Intelligent battery energy storage scheduling method for renewable energy system

Abstract

The invention relates to the technical field of renewable energy scheduling, and discloses an intelligent battery energy storage scheduling method for a renewable energy system, which comprises the following steps: step S101, collecting electrical load data; s102, constructing a photovoltaic sequence, a wind power sequence and a fire power sequence; s103, predicting the electricity load capacity through the time sequence analysis model; step S104, predicting the generating capacity of the photovoltaic generator set, the wind generating set and the thermal generator set; s105, if the sum of the generating capacity of the generator set is larger than or equal to the electricity load capacity, generating electricity in sequence, and otherwise, entering the step S106; s106, generating a power generation plan through an improved particle swarm algorithm; according to the method, the influence of weather conditions and social economic activities is comprehensively considered, the future electricity load capacity is predicted through the time sequence analysis capability of the time sequence analysis model, so that the prediction precision of the electricity load is improved, and the power generation plan is generated through the improved particle swarm algorithm, so that the power generation cost is minimized.

Description

Technical Field

[0001] The present invention relates to the technical field of renewable energy scheduling, and more specifically, to an intelligent battery energy storage scheduling method for a renewable energy system. Background Art

[0002] With the transformation of the global energy structure and the improvement of environmental awareness, the proportion of renewable energy in the power system continues to increase. Among them, renewable energy such as wind power and solar energy has the advantages of being clean and sustainable, but there are significant intermittent and volatile characteristics in the actual application process, which brings huge challenges to the stable operation of the power system. Traditional power system control methods mainly rely on fixed schedules and fixed power generation plans, which are difficult to adapt to the uncertainty of renewable energy, resulting in problems such as waste of power resources and unstable power supply.

[0003] The existing power system adopts a hybrid power generation mode, that is, on the basis of renewable energy power generation, it combines traditional thermal power generation and energy storage battery technology, provides a stable base load power supply through thermal power generation, and the energy storage battery has the functions of rapid response and peak shaving, which effectively smooths the fluctuation of renewable energy and ensures the stability and reliability of the power system. In addition, the existing method obtains the power load data of the historical time period at the current time point as the benchmark data, and then extracts the power load data of the historical time period with the highest similarity to the benchmark data through the similarity algorithm (such as cosine similarity, etc.) combined with the DTW (dynamic time warping) algorithm as the predicted value, and then extracts the power generation plan of the corresponding historical time period as the power generation plan at the current time point.

[0004] However, the above schemes mainly rely on historical electricity load data and have limited responsiveness to complex and changeable weather conditions and socio-economic activities, resulting in low accuracy in electricity load forecasting, which may make the generated power generation plan inapplicable. Summary of the invention

[0005] The present invention provides an intelligent battery energy storage scheduling method for a renewable energy system, which solves the technical problems in the above-mentioned background technology.

[0006] The present invention provides an intelligent battery energy storage scheduling method for a renewable energy system, comprising the following steps:

[0007] Step S101, taking the current time point as the reference time point, moving forward a first time period T1, acquiring power load data according to the first time interval t1, and forming a characteristic sequence;

[0008] The characteristic sequence includes M sequence units, and the mth sequence unit represents the preprocessed power load data at the mth time point, where M=T1 / t1, 1≤m≤M;

[0009] The electricity load data includes: electricity consumption, temperature, rainfall, wind speed, holiday mark, workday mark and special event mark;

[0010] Step S102, taking the current time point as the reference time point, moving forward the first time period T1, obtaining the operating parameters of the photovoltaic power generation group, the wind power generation group and the thermal power generation group according to the second time interval t2, and forming a photovoltaic sequence, a wind sequence and a thermal sequence respectively;

[0011] The photovoltaic sequence, wind sequence and thermal sequence all include N sequence units, and the nth sequence unit represents the operating parameters of the photovoltaic generator set, wind generator set and thermal generator set at the nth time point, respectively, where N = T1 / t2, 1≤n≤N;

[0012] The operating parameters of the photovoltaic generator set include: photovoltaic power generation, effective area of ​​photovoltaic panels, photoelectric conversion efficiency, solar radiation intensity, temperature, temperature coefficient, photovoltaic panel inclination and azimuth;

[0013] The operating parameters of wind turbines include: wind power generation, rated power, wind speed, wind turbine start-up wind speed, wind turbine shutdown wind speed and rotor radius;

[0014] The operating parameters of thermal power generating units include: thermal power generation power and thermal power conversion efficiency;

[0015] Step S103, inputting the characteristic sequence into the time series analysis model, and the output value represents the power load amount in the second time period T2 shifted backward from the current time point as the reference time point;

[0016] Step S104, input the photovoltaic sequence, wind sequence and thermal sequence into the photovoltaic power generation prediction model, wind power generation prediction model and thermal power generation prediction model respectively, and the output values ​​respectively represent the power generation of the photovoltaic power generation group, wind power generation group and thermal power generation group in the second time period T2 with the current time point as the reference time point.

[0017] Further, the first time period T1, the second time period T2, the first time interval t1 and the second time interval t2 are all custom parameters, wherein the duration of the first time period T1 is the same as the duration of the second time period T2.

