Park energy storage capacity optimal configuration method and system and storage medium

Abstract

In order to solve the defects in the prior art, the invention provides a park energy storage capacity optimal configuration method and system and a storage medium, and the method comprises the steps: 1, collecting and obtaining the wind and light historical data and load historical data of a to-be-analyzed park; 2, preprocessing the historical data, analyzing factors influencing the wind, light and load output of the park, screening out a feature input set, and normalizing the processed historical data; 3, constructing a wind-light and load output prediction model, taking the feature input set and historical data as the input of the prediction model, and performing short-term power prediction on the wind-light and load; 4, constructing an energy storage capacity optimization configuration model of the wind and light storage system based on a short-term power prediction result; and 5, solving the optimal configuration model of the energy storage capacity of the wind and light storage system to obtain an optimal configuration result of the energy storage capacity of the wind and light storage system. The method effectively improves the wind and light absorption of the park, reduces the wind and light abandoning rate, and guarantees the economic benefits of the park.

Description

Technical Field

[0001] The present invention relates to the field of new energy consumption and optimization technology, and in particular to a method and system for optimizing the configuration of energy storage capacity in a park, and a storage medium. Background Art

[0002] With the rapid development of the economy, my country's industrial and agricultural electricity demand has increased year by year. For industrial parks with a large number of electrical equipment and a large power consumption, the power quality issue is of paramount importance. Due to their large power demand, power shortages are more likely to occur. Therefore, reasonable planning of the power between source and load is an important part of the stable operation of the industrial park system. At the same time, achieving a balance between power generation and power consumption in the industrial park is also inseparable from source-load power prediction.

[0003] For source-load power prediction, we can use historical data combined with statistical theory to predict future data, or we can predict based on neural networks. In view of the low prediction accuracy of a single neural network, this paper establishes a hybrid neural network prediction model. Due to the time series mismatch between wind and solar power and load, problems such as wind and solar power cannot be fully absorbed or power supply is insufficient in some periods, which seriously affects the production and life of the park. In order to solve this problem of insufficient power supply or excess power generation, on the basis of source-load prediction, an energy storage system is added. Through the charging and discharging of the energy storage system, the volatility of wind and solar power output can be smoothed, so that the source-load power of the park can reach a dynamic balance, thereby effectively alleviating the problem of insufficient power supply or excess power generation in the park and ensuring the normal operation of the park. Summary of the invention

[0004] In order to solve the deficiencies of the prior art, one of the purposes of the present invention is to provide a method for optimizing the configuration of energy storage capacity in a park, comprising the following steps:

[0005] Step 1: Collect and obtain the historical wind and solar data and load data of the park to be analyzed.

[0006] Step 2 preprocesses the historical data in step 1, analyzes the factors affecting the wind and solar power and load output of the park, selects the most relevant influencing factors as the feature input set of the prediction model, and normalizes the processed historical data.

[0007] Step 3 builds a prediction model for wind, solar and load output, uses the feature input set and historical data obtained in step 2 as input to the prediction model, and performs short-term power prediction for wind, solar and load.

[0008] Step 4 builds a wind-solar-storage system energy storage capacity optimization configuration model based on the short-term power forecast results of step 3. The wind-solar-storage system energy storage capacity optimization configuration model takes the minimum total cost as the objective function, and takes energy balance constraints, energy storage power constraints and charge state constraints as the constraint conditions of the objective function.

[0009] Step 5 solves the optimal configuration model of the energy storage capacity of the wind-solar-storage system to obtain the optimal configuration result of the energy storage capacity of the wind-solar-storage system.

[0010] Furthermore, the method for preprocessing the historical data in step 1 in step 2 includes:

[0011] Step 2.1.1 Supplement the missing data in the historical data collected in step 1 by formula (1):

[0012]

[0013] Where: X t is the data value at time t where the missing value is located. t+1 is the missing data value at time t+1 where the missing value is located. t-1 is the data value at time t-1 where the missing value is located. t-n+1 is the data value at time t-n+1 where the missing value is located.

[0014] Step 2.1.2 After data filling, determine whether there are outliers and remove them if they exist.

[0015] Step 2.1.3 uses formula (1) to supplement the missing values ​​eliminated in step 2.1.2.

[0016] Furthermore, the method of analyzing the factors affecting the wind, solar and load output of the park and selecting the most relevant factors as described in step 2 includes:

[0017] Step 2.2.1 Construct the feature correlation matrix R.

[0018] Step 2.2.2 Set the correlation coefficient threshold P:

[0019]

[0020] Step 2.2.3 Calculate R using formula (3):

[0021]

[0022] Where: R is the correlation coefficient. i and Y i is the i-th pair of observations in the sample data. is the mean of X. is the mean of Y.

