Power distribution network photovoltaic power identification method and system

Abstract

This invention discloses a method for identifying photovoltaic (PV) power in a distribution network, comprising: acquiring historical distribution area data and historical meteorological data of the target distribution area; constructing a relationship fitting model; using a relationship fitting network to fit the nonlinear mapping function between irradiance and PV output power, as well as the nonlinear mapping function between meteorological data and PV output power, to obtain the identification result of PV output power; constructing an evaluation function; calculating the evaluation result and repeating the above steps to obtain the PV output power identification result; calculating the error value and repeating the above steps to obtain the distribution network PV power identification model; and using the obtained distribution network PV power identification model to perform actual distribution network PV power identification. This invention also discloses a system for implementing the aforementioned distribution network PV power identification method. This invention not only enables the identification of PV power in a distribution network but also exhibits high reliability and accuracy.

Description

Technical Field

[0001] This invention belongs to the field of electrical automation, and specifically relates to a method and system for identifying photovoltaic power in a power distribution network. Background Technology

[0002] With economic and technological development and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and daily life, bringing endless convenience. Therefore, ensuring a stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Currently, with the increasing severity of environmental problems, more and more old photovoltaic (PV) power generation systems are being connected to the power grid. However, as the total capacity of PV power generation increases, the load side of the distribution network will shift from a passive to an active network, and the dynamic characteristics of the distribution network will change accordingly. Therefore, a high proportion of distributed PV access will have an increasingly significant impact on the power system. Furthermore, the seasonality and daily output variation characteristics of high-proportion grid-connected distributed PV systems also pose new challenges to the safe and stable operation of the power grid. Therefore, accurate PV power identification is beneficial to the safe, stable, and optimized operation of PV power generation systems, and is also a necessary means to improve the stable operation of the distribution system.

[0004] Current research on distributed photovoltaic (PV) systems, both domestically and internationally, assumes that the actual output data of PV power sources is continuously available, and that its time resolution and time scale meet research requirements. However, for distributed PV power sources decentralized to low-voltage distribution areas, obtaining their output data individually is difficult due to limitations in technology and investment costs (typically only hourly power data is uploaded). Furthermore, distributed PV power sources are located at dispersed user locations, with significant differences in models and characteristics, making it difficult to investigate, statistically analyze, and aggregate data to obtain a mapping model reflecting the entire distribution area. Therefore, existing traditional PV power identification methods are not suitable for distribution areas within power systems. Summary of the Invention

[0005] One of the objectives of this invention is to provide a highly reliable and accurate method for identifying photovoltaic power in power distribution networks.

[0006] The second objective of this invention is to provide a system for implementing the photovoltaic power identification method for the power distribution network.

[0007] The photovoltaic power identification method for power distribution networks provided by this invention includes the following steps:

[0008] S1. Obtain historical station data and historical meteorological data for the target distribution station area;

[0009] S2. Construct a relationship fitting model based on RNN network;

[0010] S3. Based on the data obtained in step S1, use the relationship fitting network constructed in step S2 to fit the nonlinear mapping function between light intensity and photovoltaic output power;

[0011] S4. Based on the nonlinear mapping function obtained in step S3, the relationship fitting network constructed in step S2 is used to fit the nonlinear mapping function between meteorological data and photovoltaic output power, thereby obtaining the identification result of photovoltaic output power;

[0012] S5. Based on the photovoltaic output power identification results obtained in step S4, construct an evaluation function and calculate the evaluation results;

[0013] S6. Based on the evaluation results obtained in step S5, repeat steps S3 to S6 until the photovoltaic output power identification result that meets the set conditions is obtained;

[0014] S7. Calculate the error between the final photovoltaic output power identification result and the actual photovoltaic output power result, and repeat steps S3 to S7 according to the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained.

[0015] S8. Based on the photovoltaic power identification model of the distribution network obtained in step S7, perform actual photovoltaic power identification of the distribution network.