[0018] Further, preprocessing the power load data includes the following steps:

[0019] Step S201, converting holiday flags, working day flags and special event flags in the power load data into numerical representations;

[0020] The holiday flag, workday flag and special event flag are represented by 0 or 1, where 0 represents no and 1 represents yes;

[0021] Step S202: For missing values ​​of power consumption, temperature, rainfall, and wind speed in the power load data, interpolation filling is performed by taking the average value of data at two adjacent time points of the missing value;

[0022] Step S203: normalize the power load data at each time point using the Z-score standard score method.

[0023] Furthermore, the time series analysis model includes M hidden layers, the mth hidden layer inputs the mth sequence unit of the feature sequence, outputs an update vector, the update vector output by the Mth hidden layer is input to the classifier, and the classification space of the classifier represents the power load amount in the second time period T2 shifted backward with the current time point as the reference time point;

[0024] The calculation formula for the mth hidden layer includes:

[0025] The reset gate r of the mth hidden layer m The calculation formula is as follows:

[0026]

[0027] in and They respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the reset gate of the mth hidden layer;

[0028] The update gate z of the mth hidden layer m The calculation formula is as follows:

[0029]

[0030] in and They respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the update gate of the m-th hidden layer;

[0031] The candidate vector of the mth hidden layer The calculation formula is as follows:

[0032]

[0033] in and They represent the weight parameters and bias parameters corresponding to the candidate vectors of the mth hidden layer respectively;

[0034] The mth sequence unit x of the feature sequence input to the mth hidden layer m and the update vector h of the output of the m-1th hidden layer m-1 The attention score between The calculation formula is as follows:

[0035]

[0036] Where MLP stands for multi-layer perceptron, which is used to transform x m and h m-1 The concatenated result is mapped to a real number representation ranging from 0 to 1;

[0037] The update vector h of the output of the mth hidden layer m The calculation formula is as follows:

[0038]

[0039] Definition: h0 is assigned the value 0, x m represents the mth sequence unit of the feature sequence of the mth hidden layer input, h m-1 represents the update vector of the output of the m-1th hidden layer, ⊙ represents point-by-point multiplication, Swish represents the Swish activation function, tanh represents the hyperbolic tangent function, and concat represents the concatenation operation.

[0040] Furthermore, the calculation formula of the photovoltaic power generation prediction model includes:

[0041]

[0042]

[0043] in It represents the power generation of the photovoltaic power generation group in the second time period T2 after the current time point as the reference time point. affect n , G n , A n , θ n and T n They represent the photovoltaic power generation, photovoltaic power generation influencing factor, solar radiation intensity, photovoltaic panel effective area, photoelectric conversion efficiency, angle factor and temperature of the photovoltaic generator set at the nth time point, θ n It is equal to the photovoltaic panel inclination angle minus the azimuth angle at the nth time point, α represents the temperature coefficient, T std Indicates standard temperature, assigned a value of 25, represents the duration of the first time period T1, λ1 and λ2 represent the first penalty coefficient and the second penalty coefficient respectively, both of which are obtained by least squares fitting.

[0044] Furthermore, the calculation formula of the wind power generation prediction model is as follows:

[0045] definition: represents the power generation of the wind turbine generator set in the second time period T2 after the current time point as the reference time point, P rated Indicates the rated power of the wind turbine generator set, v rated Indicates the rated wind speed corresponding to the rated power of the wind turbine, v start and v stop They represent the wind turbine startup wind speed and wind turbine shutdown wind speed of the wind turbine generator set, v n represents the wind speed at the nth time point;

[0046] When v n Less than v start or v n Greater than v stop ,but is equal to 0;

[0047] When v n Greater than or equal to v start And v n Less than or equal to v rated ,but The calculation formula is as follows:

[0048]

[0049] in represents the wind power generated by the wind turbine at the nth time point, r represents the radius of the wind rotor, and ρ represents the air density, which is assigned a value of 1.225 kg / m 3 , C p Indicates the power coefficient, with a value of 0.35. represents the duration of the first time period T1, λ3 and λ4 represent the third penalty coefficient and the fourth penalty coefficient respectively, both of which are obtained by least squares fitting;

[0050] When v n Greater than v rated And v n Less than or equal to v stop ,but Equal to P rated .

[0051] Furthermore, the calculation formula of the thermal power generation prediction model is as follows:

[0052]

[0053] in It represents the power generation of the thermal power generating unit in the second time period T2 after the current time point as the reference time point. and They represent the thermal power generation and thermal power conversion efficiency of the thermal power generating unit at the nth time point, represents the duration of the first time period T1, λ5 and λ6 represent the fifth penalty coefficient and the sixth penalty coefficient respectively, both of which are obtained by least squares fitting.

[0054] Furthermore, the method further comprises the following steps:

[0055] Step S105, judging that the sum of the power generation of the photovoltaic power generation group, the wind power generation group and the thermal power generation group in the second time period T2 after the current time point is greater than or equal to the power load, then the wind power generation group, the photovoltaic power generation group and the thermal power generation group participate in power generation in sequence, otherwise, proceeding to step S106;

[0056] Step S106, generating a power generation plan by using an improved particle swarm algorithm based on the power load amount, the power generation amount of the photovoltaic power generation group, the wind power generation group and the thermal power generation group in the second time period T2, and the discharge amount of the photovoltaic energy storage battery and the wind energy storage battery at the current time point, with the current time point as the reference time point;

[0057] The power generation plan represents the actual power generation of each photovoltaic power generation group, wind power generation group and thermal power generation group and the actual discharge of photovoltaic energy storage batteries and wind energy storage batteries in the second time period T2 with the current time point as the reference time point.

