A Method for Energy Storage Location and Capacity Planning Considering Distribution Network Zoning

Abstract

The present invention discloses a method for planning the location and capacity of energy storage considering distribution network partitioning, including the steps of: establishing a mathematical model according to the optimization objective of the energy storage system; classifying load and renewable energy output data by using least squares support vector machines according to the output and consumption of nodes; partitioning the distribution network to determine the distribution area of the energy storage system; establishing a deep neural network with residual blocks in each area to determine the location and capacity of the energy storage device. This application considers decomposing the large-scale energy storage optimization problem into several sub-problems to meet the configuration requirements of large-scale energy storage systems under the access of a high proportion of renewable energy, and can effectively absorb the output of renewable energy under different energy storage configurations.

Description

Technical Field

[0001] The present invention relates to the technical field of siting and sizing of energy storage devices in distribution networks, and particularly relates to a method for planning the location and capacity of energy storage considering distribution network zoning. Background Art

[0002] With the large-scale grid connection of renewable energy, there may be a situation of a large amount of energy waste. A reasonably sited energy storage system can play a role in flexibly absorbing and releasing electric energy. Due to the high investment price of energy storage devices, it is necessary to reasonably set the capacity to shorten the cost recovery period. At the same time, the investment in energy storage devices can reduce network losses and is beneficial to peak shaving and valley filling of the power grid.

[0003] Existing methods for siting and sizing energy storage devices have some defects in terms of consideration of influencing factors and calculation methods. First, the volatility and randomness of renewable energy output are not considered, and the impact of large-scale grid connection of wind power and photovoltaic power on the distribution network is ignored. Second, as the scale of the distribution network continues to increase, the complexity of energy storage configuration calculation is increased. The current algorithms for siting and sizing energy storage devices are computationally complex and lack the optimization speed and accuracy of results. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for planning the location and capacity of energy storage considering distribution network zoning. The present application establishes a mathematical model according to the optimization objective of the energy storage system, classifies load and wind-solar output data using the least squares support vector machine based on node output and consumption conditions, divides the distribution network according to the classification results to determine the distribution area of the energy storage system, and establishes a deep neural network and introduces residual blocks in each area to determine the location and capacity of the energy storage at the access nodes with the minimum total economic cost as the goal, so as to obtain different energy storage configurations.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for planning the location and capacity of energy storage considering distribution network zoning, comprising the following steps:

[0007] S1: Establish a mathematical model of the distribution network according to the optimization objective of the energy storage system;

[0008] S2: Classify the node load and renewable energy output data in the distribution network using the least squares support vector machine;

[0009] S3: Divide the distribution network according to the classification situation to determine the distribution area of the energy storage system;

[0010] S4: In each divided distribution area, establish a neural network with residual blocks, determine the access node location and capacity size of the energy storage system in each distribution area based on the application of the neural network, and thus obtain multiple different energy storage configuration schemes corresponding to multiple distribution areas;

[0011] S5: Calculate the economic benefits under different energy storage configurations for analysis.

[0012] Specific measures taken to optimize the above technical solutions also include:

[0013] Furthermore, the optimization objectives of the energy storage system described in step S1 are as follows:

[0014] Set the function of the optimization objective as the overall power throughput situation f of the energy storage system over a period of time, and the calculation formula is as follows:

[0015]

[0016] In the formula, Ns represents the number of energy storage systems during operation; t0 represents the start time of the energy storage system operation; n represents the number of energy storage systems during operation; Δt represents the time period from the start time to the end time of the energy storage system operation; t0 + nΔt represents the time when the energy storage system ends charge and discharge; represents the charge and discharge power of the kth energy storage system at time t.