[0016] Step S1, which involves obtaining historical station data and historical meteorological data of the target distribution station area, specifically includes the following steps:

[0017] Obtain historical data information of the target distribution transformer area, including the output power data, photovoltaic output power data, and load power data of the target distribution transformer area;

[0018] Obtain historical meteorological data for the target distribution station area, including weather data, humidity data, air pressure data, precipitation probability data, and actual temperature data for the target distribution station area.

[0019] Step S2, which involves constructing a relationship fitting model based on an RNN network, specifically includes the following steps:

[0020] Determine the input sequence X and the output sequence Y;

[0021] Initialize the RNN network parameters: Set the initial values ​​of the first weight W and the second weight V of each layer to 0; set the number of iterations p to 0; set the initial value of the output evaluation index ρ. set Set the maximum allowed number of iterations P. max ;

[0022] The activation function of the k-th neuron in the hidden layer of an RNN network is expressed as h. k =tanh(Wxh x k +W hh h k-1 +b h ), where h k Let x be the hidden state at the current time step, and tanh(x) be the activation function. W xh Let x be the first weight matrix. k W is the input for the current time step. hh h is the second weight matrix. k-1 b is the hidden state of the previous time step. h This is the bias term for the k-th neuron in the hidden layer;

[0023] The activation function of the k-th output unit of the output layer of an RNN network is expressed as O k =W hy h k +b y O k W represents the output of the k-th output unit. hy b is the third weight matrix. y This is the bias term for the k-th neuron in the output layer;

[0024] Perform a nonlinear transformation on the output of the last RNN layer. The final output of the relationship fitting model is obtained; where, The preliminary identification results of photovoltaic output power are given, where f() is the activation function and is selected as the softmax activation function.

[0025] Step S3, which involves using the data obtained in step S1 and the relationship fitting network constructed in step S2 to fit the nonlinear mapping function between light intensity and photovoltaic output power, specifically includes the following steps:

[0026] The light intensity sequence is used as the input sequence, and the photovoltaic output power sequence is used as the output sequence.

[0027] The nonlinear mapping function between light intensity and photovoltaic output power is obtained by using the relation fitting network constructed in step S2.

[0028] Step S4, based on the nonlinear mapping function obtained in step S3, uses the relationship fitting network constructed in step S2 to fit the nonlinear mapping function between meteorological data and photovoltaic output power, thereby obtaining the identification result of photovoltaic output power. Specifically, this includes the following steps:

[0029] The acquired meteorological data is represented as X′=(X1′,X2′,…,X N ′) T, where N is the sequence length; the data at each time point in the sequence includes i feature values, corresponding to the i meteorological factors at that time point;

[0030] For each meteorological factor data, the following formula was used for normalization:

[0031]

[0032] In the formula, x* represents the normalized data; x represents the data before normalization; xmin represents the minimum value of the data; and xmax represents the maximum value of the data.

[0033] The correlation coefficient ρ0 between meteorological factor data and photovoltaic output power is calculated using the following formula:

[0034]

[0035] In the formula x t ′ represents the t-th element in the meteorological factor data; y represents the average value of the corresponding meteorological factor data; t Let t be the t-th element in the photovoltaic output power sequence; is the average value of the photovoltaic output power sequence; n is the sequence length;

[0036] Based on the obtained correlation coefficient ρ0, meteorological factor data with correlation coefficients greater than the set value are selected and used as the final input sequence; light intensity data are used as the output sequence; the relationship fitting network constructed in step S2 is used for fitting to obtain the nonlinear mapping function between meteorological data and light intensity.

[0037] Finally, based on the nonlinear mapping function between light intensity and photovoltaic output power and the nonlinear mapping function between meteorological data and light intensity, the nonlinear mapping function between meteorological data and photovoltaic output power is obtained, thus obtaining the identification result of photovoltaic output power.

[0038] Step S5, which involves constructing an evaluation function and calculating the evaluation result based on the photovoltaic output power identification result obtained in step S4, specifically includes the following steps:

[0039] The photovoltaic output power identification result obtained in step S4 is added to the acquired outlet measurement power data to obtain the actual load power L of the distribution area.