[0058] Furthermore, the power generation plan is generated by the improved particle swarm algorithm, including the following steps:

[0059] Step S301, randomly generate position vector codes and velocity vector codes of K particles of the initial population that meet the constraint conditions, and initialize the current number of iterations to 1;

[0060] Where K is a custom parameter;

[0061] The constraints include: the total power generation of all photovoltaic generators, wind generators and thermal generators, as well as the discharge of photovoltaic energy storage batteries and wind energy storage batteries must be greater than or equal to the power load; the power generation of each photovoltaic generator, wind generator and thermal generator, as well as the discharge of photovoltaic energy storage batteries and wind energy storage batteries cannot exceed the preset range, where the preset range is a custom parameter;

[0062] The position vector encoding includes U dimensional values, each of which corresponds to the actual power generation of a photovoltaic generator set, a wind generator set, or a thermal generator set, or the actual discharge of a photovoltaic energy storage battery or a wind energy storage battery, where U is equal to the total number of photovoltaic generator sets, wind generator sets, thermal generator sets, photovoltaic energy storage batteries, and wind energy storage batteries;

[0063] The velocity vector encoding includes U dimension values, and the u-th dimension value is 15% of the u-th dimension value of the position vector encoding, where 1≤u≤U;

[0064] Step S302, calculating the fitness value through the objective function according to the position vector encoding of K particles of the initialized population;

[0065] The objective function is calculated as follows:

[0066]

[0067] Where 1≤k≤K, Fit k represents the fitness value of the kth particle, The actual power generation of the photovoltaic generator set, wind generator set, and thermal generator set corresponding to the u-th dimension value encoded by the position vector of the k-th particle, or the actual discharge of the photovoltaic energy storage battery and wind energy storage battery, represents the corresponding power generation cost;

[0068] Step S303, generating a speed pause parameter according to the current number of iterations, and if the speed pause parameter is greater than or equal to 1, proceeding to step S304, otherwise proceeding to step S305;

[0069] The speed pause parameter Stop for the tth iteration t The calculation formula is as follows:

[0070]

[0071] Where e represents a natural constant, and T represents the maximum number of iterations;

[0072] Step S304, updating the position vector codes and velocity vector codes of K particles of the initialized population by using the first updating strategy;

[0073] Step S305, randomly select K / 2 particles to update the position vector code and velocity vector code of the particles through the first update strategy, and update the position vector code and velocity vector code of the remaining K / 2 particles through the second update strategy;

[0074] Step S306, the current iteration number is added by 1, and steps S302 to S303 are repeated until the current iteration number is greater than or equal to the maximum iteration number, and the position vector of the particle with the largest fitness value in the initialized population is encoded as a power generation plan;

[0075] The maximum number of iterations is a custom parameter.

[0076] Furthermore, the calculation formula of the first update strategy includes:

[0077]

[0078]

[0079] Where 1≤k≤K, 1≤u≤U, and The u-th dimension value of the velocity vector encoding of the k-th particle at the t+1-th iteration and the t-th iteration, respectively. and The u-th dimension value of the position vector encoding of the k-th particle at the t+1-th iteration and the t-th iteration, pbest t,u Indicates the u-th dimension value of the position vector encoding of the particle with the largest fitness value at the t-th iteration, gbest u The u-th dimension value of the position vector encoding of the particle with the largest fitness value in the historical iteration number, r1 and r2 represent the first random number and the second random number in the range of 0 to 1, respectively, and c1 and c2 represent the first displacement coefficient and the second displacement coefficient in the range of 0 to 2, respectively;

[0080] The calculation formula of the second update strategy includes:

[0081]

[0082]

[0083] Among them, Stop t It represents the speed pause parameter of the tth iteration, and rand represents a random number ranging from 0 to 1.

[0084] The beneficial effects of the present invention are as follows: the present invention comprehensively considers the influence of weather conditions and socio-economic activities, establishes a nonlinear mapping relationship between historical power load data and future power load through a time series analysis model, thereby improving the prediction accuracy of power load, and generates a power generation plan through an improved particle swarm algorithm, thereby ensuring that the power generation cost is minimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 is a flow chart of a smart battery energy storage scheduling method for a renewable energy system of the present invention;

[0086] Figure 2 is a flow chart of preprocessing power load data of the present invention;

[0087] Figure 3 The flowchart of the present invention is to generate a power generation plan by using an improved particle swarm algorithm. DETAILED DESCRIPTION

[0088] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0089] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0090] like Figure 1 to Figure 3 As shown, a smart battery energy storage scheduling method for a renewable energy system comprises the following steps:

[0091] Step S101, taking the current time point as the reference time point, moving forward a first time period T1, acquiring power load data according to the first time interval t1, and forming a characteristic sequence;

[0092] The characteristic sequence includes M sequence units, and the mth sequence unit represents the preprocessed power load data at the mth time point, where M=T1 / t1, 1≤m≤M;

[0093] The electricity load data includes: electricity consumption, temperature, rainfall, wind speed, holiday mark, workday mark and special event mark;