[0023] Step 2.2.4 calculates R and determines whether there is a characteristic parameter R>P in the matrix R.

[0024] Step 2.2.5: In each pair of highly correlated features, remove the feature with less correlation with the target value.

[0025] Step 2.2.6 Repeat steps 2.2.4 and 2.2.5 until there are no highly identical feature pairs.

[0026] The result output in step 2.2.7 at this time is the most relevant influencing factor for the corresponding sample.

[0027] Furthermore, the normalization method in step 2 includes: performing normalization using formula (4):

[0028]

[0029] Where: maxf i 、minf i maxf are the maximum and minimum values ​​of the sample data respectively. i -minf i is the extreme value difference. i is the original value of the sample data. f′ i is the normalized result value.

[0030] Furthermore, the prediction model in step 3 includes a prediction model of a convolutional neural network and a prediction model of a long short-term memory neural network.

[0031] Furthermore, the prediction model of the convolutional neural network includes:

[0032] The convolutional network model represented by formula (5) is:

[0033] S=X*ω (5)

[0034] Where: S is the feature mapping matrix after convolution, X is the input matrix, ω is the weight matrix (ie, kernel function), and * is the convolution operator.

[0035] The input of the convolution operation is multi-dimensional, and the number of convolution kernels plays a decisive role in the degree of feature extraction. Therefore, it is adjusted according to formula (6).

[0036]

[0037] Where: S i,j is the element of matrix S, P n,m K is the element of the 2D input matrix P of size N*M, i+n,j+m is the element of the 2D convolution kernel K.

[0038] Furthermore, the prediction model of the long short-term memory neural network includes: a forget gate, an input gate, a memory unit, and an output gate, wherein:

[0039] The forget gate is used to determine the information that needs to be discarded, mainly composed of W fand b f As shown in formula (7):

[0040] f t =σ·(W f ·[h t-1 ,x t ]+b f ) t ∈[0,1] (7)

[0041] Where: W f is the weight matrix of the forget gate, b f is the bias term of the forget gate. t is the input value at the current time t; h t-1 is the output value of the hidden layer at time t-1; is the state information at the current moment, and σ is the sigmoid function;

[0042] The input gate determines the input information, which is passed through the output Control, as shown in formula (8):

[0043]

[0044] Where: W i is the weight matrix of the input gate, b i is the bias term of the input gate.

[0045] Memory unit: Update candidate value and the current state information c t , as shown in formula (9) and formula (10):

[0046]

[0047] Where: W c is the weight matrix of the memory unit, b c is the bias term of the memory unit. It is the current state information. is the output of the input gate at the current time t. tanh is a function.

[0048] Output gate: status information c t The output o of the LSTM network is obtained after nonlinear processing t , as shown in formula (11):

[0049] o t =σ·(W o ·[h t-1 ,x t ]+b o ) (11)

[0050] Where: W ois the weight matrix of the output gate, b o is the bias term of the output gate.

[0051] Use formula (12) to output the final prediction result:

[0052] h t =o t ·tanh·c t (12)

[0053] Where: h t The final prediction result.

[0054] Furthermore, in step 3, the method of using the feature input set obtained in step 2 and the historical data as inputs of the prediction model to perform short-term power prediction on wind, solar and load includes:

[0055] Step 3.1 extracts the feature input set obtained in step 2 and historical data through the prediction model of the convolutional neural network.

[0056] Step 3.2 converts the extracted feature data into the input format of the prediction model of the long short-term memory neural network.

[0057] The converted feature data in step 3.3 enters the prediction model of the long short-term memory neural network for learning and training.

[0058] Step 3.4 outputs the trained data through the fully connected layer and the output layer to obtain the short-term power prediction data corresponding to wind, solar and load.

[0059] Furthermore, the method for constructing the wind-solar-storage system energy storage capacity optimization configuration model described in step 4 includes:

[0060] Step 4.1 Establish a single objective function for the optimization model:

[0061]

[0062] Where: Invest in maintenance costs for energy storage systems. Invest in maintenance costs for auxiliary equipment.

[0063] Step 4.2 Calculate the investment and maintenance costs of the energy storage system.

[0064]

[0065] Where: Total cost of investing in and maintaining an energy storage system. R Plan capacity for energy storage systems. Plan the unit capacity investment cost for energy storage systems. Plan the unit capacity maintenance cost for the energy storage system. r is the discount rate.ESS It is the economic service life of the energy storage system.

[0066] Step 4.3 calculates the investment and maintenance costs of auxiliary equipment, mainly the bidirectional inverter of the energy storage system.

[0067]

[0068] Where: It is the maximum power output of the bidirectional inverter in the auxiliary equipment. It is the investment cost per unit capacity of the bidirectional inverter in the auxiliary equipment. is the maintenance cost per unit capacity of the bidirectional inverter in the auxiliary equipment. r is the discount rate. f_ESS It is the economic service life of the energy storage system.

