Transient stability index prediction method and device for novel power system under high-proportion access of new energy

Abstract

The invention discloses a transient stability index prediction method and device for a novel power system under high-proportion access of new energy. The method comprises the steps of obtaining power grid continuous new energy short-circuit ratio data of a power system; preprocessing the power grid continuous new energy short-circuit ratio data to obtain a binary discrete pulse sequence; according to the historical binary discrete pulse sequence, training a pre-constructed brain-like neuromorphic network based on pulse driving, and generating a new energy short-circuit ratio data prediction model; outputting a binary new energy short-circuit ratio prediction result after a second preset time period by using a new energy short-circuit ratio data prediction model according to the historical binary discrete pulse sequence and the real-time binary discrete pulse sequence in the first preset time period; and converting the binary new energy short-circuit ratio prediction result into a continuous signal through a retainer, and obtaining a new energy short-circuit ratio prediction result after a second preset time period.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy index prediction, and more specifically, to a method and device for predicting transient stability index of a new power system under high-proportion access of new energy. Background Art

[0002] At present, my country is in a critical period of energy transformation. In order to achieve the goal of "carbon peak and carbon neutrality", the future will usher in a peak of large-scale construction of new energy power generation. In 2030 and 2060, the scale of new energy installed capacity will reach 1.2 billion kilowatts and more than 4 billion kilowatts respectively. The randomness, volatility and low anti-interference of new energy power generation pose great challenges to the safe and stable operation of the power system. As the proportion of installed capacity gradually increases, the impact of new energy power generation characteristics on the operation of the power grid has become increasingly prominent. The operation mechanism of the high-proportion new energy power system is more complicated. Therefore, the transient stability analysis of the power system under the large-scale access of new energy has become a key issue in the operation and dispatch of the power grid. The new energy short-circuit ratio refers to the ratio between the system short-circuit capacity and the capacity of electrical equipment (including power electronic equipment). At present, the new energy short-circuit ratio has been used to analyze the strength of the power system at the interconnection point of renewable resources. It is a key indicator for evaluating the transient safety and stability of new power systems.

[0003] The short-circuit ratio of new energy is a time series with dynamic characteristics, just like the online operation data of the power system. The output of the system is not only related to the input at the current moment, but also to the past state. The traditional deep neural network is a network structure inspired by the brain, but it is fundamentally different from the real brain in terms of learning rules and calculation methods. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and device for predicting transient stability indicators of a new power system under high-proportion access of new energy sources.

[0005] According to one aspect of the present invention, a method for predicting transient stability indicators of a new power system under high-proportion access of new energy is provided, comprising:

[0006] Obtaining continuous short-circuit ratio data of renewable energy sources in the power grid of the power system, the continuous short-circuit ratio data of renewable energy sources in the power grid includes: historical short-circuit ratio data and real-time short-circuit ratio data;

[0007] Preprocess the continuous renewable energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence, which includes a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence;

[0008] A pre-built pulse-driven brain-like neuromorphic network is trained according to a historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, wherein the brain-like neuromorphic network includes a hippocampal neural network and a prefrontal cortex neural network, and both the hippocampal neural network and the prefrontal cortex neural network include multiple leaky integration triggering neurons and multiple artificial leaky integration triggering neurons;

[0009] Using the new energy short-circuit ratio data prediction model, according to the historical binary discrete pulse sequence and the real-time binary discrete pulse sequence within the first preset time period, output the binary new energy short-circuit ratio prediction result after the second preset time period;

[0010] The binary new energy short circuit ratio prediction result is converted into a continuous signal through a retainer to obtain the new energy short circuit ratio prediction result after a second preset time period.

[0011] Optionally, the continuous renewable energy short-circuit ratio data of the power grid is preprocessed to obtain a binary discrete pulse sequence, including:

[0012] According to the preset time and the adoption interval, the continuous new energy short-circuit ratio data of the power grid is discretized to obtain the discretized new energy short-circuit ratio data;

[0013] The discretized new energy short-circuit ratio data is processed by a Poisson sequence generator to obtain a binary discrete pulse sequence that is compatible with the event-driven characteristics of the neuromorphic.

[0014] Optionally, the discretized new energy short-circuit ratio data is processed by a Poisson sequence generator to obtain a binary discrete pulse sequence adapted to the neuromorphic event-driven characteristics, including:

[0015] Apply sliding window to process discrete new energy short-circuit ratio data and obtain multiple processing window data matrices;

[0016] A Poisson pulse sequence generator is used to generate a random number that obeys Poisson distribution corresponding to each processing window data matrix according to multiple processing window data matrices;

[0017] Compare the value of the element in each processing window data matrix with the corresponding random number, update the processing window data matrix according to the comparison result, and obtain multiple updated processing window data matrices;

[0018] The windowed data matrix is ​​processed according to a plurality of updates to obtain a binary discrete pulse train.