[0040] The following formula is used as the evaluation function:

[0041]

[0042] In the formula, ρ is the evaluation function value; X′(t) is the meteorological data at time t; L(t) represents the average value of the meteorological data; L(t) represents the total active power load within the transformer area at time t. This represents the average value of the total active load within the transformer area.

[0043] Step S6, based on the evaluation result obtained in step S5, repeats steps S3 to S6 until a photovoltaic output power identification result that meets the set conditions is obtained. Specifically, it includes the following steps:

[0044] If the evaluation function value ρ is greater than or equal to the set threshold, the weight matrix is ​​modified, and the process returns to step S4 for the next iteration.

[0045] If the evaluation function value ρ is less than the set threshold, the loop ends and the identification result of the photovoltaic output power sequence is output.

[0046] The error between the photovoltaic output power identification result obtained in step S7 and the actual photovoltaic output power result is calculated, and steps S3 to S7 are repeated based on the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained. Specifically, the steps include the following:

[0047] Calculate the mean square error between the final photovoltaic output power identification result and the actual photovoltaic output power result;

[0048] The calculated mean square error is judged: if the mean square error is greater than or equal to the set value, then repeat steps S3 to S7 for the next round of iterative optimization; if the mean square error is less than the set value, then the current model is used as the final photovoltaic power identification model for the distribution network.

[0049] This invention also provides a system for implementing the photovoltaic power identification method for distribution networks, comprising a data acquisition module, a model building module, a first fitting module, a second fitting module, a first evaluation module, a repetition module, a second evaluation module, and an identification module; the data acquisition module, model building module, first fitting module, second fitting module, and first evaluation module are connected in series; the repetition module, second evaluation module, and identification module are also connected in series; simultaneously, the output of the repetition module is also connected to the second fitting module; the data acquisition module is used to acquire historical distribution area data information and historical meteorological data information of the target distribution area, and upload the data information to the model building module; the model building module is used to construct a relational fitting model based on an RNN network according to the received data information, and upload the data information to the first fitting module; the first fitting module is used to fit a nonlinear mapping function between light intensity and photovoltaic output power using the constructed relational fitting network according to the received data information and the acquired data information, and upload the data information to the second fitting module; the second fitting module is used to fit a nonlinear mapping function based on the constructed relational fitting network according to the received data information and the obtained nonlinear mapping function. The system firstly fits a nonlinear mapping function between meteorological data and photovoltaic (PV) output power to obtain the PV output power identification result, and uploads the data information to the first evaluation module. The first evaluation module constructs an evaluation function and calculates the evaluation result based on the received data and the obtained PV output power identification result, and uploads the data information to the repeating module. The repeating module repeats the work of the first fitting module, the second fitting module, the first evaluation module, and itself based on the received data and the obtained evaluation result until a PV output power identification result that meets the set conditions is obtained, and uploads the data information to the second evaluation module. The second evaluation module calculates the error between the final obtained PV output power identification result and the actual PV output power result based on the received data, and repeats the work of the first fitting module, the second fitting module, the first evaluation module, the repeating module, and itself based on the error until a distribution network PV power identification model that meets the set conditions is obtained, and uploads the data information to the identification module. The identification module performs actual distribution network PV power identification based on the received data and the obtained distribution network PV power identification model.

[0050] The photovoltaic power identification method and system for distribution networks provided by this invention considers the correlation between meteorological factors and photovoltaic output power, and between meteorological factors and load power, and performs power identification based on nonlinear functions identified by neural networks; therefore, this invention can not only identify photovoltaic power in distribution networks, but also has high reliability and good accuracy. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0052] Figure 2 This is a schematic diagram of the photovoltaic power identification result based on the light intensity identification method in an embodiment of the present invention.

[0053] Figure 3 This is a schematic diagram of the photovoltaic power identification result based on the meteorological factor identification method in an embodiment of the present invention.