[0094] Step S102, taking the current time point as the reference time point, moving forward the first time period T1, obtaining the operating parameters of the photovoltaic power generation group, the wind power generation group and the thermal power generation group according to the second time interval t2, and forming a photovoltaic sequence, a wind sequence and a thermal sequence respectively;

[0095] The photovoltaic sequence, wind sequence and thermal sequence all include N sequence units, and the nth sequence unit represents the operating parameters of the photovoltaic generator set, wind generator set and thermal generator set at the nth time point, respectively, where N = T1 / t2, 1≤n≤N;

[0096] The operating parameters of the photovoltaic generator set include: photovoltaic power generation, effective area of ​​photovoltaic panels, photoelectric conversion efficiency, solar radiation intensity, temperature, temperature coefficient, photovoltaic panel inclination and azimuth;

[0097] The operating parameters of wind turbines include: wind power generation, rated power, wind speed, wind turbine start-up wind speed, wind turbine shutdown wind speed and rotor radius;

[0098] The operating parameters of thermal power generating units include: thermal power generation power and thermal power conversion efficiency;

[0099] Step S103, inputting the characteristic sequence into the time series analysis model, and the output value represents the power load amount in the second time period T2 shifted backward from the current time point as the reference time point;

[0100] Step S104, input the photovoltaic sequence, wind sequence and thermal sequence into the photovoltaic power generation prediction model, wind power generation prediction model and thermal power generation prediction model respectively, and the output values ​​respectively represent the power generation of the photovoltaic power generation group, wind power generation group and thermal power generation group in the second time period T2 with the current time point as the reference time point.

[0101] In one embodiment of the present invention, the first time period T1, the second time period T2, the first time interval t1 and the second time interval t2 are all custom parameters, wherein the duration of the first time period T1 is the same as the duration of the second time period T2. Preferably, T1 and T2 are set to 1 hour, t1 is set to 5 minutes, and t2 is set to 1 minute.

[0102] In one embodiment of the present invention, Figure 2 As shown, preprocessing the power load data includes the following steps:

[0103] Step S201, converting holiday flags, working day flags and special event flags in the power load data into numerical representations;

[0104] The holiday flag, workday flag and special event flag are represented by 0 or 1, where 0 represents no and 1 represents yes; for example, a holiday flag of 0 represents that the current time point is not a holiday, and a workday flag of 1 represents that the current time point is a workday;

[0105] Step S202: For missing values ​​of power consumption, temperature, rainfall, and wind speed in the power load data, interpolation filling is performed by taking the average value of data at two adjacent time points of the missing value;

[0106] For example, if the power consumption at the first time point is missing, the average power consumption at the second and third time points is used for interpolation. If the temperature at the second time point is missing, the average temperature at the first and third time points is used for interpolation.

[0107] Step S203, normalizing the power load data at each time point by using the Z-score standard score method;

[0108] For example, the calculation formula for normalizing the power consumption at each time point is as follows:

[0109]

[0110] Among them, m and elect m They represent the power consumption at the mth time point after and before preprocessing, respectively, and avg_elec and S represent the average and standard deviation of power consumption, respectively.

[0111] It should be noted that the interpolation and filling processing is used to ensure the integrity of the power load data, and the normalization processing is used to eliminate the dimensional influence of the power load data. To a certain extent, it can accelerate the convergence speed of the timing analysis model and improve the robustness of the timing analysis model. In addition, the power load data can also be feature extracted, such as taking the average value, standard deviation, maximum value and minimum value, etc., which will not be elaborated here.

[0112] In one embodiment of the present invention, the time series analysis model includes M hidden layers, the mth hidden layer inputs the mth sequence unit of the feature sequence, outputs an update vector, the update vector output by the Mth hidden layer is input to the classifier, and the classification space of the classifier represents the power load amount of the second time period T2 shifted backward with the current time point as the reference time point;

[0113] The calculation formula for the mth hidden layer includes:

[0114] The reset gate r of the mth hidden layer m The calculation formula is as follows:

[0115]

[0116] in and They respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the reset gate of the mth hidden layer;

[0117] The update gate z of the mth hidden layer m The calculation formula is as follows:

[0118]

[0119] in and They respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the update gate of the m-th hidden layer;

[0120] The candidate vector of the mth hidden layer The calculation formula is as follows:

[0121]

[0122] in and They represent the weight parameters and bias parameters corresponding to the candidate vectors of the mth hidden layer respectively;

[0123] The mth sequence unit x of the feature sequence input to the mth hidden layer m and the update vector h of the output of the m-1th hidden layer m-1 The attention score between The calculation formula is as follows:

[0124]

[0125] Where MLP stands for multi-layer perceptron, which is used to transform x m and h m-1 The concatenated result is mapped to a real number representation ranging from 0 to 1;

[0126] The update vector h of the output of the mth hidden layer m The calculation formula is as follows:

[0127]

[0128] Definition: h0 is assigned the value 0, x m represents the mth sequence unit of the feature sequence of the mth hidden layer input, h m-1 represents the update vector of the output of the m-1th hidden layer, ⊙ represents point-by-point multiplication, Swish represents the Swish activation function, tanh represents the hyperbolic tangent function, and concat represents the concatenation operation.