[0019] Optionally, updating the processing window data matrix according to the comparison result to obtain multiple updated processing window data matrices includes:

[0020] In the comparison result, when the value of an element in the processing window data matrix is ​​greater than the value of the corresponding random number, the value of the element in the processing window data matrix is ​​set to 1;

[0021] When the value of an element in the processing window data matrix is ​​smaller than the value of the corresponding random number, the value of the element in the processing window data matrix is ​​set to 0;

[0022] A plurality of processing window data matrices are updated to obtain a plurality of updated processing window data matrices.

[0023] Optionally, the input of the hippocampal network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the state information of the prefrontal cortex network; the input signal of the prefrontal cortex network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the learning signal transmitted by the hippocampal network; both the hippocampal network and the prefrontal cortex network use leaky integration triggering neurons and artificial leaky integration triggering neurons models to construct a multi-layer brain-like neural morphological network based on the synaptic time-dependent plasticity rule, in which

[0024] The leaky integration triggering neuron model is as follows:

[0025]

[0026] Where V j (t) represents the membrane potential of neuron j at time t; is the synaptic weight between presynaptic neuron i and neuron j; The synaptic weights for input information into the network; The effect of membrane time constant on membrane potential decay is characterized by α, τ m is the membrane time constant, which controls the decay of membrane potential; is the pulse output generated by the neuron when the membrane potential exceeds the threshold; d represents the delay of the presynaptic pulse train transmission;

[0027] The artificial leaky integration triggering neuron model is as follows:

[0028] a j (t) = ρa j (t-1)+S j (t-1)

[0029] A j (t) = βa j (t)+A j (t)

[0030]

[0031] Where V j (t) is the membrane potential at time t; V this the adaptive threshold; The effect of the membrane time constant on the adaptive threshold decay is characterized by ρ, τ a is the membrane time constant; a j (t-1) is the update amount generated by the adaptive threshold of the neuron when the neuron generates a pulse output.

[0032] Optionally, a pre-built pulse-driven brain-like neuromorphic network is trained according to the historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, including:

[0033] The historical binary output discrete pulse trains are input into the hippocampal neural network and the prefrontal cortex neural network;

[0034] The features of the historical binary output discrete pulse trains are extracted through the prefrontal cortex neural network and input into the hippocampal neural network;

[0035] The hippocampal neural network generates learning signals for the synaptic weights of different neurons in the prefrontal cortex neural network according to the historical binary output discrete pulse sequences and features, and feeds the learning signals back to the prefrontal cortex neural network.

[0036] The synaptic weights of different neurons are adjusted according to the learning signal through the prefrontal cortex neural network, and the training prediction result of the binary new energy short-circuit ratio at a specific time point after the second set time period is output according to the adjusted synaptic weights of different neurons;

[0037] Processing the binary new energy short circuit ratio indicator training prediction results through the retainer, converting them into continuous prediction results, and outputting continuous new energy short circuit ratio training prediction results;

[0038] A new energy short circuit ratio data prediction model is generated based on the new energy short circuit ratio training prediction results and the corresponding new energy short circuit ratio training actual results.

[0039] According to another aspect of the present invention, a transient stability index prediction device for a new power system with a high proportion of new energy access is provided, comprising:

[0040] The first acquisition module is used to acquire the continuous short-circuit ratio data of renewable energy in the power grid of the power system, the continuous short-circuit ratio data of renewable energy in the power grid includes: historical short-circuit ratio data and real-time short-circuit ratio data;

[0041] The second acquisition module is used to pre-process the continuous renewable energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence, where the binary discrete pulse sequence includes: a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence;

[0042] A generation module is used to train a pre-built pulse-driven brain-like neuromorphic network according to a historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, wherein the brain-like neuromorphic network includes a hippocampal neural network and a prefrontal cortex neural network, and both the hippocampal neural network and the prefrontal cortex neural network include multiple leakage integration trigger neurons and multiple artificial leakage integration trigger neurons;

[0043] An output module, used to output a prediction result of the binary new energy short circuit ratio after a second preset time period according to a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence within a first preset time period by using a new energy short circuit ratio data prediction model;

[0044] The third acquisition module is used to convert the binary new energy short-circuit ratio prediction result into a continuous signal through a retainer, and obtain the new energy short-circuit ratio prediction result after a second preset time period.

[0045] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and the computer program is used to execute the method described in any one of the above aspects of the present invention.

[0046] According to another aspect of the present invention, an electronic device is provided, comprising: a processor; a memory for storing instructions executable by the processor; the processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of the above aspects of the present invention.