[0017] Furthermore, the established mathematical model in step S1 also considers constraint conditions, economic costs, and economic benefits;

[0018] The constraint conditions include node voltage constraints, power balance constraints, power flow constraints, wind and solar power output constraints, and energy storage operation balance;

[0019] Node voltage constraint:

[0020] V min ≤V i ≤V max

[0021] In the formula, V min represents the lower limit of the node voltage; V max represents the upper limit of the node voltage; V i represents the voltage of node i;

[0022] Power balance constraint P:

[0023]

[0024] In the formula, represents the active power load of node i in the distribution network; represents the active power output of the jth renewable energy source in the distribution network; represents the active power output of the kth energy storage system; N represents the total number of nodes; Nre represents the number of renewable energy sources; where the renewable energy sources are wind and solar energy;

[0025] Power flow constraint:

[0026]

[0027]

[0028] Wherein, represents the active power output of the j-th renewable energy source at time t; represents the active power load of node i at time t; represents the voltage of node i at time t; represents the voltage of node i2 at time t; represents the conductance between node i and node i2 at time t, represents the angle between the voltages of node i and node i2 at time t; represents the conductance between node i and node i2 at time t, represents the reactive power output of the j-th renewable energy source at time t; represents the reactive power load of node i at time t;

[0029] Renewable energy output constraint:

[0030]

[0031]

[0032] Wherein, represents the upper limit of the active power output of the j-th renewable energy source; represents the apparent power of the j-th renewable energy source;

[0033] Energy storage system constraint:

[0034] E min ≤E k,t ≤E max

[0035] E k,t =E0 + η ch P ch -(1 / η dis )P dis

[0036] Wherein, E k,t represents the electricity quantity of the k-th energy storage system at time t; E min represents the lower limit of the electricity quantity of the energy storage system; E max represents the upper limit of the electricity quantity of the energy storage system; E0 represents the initial electricity quantity of the energy storage system; η ch and η dis respectively represent the charge and discharge power of the energy storage system; P ch and P dis respectively represent the charge and discharge power of the energy storage system;

[0037] The economic costs include investment and maintenance costs, and the economic benefits include the benefits of reduced network losses during operation and the benefits of power regulation generated by charging and discharging at different time periods. The total cost is determined based on the economic costs and economic benefits. The specific content is as follows:

[0038] The economic cost is calculated as follows:

[0039] C s (a) = C sb (a) / n b + C sman (a)

[0040] In the formula, C s (a) represents the average investment cost of the energy storage system in the a-th year; C sb (a) represents the construction cost considering the recovery period; n b represents the recovery period; C sman (a) represents the average maintenance cost of the energy storage system in the a-th year;

[0041] The benefit of reduced network losses is calculated as follows:

[0042]

[0043] In the formula, C Loss (a) represents the benefit formed by reducing network losses in the a-th year; p(h) represents the electricity price at the h-th hour; represents the power consumption of the distribution network loss at the h-th hour on the d-th day in the a-th year;

[0044] The benefit of power regulation is calculated as follows:

[0045]

[0046] In the formula, C net (a) is the benefit formed by power regulation in the a-th year, represents the charge and discharge amount of the energy storage system at the h-th hour on the d-th day in the a-th year;

[0047] The total cost C of the energy storage system is calculated as follows:

[0048] C = C s - C loss - C net

[0049] In the formula, C s has the same meaning as C s (a); C loss has the same meaning as C Loss (a); C net has the same meaning as C net (a).

[0050] Furthermore, the specific content of step S2 is as follows:

[0051] S21: Use the least squares support vector machine to classify the node load and renewable energy output data in the mathematical model of the distribution network;

[0052] S22: Design the least squares support vector machine to solve the binary classification problem; next, each class will be further classified into two parts, and then multiple sub-classifications will be obtained. After the classification is completed, determine whether there are sub-classifications that can be regularized together among the multiple sub-classifications through the discrimination function. If they can be regularized together, the sub-classifications will be regarded as a whole, otherwise they will still be independent;

[0053] S23: For the multiple sub-classifications processed in step S22, each sub-classification is classified into two parts again, and also determine whether there are sub-classifications that can be regularized together among the multiple sub-classifications after classification through the discrimination function; divide and regularize multiple sub-classifications in this way in a loop;

[0054] Among them, the discrimination function D yz (x) is calculated as follows:

[0055]

[0056] In the formula, w yz , b yz are the parameters of the least squares support vector machine; represents the feature vector after mapping x to the high-dimensional space;

[0057] When D yz (x) is within a certain range, it is determined that the two sub-classifications y and z can be regularized into a whole.