[0129] It should be noted that the weight parameters and bias parameters in the time series analysis model are all learnable parameters. For example, the sequence unit of the feature sequence is a vector of size 1×7, and each dimension value corresponds to the applied power, temperature, rainfall, wind speed, holiday sign, working day sign and special event sign respectively. Assuming that the size of the update vector output by each hidden layer is 1×16, the size of the candidate vector is also 1×16. The reset gate and the update gate are both represented by real values. The first weight parameter corresponding to the reset gate and the update gate can be designed as a vector of size 7×1, and the second weight parameter corresponding to the reset gate and the update gate can be designed as a vector of size 16×1. The weight parameter corresponding to the candidate vector needs to be designed as a matrix of size 23×16. The activation function of the classifier is the softmax activation function, which is not described here.

[0130] In one embodiment of the present invention, the calculation formula of the photovoltaic power generation prediction model includes:

[0131]

[0132]

[0133] in It represents the power generation of the photovoltaic power generation group in the second time period T2 after the current time point as the reference time point. affect n , G n , A n , θ n and T n They represent the photovoltaic power generation, photovoltaic power generation influencing factor, solar radiation intensity, photovoltaic panel effective area, photoelectric conversion efficiency, angle factor and temperature of the photovoltaic generator set at the nth time point, θ n It is equal to the photovoltaic panel inclination angle minus the azimuth angle at the nth time point, α represents the temperature coefficient, T std Indicates standard temperature, assigned a value of 25, represents the duration of the first time period T1, λ1 and λ2 represent the first penalty coefficient and the second penalty coefficient respectively, both of which are obtained by least squares fitting.

[0134] It should be noted that the photoelectric conversion efficiency is provided by the manufacturer of the photovoltaic power generation unit. The temperature coefficient represents the decrease ratio of the photovoltaic power generation efficiency with the increase of temperature, which is also provided by the manufacturer of the photovoltaic power generation unit. The first penalty coefficient and the second penalty coefficient are obtained by fitting the actual photovoltaic power generation by the least squares method, that is, the mean square error between the predicted value of the photovoltaic power generation output by the photovoltaic power generation prediction model and the actual photovoltaic power generation is specified as the loss function. The first penalty coefficient and the second penalty coefficient are updated by the gradient descent algorithm to minimize the loss of the loss function. It will not be elaborated here.

[0135] In one embodiment of the present invention, the calculation formula of the wind power generation prediction model is as follows:

[0136] definition: represents the power generation of the wind turbine generator set in the second time period T2 after the current time point as the reference time point, P rated Indicates the rated power of the wind turbine generator set, v rated Indicates the rated wind speed corresponding to the rated power of the wind turbine, v start and v stop They represent the wind turbine startup wind speed and wind turbine shutdown wind speed of the wind turbine generator set, v n represents the wind speed at the nth time point;

[0137] When v n Less than v start or v n Greater than v stop ,but is equal to 0;

[0138] When v n Greater than or equal to v start And v n Less than or equal to v rated ,but The calculation formula is as follows:

[0139]

[0140] in represents the wind power generated by the wind turbine at the nth time point, r represents the radius of the wind rotor, and ρ represents the air density, which is assigned a value of 1.225 kg / m 3 , C p Indicates the power coefficient, with a value of 0.35. represents the duration of the first time period T1, λ3 and λ4 represent the third penalty coefficient and the fourth penalty coefficient respectively, both of which are obtained by least squares fitting;

[0141] When v n Greater than v rated And v n Less than or equal to v stop ,but Equal to P rated .

[0142] It should be noted that, similarly, the third penalty coefficient and the fourth penalty coefficient are obtained by fitting the actual wind power generation by the least squares method, that is, the mean square error between the predicted value of the wind power generation output by the wind power generation prediction model and the actual wind power generation is specified as the loss function, and the third penalty coefficient and the fourth penalty coefficient are updated by the gradient descent algorithm to minimize the loss of the loss function, which will not be elaborated here.

[0143] In one embodiment of the present invention, the calculation formula of the thermal power generation prediction model is as follows:

[0144]

[0145] in It represents the power generation of the thermal power generating unit in the second time period T2 after the current time point as the reference time point. and They represent the thermal power generation and thermal power conversion efficiency of the thermal power generating unit at the nth time point, represents the duration of the first time period T1, λ5 and λ6 represent the fifth penalty coefficient and the sixth penalty coefficient respectively, both of which are obtained by least squares fitting.

[0146] In one embodiment of the present invention, a smart battery energy storage scheduling method for a renewable energy system further includes the following steps:

[0147] Step S105, judging that the sum of the power generation of the photovoltaic power generation group, the wind power generation group and the thermal power generation group in the second time period T2 after the current time point is greater than or equal to the power load, then the wind power generation group, the photovoltaic power generation group and the thermal power generation group participate in power generation in sequence, otherwise, proceeding to step S106;

[0148] Step S106, generating a power generation plan by using an improved particle swarm algorithm based on the power load amount, the power generation amount of the photovoltaic power generation group, the wind power generation group and the thermal power generation group in the second time period T2, and the discharge amount of the photovoltaic energy storage battery and the wind energy storage battery at the current time point, with the current time point as the reference time point;

[0149] The power generation plan represents the actual power generation of each photovoltaic power generation group, wind power generation group and thermal power generation group and the actual discharge of photovoltaic energy storage batteries and wind energy storage batteries in the second time period T2 with the current time point as the reference time point.