[0054] Figure 4 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation

[0055] like Figure 1 The diagram shown is a flowchart of the method of the present invention: The photovoltaic power identification method for power distribution networks disclosed in this invention includes the following steps:

[0056] S1. Obtain historical station data and historical meteorological data for the target distribution area; specifically including the following steps:

[0057] Obtain historical data information of the target distribution transformer area, including the output power data, photovoltaic output power data, and load power data of the target distribution transformer area;

[0058] Obtain historical meteorological data information for the target distribution station area, including weather data, humidity data, air pressure data, precipitation probability data, and actual temperature data for the target distribution station area;

[0059] S2. Construct a relationship fitting model based on an RNN network; specifically including the following steps:

[0060] Determine the input sequence X and the output sequence Y;

[0061] Initialize the RNN network parameters: Set the initial values ​​of the first weight W and the second weight V of each layer to 0; set the number of iterations p to 0; set the initial value of the output evaluation index ρ. set Set the maximum allowed number of iterations P. max ;

[0062] The activation function of the k-th neuron in the hidden layer of an RNN network is expressed as h. k =tanh(W xh x k +W hh h k-1 +b h ), where h k Let x be the hidden state at the current time step, and tanh(x) be the activation function. W xh Let x be the first weight matrix. k W is the input for the current time step. hhh is the second weight matrix. k-1 b is the hidden state of the previous time step. h This is the bias term for the k-th neuron in the hidden layer;

[0063] The activation function of the k-th output unit of the output layer of an RNN network is expressed as O k =W hy h k +b y O k W represents the output of the k-th output unit. hy b is the third weight matrix. y This is the bias term for the k-th neuron in the output layer;

[0064] Perform a nonlinear transformation on the output of the last RNN layer. The final output of the relationship fitting model is obtained; where, The preliminary identification results of photovoltaic output power are given, where f() is the activation function and is selected as the softmax activation function;

[0065] S3. Based on the data obtained in step S1, use the relationship fitting network constructed in step S2 to fit the nonlinear mapping function between light intensity and photovoltaic output power; specifically, this includes the following steps:

[0066] The light intensity sequence is used as the input sequence, and the photovoltaic output power sequence is used as the output sequence.

[0067] Using the relationship fitting network constructed in step S2, a nonlinear mapping function between light intensity and photovoltaic output power is obtained;

[0068] S4. Based on the nonlinear mapping function obtained in step S3, the relationship fitting network constructed in step S2 is used to fit the nonlinear mapping function between meteorological data and photovoltaic output power, thereby obtaining the identification result of photovoltaic output power; specifically including the following steps:

[0069] The acquired meteorological data is represented as X′=(X1′,X2′,…,X N ′) T , where N is the sequence length; the data at each time point in the sequence includes i feature values, corresponding to the i meteorological factors at that time point;

[0070] For each meteorological factor data, the following formula was used for normalization:

[0071]

[0072] In the formula, x* represents the normalized data; x represents the data before normalization; xmin represents the minimum value of the data; and xmax represents the maximum value of the data.

[0073] The correlation coefficient ρ0 between meteorological factor data and photovoltaic output power is calculated using the following formula:

[0074]

[0075] In the formula x t ′ represents the t-th element in the meteorological factor data; y represents the average value of the corresponding meteorological factor data; t Let t be the t-th element in the photovoltaic output power sequence; is the average value of the photovoltaic output power sequence; n is the sequence length;

[0076] Based on the obtained correlation coefficient ρ0, meteorological factor data with correlation coefficients greater than the set value are selected and used as the final input sequence; light intensity data are used as the output sequence; the relationship fitting network constructed in step S2 is used for fitting to obtain the nonlinear mapping function between meteorological data and light intensity.

[0077] Finally, based on the nonlinear mapping function between light intensity and photovoltaic output power and the nonlinear mapping function between meteorological data and light intensity, the nonlinear mapping function between meteorological data and photovoltaic output power is obtained, thereby obtaining the identification result of photovoltaic output power;

[0078] S5. Based on the photovoltaic output power identification results obtained in step S4, construct an evaluation function and calculate the evaluation results; specifically, this includes the following steps:

[0079] The photovoltaic output power identification result obtained in step S4 is added to the acquired outlet measurement power data to obtain the actual load power L of the distribution area.

[0080] The following formula is used as the evaluation function:

[0081]

[0082] In the formula, ρ is the evaluation function value; X′(t) is the meteorological data at time t; L(t) represents the average value of the meteorological data; L(t) represents the total active power load within the transformer area at time t. This represents the average total active load within the transformer substation area.