[0069] Step 4.4 establishes constraints, including energy storage system constraints, energy storage power constraints, and state of charge constraints, where:

[0070] Energy balance constraints:

[0071]

[0072] Where: P WT (t) is the output power of the fan during period t, kW. PV (t) is the photovoltaic output power during period t, kW. is the ESS charging power during period t, kW. is the ESS discharge power in period t, kW. load (t) is the load demand during period t, kW.

[0073] Energy storage power constraints:

[0074]

[0075] Where: is the minimum value of ESS charging power, kW. is the maximum value of ESS charging power, kW. is the minimum value of ESS discharge power, kW. is the maximum value of ESS discharge power, kW.

[0076] State of Charge Constraints:

[0077] SOC min ≤SOC(t)≤SOC max (18)

[0078] Where: SOC min and SOC max They represent the lower and upper limits of the energy storage system's state of charge respectively.

[0079] Furthermore, step 5 uses the Ivy optimization algorithm to solve the optimal configuration model of the energy storage capacity of the wind-solar-storage system to obtain the optimal configuration result of the energy storage capacity of the wind-solar-storage system, which specifically includes:

[0080] Step 5.1 determines that the ivy optimization algorithm includes: population initialization, coordinated and orderly population growth, obtaining sunlight source growth, propagation and evolution of ivy plants, and selection of survivors.

[0081] Step 5.2 Population initialization:

[0082] I i =I min +rand(1,D)⊙(I max -I min )i=1,…,Npop (19)

[0083] Where: D is the dimension. max and I min are the upper and lower bounds of the search space respectively; ⊙ is the product of two vectors.

[0084] Step 5.3 Coordinated and orderly population growth:

[0085] ΔGv i (t+1)=rand 2 ⊙[N(1,D)⊙ΔG vi (t)] (20)

[0086] Where: ΔGv i (t) is the growth rate of the discrete time system at time t. ΔGv i (t+1) is the growth rate of the discrete time system at time t+1. rand() is a random real number in the interval [0,1]. 2 is a random number. N(1,D) represents a random vector with dimension D, which is a random number in the standard Gaussian distribution.

[0087] Step 5.4 Get sunlight source for growth:

[0088]

[0089] Where: |N(1,D)| is a vector whose components are the absolute values ​​of the components of vector N(1,D). is the position of the latest member i. i is the position of the i-th member. ii The position of the iith member. % is the division of two vectors.

[0090] Step 5.5 Spread and evolution of ivy plants:

[0091]

[0092] Where: I best The best member of the population. The position of the current latest member i. is the growth rate of the current newest member.

[0093] Step 5.6. Survivor selection:

[0094]

[0095] Where: β = (2 + rand) / 2.

[0096] Furthermore, according to the wind and solar data and load data predicted in step 4, combined with the solution method of the Ivy optimization algorithm and the established optimization model, the optimal solution set is obtained through multiple iterations, and then the optimal configuration plan of the energy storage capacity under the wind, solar and storage system is obtained.

[0097] In addition, the present invention also provides a system for optimizing the configuration of energy storage capacity in a park, including: a data acquisition module, a data preprocessing module, a short-term power prediction module, and an energy storage capacity optimization configuration analysis module. The data acquisition module is used to execute the method steps of step 1 in the above-mentioned method for optimizing the configuration of energy storage capacity in a park, the data preprocessing module is used to execute the method steps of step 2 in the above-mentioned method for optimizing the configuration of energy storage capacity in a park, the short-term power prediction module is used to execute the method steps of step 3 in the above-mentioned method for optimizing the configuration of energy storage capacity in a park, and the energy storage capacity optimization configuration analysis module is used to execute the method steps of step 4 and step 5 in the above-mentioned method for optimizing the configuration of energy storage capacity in a park.

[0098] A third object of the present invention is to provide a storage medium storing instructions that are readable by a computer and can be executed by calculated values, wherein the computer-executable instructions are used to execute the above-mentioned method for optimizing the configuration of energy storage capacity in a park.

[0099] The beneficial effects of the present invention are as follows: based on the short-term power forecast results, the present invention constructs an optimal configuration model for the energy storage capacity of the wind, solar and storage system, takes the minimum total cost as the objective function, and determines the relevant constraints, including energy balance constraints, energy storage power constraints and charge state constraints. The optimization model is solved using the ivy optimization algorithm to obtain the optimal configuration result of the energy storage capacity of the wind, solar and storage system. The proposed method can effectively improve the wind and solar power consumption of the park, reduce the wind and solar power abandonment rate, and ensure the economic benefits of the park. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1 Shown is a flow chart of the steps of the method for optimizing the configuration of the energy storage capacity of a park according to the present invention.