[0150] In one embodiment of the present invention, Figure 3 As shown in the figure, the power generation plan is generated by the improved particle swarm algorithm, which includes the following steps:

[0151] Step S301, randomly generate position vector codes and velocity vector codes of K particles of the initial population that meet the constraint conditions, and initialize the current number of iterations to 1;

[0152] Where K is a custom parameter, preferably, K is set to 20;

[0153] The constraints include: the total power generation of all photovoltaic generators, wind generators and thermal generators, as well as the discharge of photovoltaic energy storage batteries and wind energy storage batteries must be greater than or equal to the power load; the power generation of each photovoltaic generator, wind generator and thermal generator, as well as the discharge of photovoltaic energy storage batteries and wind energy storage batteries cannot exceed the preset range, where the preset range is a custom parameter;

[0154] The position vector encoding includes U dimensional values, each of which corresponds to the actual power generation of a photovoltaic generator set, a wind generator set, or a thermal generator set, or the actual discharge of a photovoltaic energy storage battery or a wind energy storage battery, where U is equal to the total number of photovoltaic generator sets, wind generator sets, thermal generator sets, photovoltaic energy storage batteries, and wind energy storage batteries;

[0155] The velocity vector encoding includes U dimension values, and the u-th dimension value is 15% of the u-th dimension value of the position vector encoding, where 1≤u≤U;

[0156] Step S302, calculating the fitness value through the objective function according to the position vector encoding of K particles of the initialized population;

[0157] The objective function is calculated as follows:

[0158]

[0159] Where 1≤k≤K, Fit k represents the fitness value of the kth particle, The actual power generation of the photovoltaic generator set, wind generator set, and thermal generator set corresponding to the u-th dimension value encoded by the position vector of the k-th particle, or the actual discharge of the photovoltaic energy storage battery and wind energy storage battery, represents the corresponding power generation cost;

[0160] Step S303, generating a speed pause parameter according to the current number of iterations, and if the speed pause parameter is greater than or equal to 1, proceeding to step S304, otherwise proceeding to step S305;

[0161] The speed pause parameter Stop for the tth iteration t The calculation formula is as follows:

[0162]

[0163] Where e represents a natural constant, and T represents the maximum number of iterations;

[0164] Step S304, updating the position vector codes and velocity vector codes of K particles of the initialized population by using the first updating strategy;

[0165] Step S305, randomly select K / 2 particles to update the position vector code and velocity vector code of the particles through the first update strategy, and update the position vector code and velocity vector code of the remaining K / 2 particles through the second update strategy;

[0166] Step S306, the current iteration number is added by 1, and steps S302 to S303 are repeated until the current iteration number is greater than or equal to the maximum iteration number, and the position vector of the particle with the largest fitness value in the initialized population is encoded as a power generation plan;

[0167] The maximum number of iterations is a custom parameter. Preferably, the maximum number of iterations is set to 50.

[0168] It should be noted that the dimension value of the particle is controlled by setting constraints. For example, if the power generation of the photovoltaic power generation group obtained by the photovoltaic power generation prediction model is 1000kWh, then the maximum power generation of the corresponding photovoltaic power generation group is 1000kWh, that is, the maximum value of the dimension can only be 0 to 1000. In addition, the discharge capacity of the energy storage battery is obtained by multiplying the remaining capacity of the energy storage battery by the discharge efficiency, where the discharge efficiency is a custom parameter. Preferably, the discharge efficiency is set to 80%.

[0169] In one embodiment of the present invention, the calculation formula of the first update strategy includes:

[0170]

[0171]

[0172] Where 1≤k≤K, 1≤u≤U, and The u-th dimension value of the velocity vector encoding of the k-th particle at the t+1-th iteration and the t-th iteration, respectively. and The u-th dimension value of the position vector encoding of the k-th particle at the t+1-th iteration and the t-th iteration, pbest t,u Indicates the u-th dimension value of the position vector encoding of the particle with the largest fitness value at the t-th iteration, gbest uThe u-th dimension value of the position vector encoding of the particle with the largest fitness value in the historical iteration number, r1 and r2 represent the first random number and the second random number in the range of 0 to 1, respectively, and c1 and c2 represent the first displacement coefficient and the second displacement coefficient in the range of 0 to 2, respectively;

[0173] The calculation formula of the second update strategy includes:

[0174]

[0175]

[0176] Among them, Stop t It represents the speed pause parameter of the tth iteration, and rand represents a random number ranging from 0 to 1.

[0177] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are protected by the present embodiment.