[0058] Furthermore, the specific content of step S3 is as follows: According to the classification results of each node data, divide the area according to the node connection situation, so that in the area divided by each node connection, the classification results of each node data are the same.

[0059] Furthermore, in step S3, if there are nodes that cannot be classified, they will be delimited to the nearby area according to the electrical distance.

[0060] Further, the specific content of step S4 is as follows: In each power distribution area divided in step S3, a neural network with residual blocks is established. The daily active power load and renewable energy output data in each area are obtained based on the mathematical model of the distribution network and used as the training data of the neural network. Since the location of the energy storage system connected to different nodes and the setting of the capacity size will affect the economic cost and economic benefits of the energy storage system, after the training is completed, the neural network takes the minimum total cost as the optimization goal. According to the active power load and renewable energy output data per hour of each day as the input of the neural network, the location of the node where the energy storage system is connected and the capacity configuration plan in each power distribution area are determined, and then multiple different energy storage system configuration plans corresponding to multiple power distribution areas are obtained.

[0061] The beneficial effects of the present invention are as follows: The present application proposes measures for judging the energy storage area according to node partitioning, effectively reducing the complexity of large-scale distribution network energy storage configuration calculation and being operable in practice; the least squares support vector machine classification only requires a small amount of data for accurate classification, which is beneficial for experiments on distribution networks with incomplete data collection; a method for processing data based on ResNet-DNN is proposed to calculate the objective function, solving the problem that a large amount of historical data of renewable energy affects the calculation speed and accuracy, alleviating the problem of gradient disappearance existing in DNN, being able to cope with the volatility and randomness of renewable energy, quickly forming a pattern judgment and then drawing a conclusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic diagram of the method for energy storage location and capacity planning of the present invention.

[0063] Figure 2 is a schematic diagram of the classification of IEEE 39 nodes in an embodiment of the present invention.

[0064] Figure 3 is a schematic diagram of the change in loss with the number of iterations during the learning process of the ResNet-DNN neural network in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] Now, the present invention will be further described in detail with reference to the accompanying drawings.

[0066] As Figure 1 described, the overall technical solution of the present application is as follows:

[0067] S1. Establish a mathematical model according to the optimization goal of the energy storage system;

[0068] S2. Classify the load and renewable energy output data by using the least squares support vector machine according to the output and consumption of the nodes;

[0069] S3. Allocate the power distribution network and determine the power distribution area of the energy storage system;

[0070] S4. Establish a deep neural network with residual blocks in each area to determine the location and capacity of energy storage;

[0071] S5. Calculate the economic benefits under different energy storage configurations.

[0072] To optimize the above technical solutions, the specific measures taken also include:

[0073] Further, the optimization objective of the energy storage system in step S1 is the overall power throughput of the energy storage system over a period of time, and the power calculation formula is as follows:

[0074]

[0075] The mathematical model in step S1 takes into account economic costs and benefits, including: investment and maintenance costs, the benefits to the power grid from reducing network losses during operation, and the electricity price benefits brought by charging and discharging at different times, etc.

[0076] Further, the classification of load and renewable energy output data using the least squares support vector machine in step S2 includes the following steps:

[0077] S21. According to the daily average data of node load and wind and photovoltaic power generation, use the least squares support vector machine for classification to divide the nodes into two categories;

[0078] S22. Continue to classify the two types of nodes according to the one - against - one multi - classification algorithm. The discrimination function for the y - th and z - th categories is:

[0079]

[0080] Further, in step S3, according to the classification results of step S2, divide the areas according to the node connection situation. For the nodes that may not be classified, delimit them to the adjacent areas according to the electrical distance.

[0081] Further, the determination of the location and capacity of energy storage in step S4 includes the following steps:

[0082] S41. Establish a deep neural network in each area according to the daily average data;

[0083] S42. Introduce residual blocks, calculate according to the daily data, and determine the location and capacity of energy storage.