[0101] Figure 2 Shown is a calculation flow chart of the Pearson correlation coefficient method of the present invention.

[0102] Figure 3 Shown is the structure diagram of the wind-solar and load forecasting model under the CNN-LSTM network of the present invention. DETAILED DESCRIPTION

[0103] The present invention will now be described in further detail with reference to the accompanying drawings.

[0104] Example 1

[0105] A method for optimizing the configuration of energy storage capacity in a park, characterized by comprising the following steps:

[0106] Step 1: Collect and obtain the historical wind and solar data and load data of the park to be analyzed.

[0107] Step 2 preprocesses the historical data in step 1, analyzes the factors affecting the wind and solar power and load output of the park, selects the most relevant influencing factors as the feature input set of the prediction model, and normalizes the processed historical data.

[0108] The method for preprocessing the historical data in step 1 described in step 2 includes:

[0109] Step 2.1.1 Supplement the missing data in the historical data collected in step 1 by formula (1):

[0110]

[0111] Where: X t is the data value at time t where the missing value is located. t+1 is the missing data value at time t+1 where the missing value is located. t-1 is the data value at time t-1 where the missing value is located. t-n+1 is the data value at time t-n+1 where the missing value is located.

[0112] Step 2.1.2 After data filling, determine whether there are outliers and remove them if they exist.

[0113] Step 2.1.3 uses formula (1) to supplement the missing values ​​eliminated in step 2.1.2.

[0114] The method of analyzing the factors affecting the wind, solar and load output of the park and selecting the most relevant factors as described in step 2 includes:

[0115] Step 2.2.1 Construct the feature correlation matrix R.

[0116] Step 2.2.2 Set the correlation coefficient threshold P:

[0117]

[0118] Step 2.2.3 Figure 2 For example, the Pearson correlation coefficient method is used to analyze the wind and solar data and load data and the corresponding influencing factors, obtain their correlation coefficients and perform heat map visualization, and determine the influencing factors with strong correlation with themselves as the feature input for prediction. Specifically, R is calculated using formula (3):

[0119]

[0120] Where: R is the Pearson correlation coefficient. i and Y i is the i-th pair of observations in the sample data. is the mean of X. is the mean of Y.

[0121] Step 2.2.4 calculates R and determines whether there is a characteristic parameter R>P in the matrix R.

[0122] Step 2.2.5: In each pair of highly correlated features, remove the feature with less correlation with the target value.

[0123] Step 2.2.6 Repeat steps 2.2.4 and 2.2.5 until there are no highly identical feature pairs.

[0124] The result output in step 2.2.7 at this time is the most relevant influencing factor for the corresponding sample.

[0125] The normalization method in step 2 includes: performing normalization using formula (4):

[0126]

[0127] Where: maxf i 、minf i maxf are the maximum and minimum values ​​of the sample data respectively. i -minf i is the extreme value difference. i is the original value of the sample data. f′ i is the normalized result value.

[0128] Step 3 builds a prediction model for wind, solar and load output, uses the feature input set and historical data obtained in step 2 as input to the prediction model, and performs short-term power prediction for wind, solar and load.

[0129] The prediction model in step 3 includes a prediction model of a convolutional neural network (CNN) and a prediction model of a long short-term memory neural network (LSTM).

[0130] by Figure 3 Taking CNN-LSTM network as an example, the prediction model of the present invention is as follows:

[0131] The prediction model of the convolutional neural network includes:

[0132] The convolutional network model represented by formula (5) is:

[0133] S=X*ω (5)

[0134] Where: S is the feature mapping matrix after convolution, X is the input matrix, ω is the weight matrix (ie, kernel function), and * is the convolution operator.

[0135] The input of the convolution operation is multi-dimensional, and the number of convolution kernels plays a decisive role in the degree of feature extraction. Therefore, it is adjusted according to formula (6).

[0136]

[0137] Where: S i,j is the element of matrix S, P n,m K is the element of the 2D input matrix P of size N*M, i+n,j+m is the element of the 2D convolution kernel K.

[0138] The prediction model of the long short-term memory neural network includes: a forget gate, an input gate, a memory unit, and an output gate, wherein:

[0139] The forget gate is used to determine the information that needs to be discarded, mainly composed of W f and b f As shown in formula (7):

[0140] f t =σ·(W f ·[h t-1 ,x t ]+b f ) t ∈[0,1] (7)

[0141] Where: W f is the weight matrix of the forget gate, b f is the bias term of the forget gate. t is the input value at the current time t; h t-1 is the output value of the hidden layer at time t-1; is the state information at the current moment, and σ is the sigmoid function;

[0142] The input gate determines the input information, which is passed through the output Control, as shown in formula (8):

[0143]

[0144] Where: W i is the weight matrix of the input gate, b i is the bias term of the input gate.