Claims

1. A smart battery energy storage scheduling method for a renewable energy system, characterized in that: The following steps are involved: Step S101, taking the current time point as the reference time point, moving forward a first time period T1, acquiring power load data according to the first time interval t1, and forming a characteristic sequence; The characteristic sequence includes M sequence units, and the mth sequence unit represents the preprocessed power load data at the mth time point, where M=T1 / t1, 1≤m≤M; The electricity load data includes: electricity consumption, temperature, rainfall, wind speed, holiday mark, workday mark and special event mark; Step S102, taking the current time point as the reference time point, moving forward the first time period T1, obtaining the operating parameters of the photovoltaic power generation group, the wind power generation group and the thermal power generation group according to the second time interval t2, and forming a photovoltaic sequence, a wind sequence and a thermal sequence respectively; The photovoltaic sequence, wind sequence and thermal sequence all include N sequence units, and the nth sequence unit represents the operating parameters of the photovoltaic generator set, wind generator set and thermal generator set at the nth time point, respectively, where N = T1 / t2, 1≤n≤N; The operating parameters of the photovoltaic generator set include: photovoltaic power generation, effective area of ​​photovoltaic panels, photoelectric conversion efficiency, solar radiation intensity, temperature, temperature coefficient, photovoltaic panel inclination and azimuth; The operating parameters of wind turbines include: wind power generation, rated power, wind speed, wind turbine start-up wind speed, wind turbine shutdown wind speed and rotor radius; The operating parameters of thermal power generating units include: thermal power generation power and thermal power conversion efficiency; Step S103, inputting the characteristic sequence into the time series analysis model, and the output value represents the power load amount in the second time period T2 shifted backward from the current time point as the reference time point; Step S104, input the photovoltaic sequence, wind sequence and thermal sequence into the photovoltaic power generation prediction model, wind power generation prediction model and thermal power generation prediction model respectively, and the output values ​​respectively represent the power generation of the photovoltaic power generation group, wind power generation group and thermal power generation group in the second time period T2 with the current time point as the reference time point.

2. The intelligent battery energy storage scheduling method for renewable energy system according to claim 1, characterized in that: The first time period T1, the second time period T2, the first time interval t1 and the second time interval t2 are all custom parameters, wherein the duration of the first time period T1 is the same as the duration of the second time period T2.

3. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 1, characterized in that: Preprocessing the power load data includes the following steps: Step S201, converting holiday flags, working day flags and special event flags in the power load data into numerical representations; The holiday flag, workday flag and special event flag are represented by 0 or 1, where 0 represents no and 1 represents yes; Step S202: For missing values ​​of power consumption, temperature, rainfall, and wind speed in the power load data, interpolation filling is performed by taking the average value of data at two adjacent time points of the missing value; Step S203: normalize the power load data at each time point using the Z-score standard score method.

4. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 1, characterized in that: The time series analysis model includes M hidden layers, the mth hidden layer inputs the mth sequence unit of the feature sequence, outputs an update vector, the update vector output by the Mth hidden layer is input to the classifier, and the classification space of the classifier represents the power load amount in the second time period T2 shifted backward with the current time point as the reference time point; The calculation formula for the mth hidden layer includes: The reset gate r of the mth hidden layer m The calculation formula is as follows: in and They respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the reset gate of the mth hidden layer; The update gate z of the mth hidden layer m The calculation formula is as follows: in and They respectively represent the first weight parameter, the second weight parameter and the bias parameter corresponding to the update gate of the m-th hidden layer; The candidate vector of the mth hidden layer The calculation formula is as follows: in and They represent the weight parameters and bias parameters corresponding to the candidate vectors of the mth hidden layer respectively; The mth sequence unit x of the feature sequence input to the mth hidden layer m and the update vector h of the output of the m-1th hidden layer m-1 The attention score between The calculation formula is as follows: Where MLP stands for multi-layer perceptron, which is used to transform x m and h m-1 The concatenated result is mapped to a real number representation ranging from 0 to 1; The update vector h of the output of the mth hidden layer m The calculation formula is as follows: Definition: h0 is assigned the value 0, x m represents the mth sequence unit of the feature sequence of the mth hidden layer input, h m-1 represents the update vector of the output of the m-1th hidden layer, ⊙ represents point-by-point multiplication, Swish represents the Swish activation function, tanh represents the hyperbolic tangent function, and concat represents the concatenation operation.

5. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 1, characterized in that: The calculation formula of the photovoltaic power generation prediction model includes: in It represents the power generation of the photovoltaic power generation group in the second time period T2 after the current time point as the reference time point. affect n , G n , A n , θ n and T n They represent the photovoltaic power generation, photovoltaic power generation influencing factor, solar radiation intensity, photovoltaic panel effective area, photoelectric conversion efficiency, angle factor and temperature of the photovoltaic generator set at the nth time point, θ n It is equal to the photovoltaic panel inclination angle minus the azimuth angle at the nth time point, α represents the temperature coefficient, T std Indicates standard temperature, assigned a value of 25, represents the duration of the first time period T1, λ1 and λ2 represent the first penalty coefficient and the second penalty coefficient respectively, both of which are obtained by least squares fitting.

6. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 1, characterized in that: The calculation formula of the wind power generation prediction model is as follows: definition: represents the power generation of the wind turbine generator set in the second time period T2 after the current time point as the reference time point, P rated Indicates the rated power of the wind turbine generator set, v rated Indicates the rated wind speed corresponding to the rated power of the wind turbine, v start and v stop They represent the wind turbine startup wind speed and wind turbine shutdown wind speed of the wind turbine generator set, v n represents the wind speed at the nth time point; When v n Less than v start or v n Greater than v stop ,but is equal to 0; When v n Greater than or equal to v start And v n Less than or equal to v rated ,but The calculation formula is as follows: Where P w n ind represents the wind power generated by the wind turbine at the nth time point, r represents the radius of the wind rotor, and ρ represents the air density, which is assigned a value of 1.225 kg / m 3 , C p Indicates the power coefficient, with a value of 0.