[0084] The overall technical solution of this application is specifically described below:

[0085] S1. Establish a mathematical model according to the optimization objective of the energy storage system;

[0086] For the site selection and capacity determination of energy storage power stations under the new power system, it is necessary to consider both the stability of the participation of renewable energy in system operation and the economy of system operation;

[0087] S11. In order to consume more electricity while minimizing the investment cost of energy storage, the objective function is set as the overall electricity throughput of the energy storage system over a period of time, and the calculation formula is as follows:

[0088]

[0089] S12. Consider power balance constraints, power flow constraints, renewable energy output constraints, and energy storage operation balance. The node voltage constraint is as follows:

[0090] V min ≤V i ≤V max

[0091] The power balance constraint P is as follows:

[0092]

[0093] Power flow constraint:

[0094]

[0095]

[0096] The renewable energy output constraint is as follows:

[0097]

[0098]

[0099] The energy storage device constraint is as follows:

[0100] E min ≤E k,t ≤E max

[0101] E k,t =E0+η ch P ch -(1 / η dis )P dis

[0102] The economic cost of the energy storage system includes investment and maintenance costs, and the economic benefits include the benefits of reducing network losses during operation and the electricity price benefits generated by charging and discharging at different times. The total cost is determined based on the economic cost and economic benefits;

[0103] S131. The economic cost is calculated as follows:

[0104] Cs (a) = C sb (a) / n b +C sman (a)

[0105] S1312. The calculation of the network loss reduction benefit is as follows:

[0106]

[0107] S133. The power regulation benefit is as follows:

[0108]

[0109] S134. The total cost is as follows:

[0110] C = C s -C loss -C net .

[0111] S2. Classify the load and renewable energy output data by using the least squares support vector machine according to the output and consumption of the nodes;

[0112] S21. The classification of the least squares support vector machine in the prior art is as follows:

[0113]

[0114]

[0115] In the formula, both w and b are model parameters, C > 0 is the penalty parameter, and ξ is the slack factor used to measure the error;

[0116] The Gaussian kernel function in the model is as follows:

[0117]

[0118] S22. Since the least squares support vector machine is designed to solve the binary classification problem, the one - against - one multi - classification algorithm is adopted, and each class is further classified pairwise. Then the discrimination function between the y - th class and the z - th class is as follows:

[0119]

[0120] Regularize the binary results by voting. There may be data that cannot be correctly classified and need to be adjusted separately.

[0121] S3. Divide the distribution network and determine the distribution area of the energy storage system;

[0122] According to the classification results of the output data of each node, divide the area according to the node connection situation. For the nodes that may not be classified, delimit them into the adjacent area according to the electrical distance.

[0123] S4. Establish a deep neural network with residual blocks in each region to determine the location and capacity of energy storage;

[0124] During the model training process, first initialize the weights of each layer with random numbers, calculate the error between the predicted value and the true value through DNN, the error can be backpropagated to calculate the gradient, and update the weights through the gradient. Repeat the above process until the network training is completed;

[0125] Adopt the Relu activation function whose derivative is always equal to 1 in the positive part to reduce the problems of gradient disappearance and gradient explosion. The expression is:

[0126] Relu(x) = max(x, 0)

[0127] Introduce the ResNet model to learn the residuals between the output parameters and the initial values, and solve the possible problem of gradient disappearance.

[0128] S5. Calculate the economic benefits under different energy storage configurations;

[0129] Calculate the scheme with the maximum economic benefit under different energy storage specifications according to the objective function and conduct a comparative analysis.

[0130] A specific embodiment of applying the energy storage location and capacity planning method of the present invention is as follows:

[0131] Conduct experiments on the standard IEEE33 node. The node voltage fluctuation range is set to 0.9 - 1.05 p.u. Photovoltaics of 200 kW are connected to nodes 10, 17, 21, 27, and 33, and wind power of 600 kW is connected to nodes 14 and 25. The expected payback period of the energy storage device is 20 years, the construction cost is 2.5 yuan / MWh, and the maintenance cost is 0.05 yuan / MWh per year. The electricity price is divided into three segments: peak, flat, and valley. The time periods and prices are shown in Table 1

[0132] Table 1

[0133]

[0134] According to the one - to - one multi - classification algorithm of the least - squares support vector machine, the division results are as Figure 2 shown. It can be seen from the figure that the 33 nodes are divided into 6 regions.