[0145] Memory unit: Update candidate value and the current state information c t , as shown in formula (9) and formula (10):

[0146]

[0147] Where: W c is the weight matrix of the memory unit, b c is the bias term of the memory unit. It is the current state information. is the output of the input gate at the current time t. tanh is a function.

[0148] Output gate: status information c t The output o of the LSTM network is obtained after nonlinear processing t , as shown in formula (11):

[0149] o t =σ·(W o ·[h t-1 ,x t ]+b o ) (11)

[0150] Where: W o is the weight matrix of the output gate, b o is the bias term of the output gate.

[0151] Use formula (12) to output the final prediction result:

[0152] h t =o t ·tanh·c t (12)

[0153] Where: h t The final prediction result.

[0154] In step 3, the method for using the feature input set obtained in step 2 and the historical data as the input of the prediction model to perform short-term power prediction on wind, solar and load includes:

[0155] Step 3.1 extracts the feature input set obtained in step 2 and historical data through the prediction model of the convolutional neural network.

[0156] Step 3.2 converts the extracted feature data into the input format of the prediction model of the long short-term memory neural network.

[0157] The converted feature data in step 3.3 enters the prediction model of the long short-term memory neural network for learning and training.

[0158] Step 3.4 outputs the trained data through the fully connected layer and the output layer to obtain the short-term power prediction data corresponding to wind, solar and load.

[0159] Step 4 builds a wind-solar-storage system energy storage capacity optimization configuration model based on the short-term power forecast results of step 3. The wind-solar-storage system energy storage capacity optimization configuration model takes the minimum total cost as the objective function, and takes energy balance constraints, energy storage power constraints and charge state constraints as the constraint conditions of the objective function.

[0160] The method for constructing the optimal configuration model of the wind-solar-storage system energy storage capacity in step 4 includes:

[0161] Step 4.1 Establish a single objective function for the optimization model:

[0162]

[0163] Where: Invest in maintenance costs for energy storage systems. Invest in maintenance costs for auxiliary equipment.

[0164] Step 4.2 Calculate the investment and maintenance costs of the energy storage system.

[0165]

[0166] Where: Total cost of investing in and maintaining an energy storage system. R Plan capacity for energy storage systems. Plan the unit capacity investment cost for energy storage systems. Plan the unit capacity maintenance cost for the energy storage system. r is the discount rate. ESS It is the economic service life of the energy storage system.

[0167] Step 4.3 calculates the investment and maintenance costs of auxiliary equipment, mainly the bidirectional inverter of the energy storage system.

[0168]

[0169] Where: It is the maximum power output of the bidirectional inverter in the auxiliary equipment. It is the investment cost per unit capacity of the bidirectional inverter in the auxiliary equipment. is the maintenance cost per unit capacity of the bidirectional inverter in the auxiliary equipment. r is the discount rate. f_ESS It is the economic service life of the energy storage system.

[0170] Step 4.4 establishes constraints, including energy storage system constraints, energy storage power constraints, and state of charge constraints, where:

[0171] Energy balance constraints:

[0172]

[0173] Where: P WT (t) is the output power of the fan during period t, kW. PV (t) is the photovoltaic output power during period t, kW. is the ESS charging power during period t, kW. is the ESS discharge power in period t, kW. load (t) is the load demand during period t, kW.

[0174] Energy storage power constraints:

[0175]

[0176] Where: is the minimum value of ESS charging power, kW. is the maximum value of ESS charging power, kW. is the minimum value of ESS discharge power, kW. is the maximum value of ESS discharge power, kW.

[0177] State of Charge Constraints:

[0178] SOC min ≤SOC(t)≤SOC max (18)

[0179] Where: SOC min and SOC max They represent the lower and upper limits of the energy storage system's state of charge respectively.

[0180] Step 5 solves the optimal configuration model of the energy storage capacity of the wind-solar-storage system to obtain the optimal configuration result of the energy storage capacity of the wind-solar-storage system, which specifically includes:

[0181] Step 5.1 determines that the ivy optimization algorithm includes: population initialization, coordinated and orderly population growth, obtaining sunlight source growth, propagation and evolution of ivy plants, and selection of survivors.

[0182] Step 5.2 Population initialization:

[0183] Ii =I min +rand(1,D)⊙(I max -I min ) i=1,…,Npop (19)

[0184] Where: D is the dimension. max and I min are the upper and lower bounds of the search space respectively; ⊙ is the product of two vectors.

[0185] Step 5.3 Coordinated and orderly population growth:

[0186] ΔGv i (t+1)=rand 2 ⊙[N(1,D)⊙ΔG vi (t)] (20)

[0187] Where: ΔGv i (t) is the growth rate of the discrete time system at time t. ΔGv i (t+1) is the growth rate of the discrete time system at time t+1. rand() is a random real number in the interval [0,1]. 2 is a random number. N(1,D) represents a random vector with dimension D, which is a random number in the standard Gaussian distribution.