[0083] S6. Based on the evaluation results obtained in step S5, repeat steps S3 to S6 until a photovoltaic output power identification result that meets the set conditions is obtained; specifically, the following steps are included:

[0084] If the evaluation function value ρ is greater than or equal to the set threshold, the weight matrix is ​​modified, and the process returns to step S4 for the next iteration.

[0085] If the evaluation function value ρ is less than the set threshold, the loop ends and the identification result of the photovoltaic output power sequence is output.

[0086] S7. Calculate the error between the final photovoltaic output power identification result and the actual photovoltaic output power result, and repeat steps S3 to S7 according to the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained; specifically including the following steps:

[0087] Calculate the mean square error between the final photovoltaic output power identification result and the actual photovoltaic output power result;

[0088] The calculated mean square error is judged: if the mean square error is greater than or equal to the set value, then repeat steps S3 to S7 for the next round of iterative optimization; if the mean square error is less than the set value, then the current model is used as the final photovoltaic power identification model for the distribution network.

[0089] S8. Based on the photovoltaic power identification model of the distribution network obtained in step S7, perform actual photovoltaic power identification of the distribution network.

[0090] The method of the present invention will be further described below with reference to an embodiment:

[0091] Taking a subsystem service station as an example, its photovoltaic power capacity is approximately 226.8kW. This service station is powered by a dedicated transformer and line. The load at the service station mainly consists of the resort's living load, lighting system, and fire protection load. The above load characteristics also indicate that the load power is not directly related to the light intensity, thus justifying the premise of the model in this invention that the correlation between the two is approximately zero. The photovoltaic array is equipped with dedicated light intensity and power generation measurement devices, collecting data every 5 minutes, using 24 hours of data. Since the transformer is equipped with a power measurement device with a time resolution of 1 minute, the measurement data needs to be preprocessed, i.e., its resolution is reduced to 5 minutes, and aligned with the 5-minute resolution light intensity measurement data in terms of time scale. The sum of the photovoltaic output power and the transformer power is the load power of the service station.

[0092] The light intensity sequence is used as the input data of the RNN, and the identified photovoltaic output power sequence is used as the output data of the RNN. The parameters of the RNN are initialized as follows: the initial values ​​of the weights of each layer W and V are set to 0, the initial learning rate η is set to 0.0001, the initial number of iterations p is set to 0, and the maximum allowed number of iterations P is set to 0. max =7390, the convergence criterion is ρ<0.0001, and the number of neurons is determined by an empirical formula:

[0093]

[0094] Where: Nhid N represents the number of neurons in the hidden layer of a neural network. in N represents the number of neurons in the input layer of the neural network. out This represents the number of neurons in the output layer of the neural network; a is a constant between 1 and 10.

[0095] This embodiment extracts 591,825 data points each of light intensity and photovoltaic power with a time resolution of 5 minutes from April 4, 2016 to November 18, 2021, forming a sequence. The first 500,000 data points are used as the training set, and the last 91,825 data points are used as the test set. A loop process is executed to complete the forward propagation process, and the loop result is output.

[0096] Combining the measured power data Z from the transformer substation's output in step 1, after the RNN outputs the preliminary identification result Y of the photovoltaic power generation, the result Y is added to the measured power Z to obtain the actual load power L = Y + Z for the transformer substation. The RNN iteration count p is incremented by 1 to determine if the RNN iteration count exceeds the limit. If it exceeds the maximum allowed iteration count, the loop ends and training exits; if it does not exceed the maximum allowed iteration count, the next step continues. Combining the actual load power L and the irradiance X of the transformer substation, the evaluation function of the RNN is constructed using the coefficient ρ of the linear correlation between the two. The evaluation index ρ is calculated. If ρ < 0.0001, the loop ends; if not, the weights are adjusted and the process returns to the next step for RNN retraining.