[0135] Process the wind power, photovoltaic power generation data and active load data and use them for DNN learning. Select the mean square error as the loss function, the learning rate is 0.001, and the number of network iterations is 500. The input layer, hidden layer and output layer are set as shown in Table 2

[0136] Table 2

[0137]

[0138] During the training process, the change process of the loss value of the training set is as follows Figure 3 shown, and the model takes into account both accuracy and generalization.

[0139] Calculate the objective function. When installing 1 energy storage device, the power throughput is 0.392 MWh; when installing two, the power throughput is 0.713 MWh; when installing 3, the power throughput is 1.330 MWh; when installing 4, the power throughput is 1.523 MWh. It can be seen from the calculation results that as the energy storage device increases, the power consumption capacity increases, but the effect gradually decreases. After installing 4 energy storage devices, the power throughput increment is not obvious.

[0140] The economic cost and benefit analysis of the calculation results are shown in Table 3

[0141] Table 3

[0142]

[0143] In summary, the present invention proposes measures for judging the energy storage area according to node partitioning, effectively reducing the complexity of the calculation of the energy storage configuration of large-scale distribution networks and being operable in practice; the least squares support vector machine classification only requires a small amount of data for accurate classification, which is beneficial for distribution networks with incomplete data collection to conduct experiments; a method for processing data based on ResNet-DNN is proposed to calculate the objective function, solving the problem that a large amount of historical data of renewable energy affects the calculation speed and accuracy, alleviating the problem of gradient disappearance existing in DNN, being able to cope with the volatility and randomness of renewable energy, quickly forming a pattern judgment and then drawing a conclusion.

[0144] It should be noted that the terms such as "up", "down", "left", "right", "front", "back", etc. cited in the invention are only for the convenience of description and are not used to limit the scope of implementation of the present invention. The change or adjustment of their relative relationship, without substantial change in the technical content, should also be regarded as the scope of implementation of the present invention.

[0145] The above is only the preferred implementation manner of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A method for energy storage location and capacity planning considering distribution network zoning, characterized in that It includes the following steps: S1: Establish a mathematical model of the distribution network according to the optimization objective of the energy storage system; S2: Use the least squares support vector machine to classify the node load and renewable energy output data in the distribution network; S3: Divide the distribution network according to the classification results, and determine the distribution area of the energy storage system; S4: In each distribution area divided in step S3, a neural network with residual blocks is established. Based on the mathematical model of the distribution network, the daily average active load and renewable energy output data of each area are obtained and used as the training data of the neural network. Since the location of the energy storage system connected to different nodes and the setting of the capacity size will affect the economic cost and economic benefit of the energy storage system, after the training is completed, the neural network takes the minimum total cost as the optimization objective. According to the active load and renewable energy output data per hour of each day as the input of the neural network, determine the location and capacity configuration scheme of the energy storage system access node in each distribution area, and then obtain multiple different energy storage system configuration schemes corresponding to multiple distribution areas; S5: Calculate the economic benefits under different energy storage configurations for analysis; The optimization objective of the energy storage system described in step S1 is as follows: Set the function of the optimization objective as the overall power throughput situation f of the energy storage system within a period of time, and the calculation formula is as follows: Wherein, Ns represents the number of energy storage systems during operation; t0 represents the start time of the operation of the energy storage system; n represents the number of energy storage systems during operation; Δt represents the time period from the start time to the end time of the operation of the energy storage system; t0 + nΔt represents the time when the energy storage system ends charge and discharge; represents the charge and discharge power of the kth energy storage system at time t; The specific content of step S2 is: S21: Use the least squares support vector machine to classify the node load and renewable energy output data in the mathematical model of the distribution network; S22: Based on the least squares support vector machine, it is designed to solve the binary classification problem. Next, each category is further classified into two parts, and then multiple sub-classifications are obtained. After the classification is completed, determine whether there are sub-classifications that can be regularized together among the multiple sub-classifications through the discrimination function. If they can be regularized together, the sub-classifications are regarded as a whole, otherwise they remain independent; S23: For the multiple sub-classifications processed in step S22, each sub-classification is classified into two parts again, and also determine whether there are sub-classifications that can be regularized together among the multiple sub-classifications after classification through the discrimination function; and so on, cycle to divide and regularize multiple sub-classifications; Among them, the discrimination function D yz (x) is calculated as follows: where w yz , b yz are the parameters of the least squares support vector machine; represents the feature vector after mapping x to a high-dimensional space; when D yz (x) is within a certain range, it is determined that the two sub-classifications y and z can be regularized into a whole.