[0188] Step 5.4 Get sunlight source for growth:

[0189]

[0190]

[0191] Where: |N(1,D)| is a vector whose components are the absolute values ​​of the components of vector N(1,D). is the position of the latest member i. i is the position of the i-th member. ii The position of the iith member. % is the division of two vectors.

[0192] Step 5.5 Spread and evolution of ivy plants:

[0193]

[0194] Where: I best The best member of the population. The position of the current latest member i. is the growth rate of the current newest member.

[0195] Step 5.6. Survivor selection:

[0196]

[0197] Where: β = (2 + rand) / 2.

[0198] According to the wind and solar data and load data predicted in step 4, combined with the solution method of the Ivy optimization algorithm and the established optimization model, the optimal solution set is obtained through multiple iterations, and then the optimal configuration plan of energy storage capacity under the wind, solar and storage system is obtained.

[0199] The present invention collects and obtains the historical data of wind and solar power and historical data of load of the park to be analyzed; pre-processes the historical data, uses the Pearson correlation coefficient method to analyze the factors affecting the wind and solar power and load output of the park, selects the most relevant influencing factors as the predicted feature input set, and normalizes the processed historical data; constructs a prediction model of wind and solar power and load output of the CNN-LSTM network, combines the model, the selected feature input and historical data to perform short-term power prediction of wind and solar power and load; based on the short-term power prediction results, constructs an optimal configuration model for the energy storage capacity of the wind, solar and storage system, takes the minimum total cost as the objective function, and determines the relevant constraints, including energy balance constraints, energy storage power constraints and charge state constraints. The optimization model is solved using the Ivy optimization algorithm to obtain the optimal configuration result of the energy storage capacity of the wind, solar and storage system.

[0200] The present invention uses the above-mentioned park energy storage capacity optimization configuration method, predicts the wind, light and load in the park through the CNN-LSTM neural network model, and constructs a wind, light and storage system energy storage capacity optimization configuration model based on the short-term power prediction results. The minimum total cost is used as the objective function to determine the relevant constraints, including energy balance constraints, energy storage power constraints and charge state constraints. The optimization model is solved using the ivy optimization algorithm to obtain the optimal configuration result of the wind, light and storage system energy storage capacity. The proposed method can effectively improve the wind and light consumption in the park, reduce the wind and light abandonment rate, and ensure the economic benefits of the park.

[0201] Example 2

[0202] A system for optimizing and configuring a park energy storage capacity, comprising: a data acquisition module, a data preprocessing module, a short-term power prediction module, and an energy storage capacity optimization configuration analysis module; wherein the data acquisition module is used to execute the method steps of step 1 in the park energy storage capacity optimization configuration method described in Example 1, the data preprocessing module is used to execute the method steps of step 2 in the park energy storage capacity optimization configuration method described in Example 1, the short-term power prediction module is used to execute the method steps of step 3 in the park energy storage capacity optimization configuration method described in Example 1, and the energy storage capacity optimization configuration analysis module is used to execute the method steps of step 4 and step 5 in the park energy storage capacity optimization configuration method described in Example 1.

[0203] Example 3

[0204] A storage medium stores instructions that are readable by a computer and can be executed by calculated values, and the computer-executable instructions are used to execute the park energy storage capacity optimization configuration method described in Example 1.

[0205] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for optimizing the configuration of energy storage capacity in a park, characterized in that: The steps include: Step 1: Collect and obtain the historical wind and solar data and load data of the park to be analyzed; Step 2 preprocesses the historical data in step 1, analyzes the factors affecting the wind and solar power and load output of the park, selects the most relevant influencing factors as the feature input set of the prediction model, and normalizes the processed historical data; Step 3: construct a prediction model for wind, solar and load output, use the feature input set and historical data obtained in step 2 as inputs to the prediction model, and perform short-term power prediction for wind, solar and load; Step 4: Based on the short-term power forecast result of step 3, a wind-solar-storage system energy storage capacity optimization configuration model is constructed; the wind-solar-storage system energy storage capacity optimization configuration model takes the minimum total cost as the objective function, and takes the energy balance constraint, energy storage power constraint and charge state constraint as the constraint conditions of the objective function; Step 5 solves the optimal configuration model of the energy storage capacity of the wind-solar-storage system to obtain the optimal configuration result of the energy storage capacity of the wind-solar-storage system.