[0097] Now, assuming that there are insufficient light intensity measurement devices and light intensity data cannot be obtained, light intensity will be replaced by meteorological factors closely related to light intensity. Based on historical weather data, the following ten meteorological factors closely related to light intensity within the distribution area are selected: weather, humidity, air pressure, precipitation probability, actual temperature, perceived temperature, UV index, wind speed, wind direction, and wind force level.

[0098] Ten meteorological influencing factors were normalized using the average interpolation method. The Pearson correlation coefficient was used to calculate the linear correlation coefficient between the historical meteorological data and the historical photovoltaic output power data. Data with correlation coefficients greater than a threshold were retained, and a training set was constructed. In this embodiment, the threshold was set to 0.2, and the retained meteorological factors were weather, humidity, air pressure, precipitation probability, and actual temperature. Light intensity was replaced with the above five meteorological factors. Each time step in the sequence contained five feature values, representing the five meteorological factors at that moment. The dataset was equivalent to an N×5 matrix. Only the input data was changed from one-dimensional to five-dimensional, and the same method was used to train the neural network, enabling low-cost identification of the photovoltaic output power sequence from the power measurement sequence of the transformer substation.

[0099] Based on the training results, the mean squared error (RMSE) was used as the evaluation metric for network accuracy. A more accurate identification model was obtained through repeated training and saved. The results of identifying photovoltaic output power based on light intensity are as follows: Figure 2 As shown, the mean square error between the identified photovoltaic power and the actual photovoltaic power is 38.36 kWh; the results of identifying the photovoltaic output power based on meteorological factors are as follows. Figure 3 As shown, the mean square error between the identified photovoltaic power and the actual photovoltaic power is 41.77 kWh;

[0100] like Figure 2 The identified photovoltaic power, obtained by training a neural network based on historical irradiance data, was compared with the actual photovoltaic power generation sequence. The standard deviation (RMSE), a quantitative indicator reflecting the goodness of fit between the two, was 38.36, meaning that the average deviation between the identified value and the actual value was 38.36 kWh. This indicates that the identified photovoltaic power is relatively close to the actual photovoltaic power. Although the error did not reach the ideal value of near zero, in practical applications, this identification method can maximize the separation between the transformer load and photovoltaic power generation, and helps improve the accuracy of load forecasting.

[0101] like Figure 3 The identified photovoltaic (PV) power, obtained by training a neural network using historical weather data, was compared with the actual PV power generation. The RMSE (Real-Time Standard Deviation), a quantifiable indicator reflecting the goodness of fit between the two, was 41.77, meaning the average deviation between the identified and actual values ​​was 41.77 kWh. This error is significantly larger than that of the model based on illuminance intensity before the improvement. Therefore, the accuracy of PV output power identification based on weather factors is lower than that based on illuminance intensity. This is because illuminance intensity is a direct influencing factor of PV output, and relying on illuminance intensity for identification and prediction is the most direct and accurate method. However, since meteorological data can be obtained only from weather forecasts, weather-based PV output power identification eliminates the high cost of equipment for collecting illuminance data, making the PV output identification method more economical.

[0102] like Figure 4The diagram shows the functional modules of the system of the present invention: The system for implementing the photovoltaic power identification method of the distribution network disclosed in this invention includes a data acquisition module, a model building module, a first fitting module, a second fitting module, a first evaluation module, a repetition module, a second evaluation module, and an identification module; the data acquisition module, model building module, first fitting module, second fitting module, and first evaluation module are connected in series; the repetition module, second evaluation module, and identification module are connected in series; at the same time, the output of the repetition module is also connected to the second fitting module; the data acquisition module is used to acquire historical distribution area data information and historical meteorological data information of the target distribution area, and upload the data information to the model building module; the model building module is used to construct a relation fitting model based on an RNN network according to the received data information, and upload the data information to the first fitting module; the first fitting module is used to fit the nonlinear mapping function between light intensity and photovoltaic output power using the constructed relation fitting network according to the received data information and the acquired data information, and upload the data information to the second fitting module; the second fitting module is used to fit the nonlinear mapping function according to the received data information and the obtained nonlinear mapping function using a constructed... A relational fitting network is constructed to fit the nonlinear mapping function between meteorological data and photovoltaic output power, thereby obtaining the photovoltaic output power identification result and uploading the data information to the first evaluation module. The first evaluation module is used to construct an evaluation function and calculate the evaluation result based on the received data information and the obtained photovoltaic output power identification result, and upload the data information to the repeating module. The repeating module is used to repeat the work of the first fitting module, the second fitting module, the first evaluation module, and itself based on the received data information and the obtained evaluation result until a photovoltaic output power identification result that meets the set conditions is obtained, and upload the data information to the second evaluation module. The second evaluation module is used to calculate the error between the final obtained photovoltaic output power identification result and the actual photovoltaic output power result based on the received data information, and repeat the work of the first fitting module, the second fitting module, the first evaluation module, the repeating module, and itself based on the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained, and upload the data information to the identification module. The identification module is used to perform actual distribution network photovoltaic power identification based on the received data information and the obtained distribution network photovoltaic power identification model.