35. represents the duration of the first time period T1, λ3 and λ4 represent the third penalty coefficient and the fourth penalty coefficient respectively, both of which are obtained by least squares fitting; When v n Greater than v rated And v n Less than or equal to v stop ,but Equal to P rated .

7. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 1, characterized in that: The calculation formula of the thermal power generation prediction model is as follows: in It represents the power generation of the thermal power generating unit in the second time period T2 after the current time point as the reference time point. and They represent the thermal power generation and thermal power conversion efficiency of the thermal power generating unit at the nth time point, represents the duration of the first time period T1, λ5 and λ6 represent the fifth penalty coefficient and the sixth penalty coefficient respectively, both of which are obtained by least squares fitting.

8. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 1, characterized in that: The following steps are also included: Step S105, judging that the sum of the power generation of the photovoltaic power generation group, the wind power generation group and the thermal power generation group in the second time period T2 after the current time point is greater than or equal to the power load, then the wind power generation group, the photovoltaic power generation group and the thermal power generation group participate in power generation in sequence, otherwise, proceeding to step S106; Step S106, generating a power generation plan by using an improved particle swarm algorithm based on the power load amount, the power generation amount of the photovoltaic power generation group, the wind power generation group and the thermal power generation group in the second time period T2, and the discharge amount of the photovoltaic energy storage battery and the wind energy storage battery at the current time point, with the current time point as the reference time point; The power generation plan represents the actual power generation of each photovoltaic power generation group, wind power generation group and thermal power generation group and the actual discharge of photovoltaic energy storage batteries and wind energy storage batteries in the second time period T2 with the current time point as the reference time point.

9. The intelligent battery energy storage scheduling method for renewable energy systems according to claim 8, characterized in that: The power generation plan is generated by the improved particle swarm algorithm, which includes the following steps: Step S301, randomly generate position vector codes and velocity vector codes of K particles of the initial population that meet the constraint conditions, and initialize the current number of iterations to 1; Where K is a custom parameter; The constraints include: the total power generation of all photovoltaic generators, wind generators and thermal generators, as well as the discharge of photovoltaic energy storage batteries and wind energy storage batteries must be greater than or equal to the power load; the power generation of each photovoltaic generator, wind generator and thermal generator, as well as the discharge of photovoltaic energy storage batteries and wind energy storage batteries cannot exceed the preset range, where the preset range is a custom parameter; The position vector encoding includes U dimensional values, each of which corresponds to the actual power generation of a photovoltaic generator set, a wind generator set, or a thermal generator set, or the actual discharge of a photovoltaic energy storage battery or a wind energy storage battery, where U is equal to the total number of photovoltaic generator sets, wind generator sets, thermal generator sets, photovoltaic energy storage batteries, and wind energy storage batteries; The velocity vector encoding includes U dimension values, and the u-th dimension value is 15% of the u-th dimension value of the position vector encoding, where 1≤u≤U; Step S302, calculating the fitness value through the objective function according to the position vector encoding of K particles of the initialized population; The objective function is calculated as follows: Where 1≤k≤K, Fit k represents the fitness value of the kth particle, The actual power generation of the photovoltaic generator set, wind generator set, and thermal generator set corresponding to the u-th dimension value encoded by the position vector of the k-th particle, or the actual discharge of the photovoltaic energy storage battery and wind energy storage battery, represents the corresponding power generation cost; Step S303, generating a speed pause parameter according to the current number of iterations, and if the speed pause parameter is greater than or equal to 1, proceeding to step S304, otherwise proceeding to step S305; The speed pause parameter Stop for the tth iteration t The calculation formula is as follows: Where e represents a natural constant, and T represents the maximum number of iterations; Step S304, updating the position vector codes and velocity vector codes of K particles of the initialized population by using the first updating strategy; Step S305, randomly select K / 2 particles to update the position vector code and velocity vector code of the particles through the first update strategy, and update the position vector code and velocity vector code of the remaining K / 2 particles through the second update strategy; Step S306, the current iteration number is added by 1, and steps S302 to S303 are repeated until the current iteration number is greater than or equal to the maximum iteration number, and the position vector of the particle with the largest fitness value in the initialized population is encoded as a power generation plan; The maximum number of iterations is a custom parameter.

10. The intelligent battery energy storage scheduling method for renewable energy system according to claim 9, characterized in that: The calculation formula of the first update strategy includes: Where 1≤k≤K, 1≤u≤U, and The u-th dimension value of the velocity vector encoding of the k-th particle at the t+1-th iteration and the t-th iteration, respectively. and The u-th dimension value of the position vector encoding of the k-th particle at the t+1-th iteration and the t-th iteration, pbest t,u Indicates the u-th dimension value of the position vector encoding of the particle with the largest fitness value at the t-th iteration, gbest u The u-th dimension value of the position vector encoding of the particle with the largest fitness value in the historical iteration number, r1 and r2 represent the first random number and the second random number in the range of 0 to 1, respectively, and c1 and c2 represent the first displacement coefficient and the second displacement coefficient in the range of 0 to 2, respectively; The calculation formula of the second update strategy includes: Among them, Stop t It represents the speed pause parameter of the tth iteration, and rand represents a random number ranging from 0 to 1.

Images