2. The method for optimizing the configuration of energy storage capacity in a park according to claim 1, characterized in that: The method for preprocessing the historical data in step 1 described in step 2 includes: Step 2.1.1 Supplement the missing data in the historical data collected in step 1 by formula (1): Where: X t is the data value at time t where the missing value is located; X t+1 is the missing data value at time t+1 where the missing value is located; X t-1 is the data value at time t-1 where the missing value is located; X t-n+1 is the data value at time t-n+1 where the missing value is located; Step 2.1.2 After filling the data, determine whether there are outliers, and if so, remove them; Step 2.1.3 uses formula (1) to supplement the missing values ​​eliminated in step 2.1.

2.

3. The method for optimizing the configuration of energy storage capacity in a park according to claim 1, characterized in that: The method of analyzing the factors affecting the wind, solar and load output of the park and selecting the most relevant factors as described in step 2 includes: Step 2.2.1 construct the feature correlation matrix R; Step 2.2.2 Set the correlation coefficient threshold P: Step 2.2.3 Calculate R using formula (3): Where: R is the correlation coefficient; X i and Y i is the i-th pair of observations in the sample data; is the mean of X; is the mean of Y; Step 2.2.4 calculates R and determines whether there is a characteristic parameter of R>P in the matrix R; Step 2.2.5: In each pair of highly correlated features, delete the feature with less correlation with the target value; Step 2.2.6 Repeat steps 2.2.4 and 2.2.5 until there are no highly identical feature pairs; The result output in step 2.2.7 at this time is the most relevant influencing factor for the corresponding sample.

4. The method for optimizing the configuration of energy storage capacity in a park according to claim 1, characterized in that: The normalization method in step 2 includes: performing normalization using formula (4): Where: maxf i 、minf i are the maximum and minimum values ​​of the sample data respectively; maxf i -minf i is the extreme value difference; f i is the original value of the sample data; f· i is the normalized result value.

5. The method for optimizing the configuration of energy storage capacity in a park according to claim 1, characterized in that: The prediction model described in step 3 includes a prediction model of a convolutional neural network and a prediction model of a long short-term memory neural network.

6. The method for optimizing the configuration of energy storage capacity in a park according to claim 5, characterized in that: The prediction model of the convolutional neural network includes: The convolutional network model represented by formula (5) is: S=X*ω (5) Where: S is the feature mapping matrix after convolution, X is the input matrix, ω is the weight matrix (i.e., kernel function), and * is the convolution operator; The input of the convolution operation is multi-dimensional input, and the number of convolution kernels plays a decisive role in the degree of feature extraction. Therefore, it is adjusted according to formula (6); Where: S i,j is the element of matrix S, P n,m K is the element of the 2D input matrix P of size N*M, i+n,j+m is the element of the 2D convolution kernel K.

7. The method for optimizing the configuration of energy storage capacity in a park according to claim 5, characterized in that: The prediction model of the long short-term memory neural network includes: a forget gate, an input gate, a memory unit, and an output gate, wherein: The forget gate is used to determine the information that needs to be discarded, mainly composed of W f and b f As shown in formula (7): f t =σ·(W f ·[h t-1 ,x t ]+b f ) f t ∈[0,1] (7) Where: W f is the weight matrix of the forget gate, b f is the bias term of the forget gate; x t is the input value at the current time t; h t-1 is the output value of the hidden layer at time t-1; is the state information at the current moment, and σ is the sigmoid function; The input gate determines the input information, which is passed through the output Control, as shown in formula (8): Where: W i is the weight matrix of the input gate, b i is the bias term of the input gate; Memory unit: Update candidate value and the current state information c t , as shown in formula (9) and formula (10): Where: W c is the weight matrix of the memory unit, b c is the bias term of the memory unit; It is the current state information; is the output of the input gate at the current time t; tanh is a function; Output gate: status information c t The output o of the LSTM network is obtained after nonlinear processing t , as shown in formula (11): the t =σ·(W o ·[h t-1 ,x t ]+b o ) (11) Where: W o is the weight matrix of the output gate, b o is the bias term of the output gate; Use formula (12) to output the final prediction result: h t =o t ·tanh·c t (12) Where: h t The final prediction result.

8. The method for optimizing the configuration of energy storage capacity in a park according to claim 5, characterized in that: In step 3, the method for using the feature input set obtained in step 2 and the historical data as the input of the prediction model to perform short-term power prediction on wind, solar and load includes: Step 3.1 extracts the feature input set obtained in step 2 and the historical data through the prediction model of the convolutional neural network; Step 3.2 converts the extracted feature data into the input format of the prediction model of the long short-term memory neural network; Step 3.3 The converted feature data enters the prediction model of the long short-term memory neural network for learning and training; Step 3.4 outputs the trained data through the fully connected layer and the output layer to obtain the short-term power prediction data corresponding to wind, solar and load.