Claims

1. A method for identifying photovoltaic power in a distribution network, comprising the following steps: S1. Obtain historical station data and historical meteorological data for the target distribution area; specifically including the following steps: Obtain historical data information of the target distribution transformer area, including the output power data, photovoltaic output power data, and load power data of the target distribution transformer area; Obtain historical meteorological data information for the target distribution station area, including weather data, humidity data, air pressure data, precipitation probability data, and actual temperature data for the target distribution station area; S2. Construct a relationship fitting model based on an RNN network; specifically, this includes the following steps: Determine the input sequence X and the output sequence Y; Initialize the RNN network parameters: Set the initial values ​​of the first weight W and the second weight V of each layer to 0; set the number of iterations. Set the initial value of the output result evaluation index to 0. ; Setting the maximum number of allowed iterations ; The activation function of the kth neuron of the hidden layer of the RNN network is denoted as wherein, The hidden state at the current time step. For activation function and , This is the first weight matrix. For the current time step input, This is the second weight matrix. This is the hidden state from the previous time step. This is the bias term for the k-th neuron in the hidden layer; The activation function of the kth output unit of the output layer of the RNN network is represented as wherein is the output result of the kth output unit, is a third weight matrix, is a bias term of the kth neuron of the output layer; performing a nonlinear transformation on the output of the last RNN network , to obtain an output result of the final relationship fitting model; wherein, is a preliminary identification result of photovoltaic output power, is an activation function and is selected as a softmax activation function; S3. Based on the data obtained in step S1, use the relationship fitting network constructed in step S2 to fit the nonlinear mapping function between light intensity and photovoltaic output power; specifically, this includes the following steps: The light intensity sequence is used as the input sequence, and the photovoltaic output power sequence is used as the output sequence. Using the relationship fitting network constructed in step S2, a nonlinear mapping function between light intensity and photovoltaic output power is obtained; S4. Based on the nonlinear mapping function obtained in step S3, the relationship fitting network constructed in step S2 is used to fit the nonlinear mapping function between meteorological data and photovoltaic output power, thereby obtaining the identification result of photovoltaic output power; specifically including the following steps: The acquired weather data is represented as where N is the sequence length; each time point in the sequence includes i feature values, corresponding to i kinds of weather factor data of the time point. For each meteorological factor data, the following formula was used for normalization: In the formula is the normalized data; is the data before normalization; is the minimum value of the data; is the maximum value of the data; The correlation coefficient between the meteorological factor data and the photovoltaic output power is calculated using the following formula : In the formula Let t be the t-th element in the meteorological factor data; This represents the average value of the corresponding meteorological factor data; Let t be the t-th element in the photovoltaic output power sequence; is the average value of the photovoltaic output power sequence; n is the sequence length; According to the correlation coefficient obtained , select the meteorological factor data with the correlation coefficient greater than a set value as the final input sequence; take the light intensity data as the output sequence; and use the relationship fitting network constructed in step S2 to perform fitting to obtain a nonlinear mapping function between the meteorological data and the light intensity. Finally, based on the nonlinear mapping function between light intensity and photovoltaic output power and the nonlinear mapping function between meteorological data and light intensity, the nonlinear mapping function between meteorological data and photovoltaic output power is obtained, thereby obtaining the identification result of photovoltaic output power; S5. Based on the photovoltaic output power identification results obtained in step S4, construct an evaluation function and calculate the evaluation results; S6. Based on the evaluation results obtained in step S5, repeat steps S3 to S6 until the photovoltaic output power identification result that meets the set conditions is obtained; S7. Calculate the error between the final photovoltaic output power identification result and the actual photovoltaic output power result, and repeat steps S3 to S7 according to the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained; S8. Based on the photovoltaic power identification model of the distribution network obtained in step S7, perform actual photovoltaic power identification of the distribution network.

