the test set detection is used as the measurement scale of computational efficiency. The results are as follows: Figure 4 shown.

[0113] Depend on Figure 4 It can be seen that the computation time of this method is less than that of BPNN and ELM algorithms, and is basically close to that of SVM. This shows that this method can improve the computation efficiency while ensuring the computation accuracy, so it can efficiently realize the intelligent processing of power grid data, and the reliability of engineering applications is also high.

[0114] The above are only specific embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for interacting power grid land information based on an improved bat algorithm, characterized in that the steps are: as follows: Step 1: Build a grid provincial demand response interactive framework, which includes power suppliers, power regulators, end users, and aggregators. The power regulator is mainly responsible for supervising and controlling the response of power demand, the aggregator integrates the DR resources of users, and the power supplier is responsible for guidance and management; Step 2: Based on the interactive architecture corresponding to the grid and land demand, it is necessary to determine the information interaction business scenario. In a typical demand response information interaction scenario, different demand response terminals or systems use the same information exchange service to send the collected data to the information interaction model during information exchange. Step 3: In the established provincial and local demand corresponding information interaction model, real-time data is included. Based on security and real-time considerations, the data needs to be processed intelligently in real time, and SVM is used to analyze the data and perform anomaly detection on data anomalies; Step 4 uses the improved bat algorithm to optimize the parameters of SVM.

2. The method for interacting power grid land information based on the improved bat algorithm according to claim 1 is characterized in that: Step 1 includes: Step 1.1 First, the power supplier determines the response requirements of the power system, thereby determining the relevant response plan; after determining the response plan, the response plan is provided to the aggregator; the aggregator obtains the response plan from the power supplier, and the power supplier manages the planning process; Step 1.2 After receiving the plan, the aggregator formulates a corresponding DR plan and publishes the plan to users to stimulate users to participate in demand response; users set the conditions for participating in demand response according to the benefits they can obtain in the plan and selectively participate in demand response; after receiving the DR signal, users choose DR services and sign contracts based on the actual power supply and available response time; Step 1.3 Finally, the aggregator and service provider manage the contract and evaluate and settle the user's participation in demand response.

3. The power grid land information interaction method based on the improved bat algorithm according to claim 1 is characterized in that: The specific process of establishing the information interaction model corresponding to provincial and local needs in step 2 is: Step 2.1 In the demand response system, the general demand response service management system or demand response aggregator is the upper node UN, and the demand response end user or small aggregator is the lower node DN; in the information interaction model, UN is equivalent to the server, and DN is equivalent to the client; N performs operations on n DNs as a whole, sends signals to each DN unit, and the DN unit executes them; The demand response service provider generally acts as the upper node UN, generating demand response signals; the aggregator, as the middle layer, can act as both the upper node UN and the lower node DN, and is divided into large and small aggregators; When the demand response service provider interacts with the aggregator, the aggregator acts as the lower node DN. When the aggregator interacts with the end user, the aggregator acts as the upper node UN. Large aggregators can directly participate in demand response, while small aggregators need to aggregate loads before participating in demand response. The end user acts as the lower node DN. Step 2.2 establishes a periodic data flow model F, F = (L, B, E, D), where L is the length of the DR periodic data message, B is the end-to-end delay of the message, E is the arrival time interval of the data flow, and D is the maximum delay for the service to reach the destination node. The value of the end-to-end delay must be less than or equal to the maximum delay for the service to reach the destination node.

4. The method for interacting power grid land-saving information based on the improved bat algorithm according to claim 1 is characterized in that: Step 3 includes: Step 3.1 Set the n-dimensional input x in the acquired raw data i With y j , and {x i ,y j }, i, j = 1, 2, ..., n; then the SVM is defined as: in is the feature mapping function of SVM, that is, the regression function of the sample, which can be characterized as Among them, ||ω|| 2 is a complexity parameter, which is used to characterize the complexity of F(·); C is a penalty factor, which is used to express the penalty level of SVM for classification errors and improve the probability of finding the global optimal solution; is the empirical risk value, which is used to characterize the error between the predicted result and the actual result. If ε represents a slack variable, which is used to eliminate the regional error of the actual value, it conforms to the following formula: |y-|F(x)| ε =max{0,|y-F(x)|-ε} therefore, Can be rewritten as Step 3.2, combined, we can get: Substituting the Lagrange multiplier method into the above formula, the dual optimization problem can be solved: Among them, K(x i ,y j ) represents the kernel function of SVM; RBF kernel function is selected as the kernel function of SVM, as shown below: K(x i ,y j )=exp(-γ||x i -y j || 2 Step 3.3 Set the solution format to (a, a*) and substitute: in, Substitute it into the above formula F(x), It realizes binary classification of data samples, where F(x) is either +1 or -1. The sample type can be determined by the classification result, thus detecting abnormal data.

5. The method for interacting power grid land information based on the improved bat algorithm according to claim 1 is characterized in that: Step 4 includes: Step 4.1 Set the basic parameters of the bat algorithm, including the population size S p , the dimension of the population D, the upper limit of iteration I max , impulse loudness A(0), emissivity R(0), A(0) decay coefficient A f , the enhancement coefficient R of R(0) f And the pulse frequency {f min , f max }; Step 4.2: Initialize the bat population, calculate and evaluate the initial positions of individual bats in the population, calculate the fitness value of the individual and find the existing optimal solution; Step 4.3 updates the bat's transmission pulse frequency, flight speed and spatial position and optimizes them; Step 4.4: According to the conditions obtained from step 4.1 to step 4.3, determine (rand i )∩(f(x i ) <f(x * )) is true, if so, then update the loudness pulse emission rate according to the following formula​ Among them, f(x i ) represents the fitness of the current bat individual, f(x * ) represents the optimal fitness of space; Step 4.5 is based on whether the iteration limit is reached. If so, the optimal fitness value is output; otherwise, jump to step 6 to continue to search for the best result; and then detect the abnormal data collected by the interactive information system; Step 4.6 updates the flight speed and position of the individuals in the population according to the following formula. The update calculation formula is as follows: f i =f min +β(f max -f min ); In the formula, f i 、f min 、f max represent the current, minimum and maximum pulse transmission frequencies respectively; β represents a random number in the range of (0,1), ω represents the rate weight parameter, and v i represents the real-time rate of the current individual, x i represents the current position of the i-th individual, x * represents the optimal solution in space, and t represents the current number of iterations; Step 4.7 If the current bat individual pulse emission rate satisfies the random number, the individual will perform local optimization according to the following formula:

Images