2. The energy storage location and capacity planning method considering distribution network zoning according to claim 1, wherein, The established mathematical model in step S1 also considers constraint conditions, economic costs and economic benefits; The constraint conditions include node voltage constraints, power balance constraints, power flow constraints, wind and light output constraints, and energy storage operation balance; Node voltage constraint: V min ≤V i ≤V max Where V min represents the lower limit of the node voltage; V max represents the upper limit of the node voltage; V i represents the voltage of node i; Power balance constraint P: Wherein, P i L represents the active power load of node i in the distribution network; represents the active power output of the j-th renewable energy source in the distribution network; represents the active power output of the k-th energy storage system; N represents the total number of nodes; Nre represents the number of renewable energy sources; wherein the renewable energy sources are wind and solar energy; Power flow constraint: In the formula, represents the active power output of the j-th renewable energy source at time t; represents the active power load of node i at time t; V i,t represents the voltage of node i at time t; represents the voltage of node i2 at time t; represents the conductance between node i and node i2 at time t, represents the angle between the voltages of node i and node i2 at time t, represents the conductance between node i and node i2 at time t, represents the reactive power output of the j-th renewable energy source at time t; represents the reactive power load of node i at time t; Wind and light output constraint: In the formula, represents the upper limit of the active power output of the j-th renewable energy source; represents the apparent power of the j-th renewable energy source; Energy storage system constraint: E min ≤E k,t ≤E max E k,t = E0 + η ch P ch -(1 / η dis )P dis Where, E k,t represents the power of the k-th energy storage system at time t; E min represents the lower limit of the power of the energy storage system; E max represents the upper limit of the power of the energy storage system; E0 represents the initial power of the energy storage system; η ch and η dis respectively represent the charge and discharge power of the energy storage system; P ch and P dis respectively represent the charge and discharge power of the energy storage system; The economic cost includes investment and maintenance costs, and the economic benefit includes the benefit of reducing network loss during operation and the benefit of power regulation generated by charging and discharging in different time periods. The total cost is determined based on the economic cost and economic benefit; the specific content is: The economic cost is calculated as follows: C s (a) = C sb (a) / n b + C sman (a) Where C s (a) represents the average investment cost of the energy storage system in the a-th year; C sb (a) represents the construction cost considering the payback period; n b represents the payback period; C sman (a) represents the average maintenance cost of the energy storage system in the a-th year; The benefit of reducing network loss is calculated as follows: where C Loss (a) represents the revenue generated by reducing network losses in the a-th year; p(h) represents the electricity price at the h-th hour; represents the power consumption of the distribution network loss at the h-th hour on the d-th day in the a-th year; The benefit of power regulation is calculated as follows: Where C net (a) is the revenue formed by the electric energy regulation in the ath year, represents the charge and discharge amount of the energy storage system at d day and h hour in the ath year; The total cost C of the energy storage system is calculated as follows: C=C s -C loss -C net where C s has the same meaning as C s (a); C loss has the same meaning as C Loss (a); C net has the same meaning as C net (a).

3. A method for planning the energy storage location and capacity considering the distribution network partition according to claim 1, characterized in that The specific content of step S3 is: According to the classification results of each node data, divide the area according to the node connection situation, so that in the area divided by the connection of each node, the classification results of each node data are the same.

4. A method for planning the energy storage location and capacity considering the distribution network partition according to claim 3, characterized in that In step S3, if there are nodes that cannot be classified, they are assigned to a nearby area according to the electrical distance.

Images