2. The power distribution network photovoltaic power identification method of claim 1, wherein Step S5, which involves constructing an evaluation function and calculating the evaluation result based on the photovoltaic output power identification result obtained in step S4, specifically includes the following steps: The photovoltaic output power identification result obtained in step S4 is added to the acquired outlet measurement power data to obtain the actual load power L of the distribution area. The following formula is used as the evaluation function: In the formula To evaluate the function value; The meteorological data is at time t; This represents the average value of meteorological data. Let t be the total active load within the transformer area at time t; This represents the average total active load within the transformer area.

3. The power distribution network photovoltaic power identification method of claim 2, wherein Step S6, based on the evaluation result obtained in step S5, repeats steps S3 to S6 until a photovoltaic output power identification result that meets the set conditions is obtained. Specifically, it includes the following steps: If the evaluation function value is greater than or equal to a set threshold value, the weight matrix is modified and the next iteration is returned to step S4. If the evaluation function value is less than a set threshold value, the loop ends and the recognition result of the photovoltaic output power sequence is output.

4. The power distribution network photovoltaic power identification method of claim 3, wherein The error between the photovoltaic output power identification result obtained in step S7 and the actual photovoltaic output power result is calculated, and steps S3 to S7 are repeated based on the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained. Specifically, the steps include the following: Calculate the mean square error between the final photovoltaic output power identification result and the actual photovoltaic output power result; The calculated mean square error is judged: if the mean square error is greater than or equal to the set value, then repeat steps S3 to S7 for the next round of iterative optimization; if the mean square error is less than the set value, then the current model is used as the final photovoltaic power identification model for the distribution network.

5. A system for implementing the power distribution network photovoltaic power identification method of any one of claims 1-4, characterized in that The system includes a data acquisition module, a model building module, a first fitting module, a second fitting module, a first evaluation module, a repetition module, a second evaluation module, and an identification module. These modules are connected in series. Similarly, the repetition module, the second evaluation module, and the identification module are also connected in series. The output of the repetition module is connected to the second fitting module. The data acquisition module acquires historical station area data and historical meteorological data for the target distribution area and uploads this data to the model building module. The model building module constructs a relationship fitting model based on an RNN network using the received data and uploads the data to the first fitting module. The first fitting module uses the constructed relationship fitting network to fit a nonlinear mapping function between light intensity and photovoltaic output power based on the received and acquired data and uploads the data to the second fitting module. The second fitting module is used to fit the nonlinear mapping function between meteorological data and photovoltaic output power using the received data information and the obtained nonlinear mapping function, and then upload the data information to the first evaluation module. The first evaluation module is used to construct an evaluation function and calculate the evaluation result based on the received data information and the identification result of the photovoltaic output power, and then upload the data information to the repeat module. The repeat module is used to repeat the work of the first fitting module, the second fitting module, the first evaluation module and itself based on the received data information and the obtained evaluation results until the photovoltaic output power identification result that meets the set conditions is obtained, and then upload the data information to the second evaluation module. The second evaluation module is used to calculate the error between the final photovoltaic output power identification result and the actual photovoltaic output power result based on the received data information, and repeat the work of the first fitting module, the second fitting module, the first evaluation module, the repeating module and itself according to the error until a distribution network photovoltaic power identification model that meets the set conditions is obtained, and then upload the data information to the identification module. The identification module is used to identify the actual photovoltaic power of the distribution network based on the received data and the obtained distribution network photovoltaic power identification model.

Images