9. The method for optimizing the configuration of energy storage capacity in a park according to claim 1, characterized in that: The method for constructing the optimal configuration model of the wind-solar-storage system energy storage capacity in step 4 includes: Step 4.1 Establish a single objective function for the optimization model: Where: Investing in maintenance costs for energy storage systems; Invest in maintenance costs for auxiliary equipment; Step 4.2 Calculate the investment and maintenance cost of the energy storage system; Where: Total investment and maintenance cost for energy storage system; ESS R Planning capacity for energy storage systems; Plan the unit capacity investment cost for energy storage systems; is the unit capacity maintenance cost of the energy storage system; r is the discount rate; n ESS The economic service life of the energy storage system; Step 4.3 Calculate the investment and maintenance costs of auxiliary equipment, mainly the bidirectional inverter of the energy storage system; Where: The maximum power output of the bidirectional inverter in the auxiliary equipment; The unit capacity investment cost of the bidirectional inverter in the auxiliary equipment; is the maintenance cost per unit capacity of the bidirectional inverter in the auxiliary equipment; r is the discount rate; n f_ESS The economic service life of the energy storage system; Step 4.4 establishes constraints, including energy storage system constraints, energy storage power constraints, and state of charge constraints, where: Energy balance constraints: Where: P WT (t) is the output power of the fan during period t, kW; P PV (t) is the photovoltaic output power during period t, kW; is the ESS charging power during period t, kW; is the ESS discharge power in period t, kW; L load (t) is the load demand during period t, kW; Energy storage power constraints: Where: is the minimum value of ESS charging power, kW; is the maximum value of ESS charging power, kW; is the minimum value of ESS discharge power, kW; is the maximum value of ESS discharge power, kW; State of Charge Constraints: SOC min ≤SOC(t)≤SOC max (18) Where: SOC min and SOC max They represent the lower and upper limits of the energy storage system's state of charge respectively.

10. The method for optimizing the configuration of energy storage capacity in a park according to claim 1, characterized in that: Step 5 uses the Ivy optimization algorithm to solve the optimal configuration model of the energy storage capacity of the wind-solar-storage system to obtain the optimal configuration result of the energy storage capacity of the wind-solar-storage system, which specifically includes: Step 5.1 determines that the ivy optimization algorithm includes: population initialization, coordinated and orderly population growth, obtaining sunlight source growth, propagation and evolution of ivy plants, and selection of survivors; Step 5.2 Population initialization: I i =I min +rand(1,D)⊙(I max -I min ) i=1,…,Npop (19) Where: D is the dimension; I max and I min are the upper and lower bounds of the search space respectively; ⊙ is the product of two vectors; Step 5.3 Coordinated and orderly population growth: ΔGv i (t+1)=rand 2 ⊙[N(1,D)⊙ΔG vi (t)] (20) Where: ΔGv i (t) is the growth rate of the discrete-time system at time t; ΔGv i (t+1) is the growth rate of the discrete time system at time t+1; rand() is a random real number in the interval [0,1]; rand 2 is a random number; N(1,D) represents a random vector with dimension D, which is a random number in the standard Gaussian distribution; Step 5.4 Get sunlight source for growth: Where: |N(1,D)| is a vector whose components are the absolute values ​​of the components of vector N(1,D); is the position of the latest member i; I i is the position of the i-th member; ii is the position of the ii-th member; % is the division of two vectors; Step 5.5 Spread and evolution of ivy plants: Where: I best The best member of the population; is the position of the latest member i; is the growth rate of the current newest members; Step 5.

6. Survivor selection: Where: β = (2 + rand) / 2.

11. The method for optimizing the configuration of park energy storage capacity according to claim 10, characterized in that: According to the wind and solar data and load data predicted in step 4, combined with the solution method of the Ivy optimization algorithm and the established optimization model, the optimal solution set is obtained through multiple iterations, and then the optimal configuration plan of energy storage capacity under the wind, solar and storage system is obtained.

12. A park energy storage capacity optimization configuration system, characterized in that: include: Data acquisition module, data preprocessing module, short-term power prediction module, energy storage capacity optimization configuration analysis module; wherein the data acquisition module is used to execute the method steps of step 1 in the method for optimizing the configuration of the energy storage capacity of a park as described in any one of claims 1-11, the data preprocessing module is used to execute the method steps of step 2 in the method for optimizing the configuration of the energy storage capacity of a park as described in any one of claims 1-11, the short-term power prediction module is used to execute the method steps of step 3 in the method for optimizing the configuration of the energy storage capacity of a park as described in any one of claims 1-11, and the energy storage capacity optimization configuration analysis module is used to execute the method steps of step 4 and step 5 in the method for optimizing the configuration of the energy storage capacity of a park as described in any one of claims 1-11.

13. A storage medium, characterized in that: The storage medium stores instructions that are computer-readable and executable by calculated values, and the computer-executable instructions are used to execute the park energy storage capacity optimization configuration method described in any one of claims 1-11.

Images