[0047] Thus, the present invention adopts a leakage integration trigger model and an artificial leakage integration trigger model, which is biologically interpretable. The superposition of the artificial leakage integration trigger model enables the adaptive characteristics of the biological neuron trigger threshold to play an adjustment role in the network. A dual network structure simulating the human brain hippocampus and prefrontal cortex is established, so that the learning process of the prediction system is not only affected by the synaptic plasticity rules, but also regulated by the learning signal of the hippocampal network. This is similar to the dopamine signal function of the human brain nervous system during learning and cognitive activities. A learning method for the hippocampal network and the prefrontal cortex network is established, and the continuous learning ability of the network under a small number of calibration samples is achieved through the collaboration of the two networks. This enables the network to obtain rapid and accurate historical data in a new power system with a limited time window, and through a powerful online learning ability, a more accurate and robust new energy short-circuit ratio prediction result is given. This training method greatly reduces the network's dependence on label data and improves the generalization ability and robustness of the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0049] Figure 1 It is a flow chart of a method for predicting transient stability index of a new power system under high proportion access of new energy provided by an exemplary embodiment of the present invention;

[0050] Figure 2 It is a dual network neuromorphic structure diagram including a hippocampal network structure and a prefrontal cortex network structure provided by an exemplary embodiment of the present invention;

[0051] Figure 3 is a schematic diagram of a synaptic time-dependent plasticity rule provided by an exemplary embodiment of the present invention;

[0052] Figure 4 It is a schematic diagram of determining hyperparameters and initialization weights for secondary optimization provided by an exemplary embodiment of the present invention;

[0053] Figure 5 is a schematic diagram of a learning algorithm of a recursive neuromorphic network under a specific task provided by an exemplary embodiment of the present invention;

[0054] Figure 6 It is a schematic diagram of neurons realization of a hippocampal network and a prefrontal cortex network under a network architecture provided by an exemplary embodiment of the present invention;

[0055] Figure 7 It is a structural schematic diagram of a transient stability index prediction device for a new power system under high-proportion access of new energy provided by an exemplary embodiment of the present invention;

[0056] Figure 8 This is a structure of an electronic device provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0057] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0058] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0059] Those skilled in the art can understand that the terms "first" and "second" in the embodiments of the present invention are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0060] It should also be understood that, in the embodiments of the present invention, “plurality” may refer to two or more than two, and “at least one” may refer to one, two or more than two.

[0061] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0062] In addition, the term "and / or" in the present invention is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects before and after are in an "or" relationship.

[0063] It should also be understood that the description of the various embodiments of the present invention focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced to each other, and for the sake of brevity, they will not be described one by one.

[0064] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0065] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0066] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0067] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0068] Embodiments of the present invention can be applied to electronic devices such as terminal devices, computer systems, servers, etc., which can operate with many other general or special computing system environments or configurations. Examples of well-known terminal devices, computing systems, environments and / or configurations suitable for use with electronic devices such as terminal devices, computer systems, servers, etc. include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems, and distributed cloud computing technology environments including any of the above systems, etc.

[0069] Electronic devices such as terminal devices, computer systems, servers, etc. can be described in the general context of computer system executable instructions (such as program modules) executed by computer systems. Generally, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0070] Exemplary Methods

[0071] Figure 1 1 is a flow chart of a method for predicting transient stability indicators of a new power system under high-proportion access of new energy sources provided by an exemplary embodiment of the present invention. This embodiment can be applied to electronic devices, such as Figure 1 As shown, the transient stability index prediction method 100 of the new power system under high proportion access of new energy sources includes the following steps:

[0072] Step 101, obtaining continuous short-circuit ratio data of renewable energy sources in the power grid of the power system, the continuous short-circuit ratio data of renewable energy sources in the power grid including: historical short-circuit ratio data and real-time short-circuit ratio data;

[0073] Step 102, preprocessing the continuous renewable energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence, the binary discrete pulse sequence including: a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence;

[0074] Step 103, training a pre-built pulse-driven brain-like neuromorphic network according to a historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, wherein the brain-like neuromorphic network includes a hippocampal neural network and a prefrontal cortex neural network, and both the hippocampal neural network and the prefrontal cortex neural network include a plurality of leakage integration triggering neurons and a plurality of artificial leakage integration triggering neurons;

[0075] Step 104, using the new energy short circuit ratio data prediction model, according to the historical binary discrete pulse sequence and the real-time binary discrete pulse sequence within the first preset time period, outputting the binary new energy short circuit ratio prediction result after the second preset time period;

[0076] Step 105: Convert the binary new energy short circuit ratio prediction result into a continuous signal through a retainer to obtain the new energy short circuit ratio prediction result after a second preset time period.

[0077] Specifically, the present invention provides a method for predicting transient stability indicators of a new power system under high-proportion access of new energy sources, the method comprising:

[0078] Step S1: Acquire new energy data of the power system; the new energy data refers to the continuous new energy short-circuit ratio data of the power grid, including historical data and online data.

[0079] Specifically, according to the circuit model of the new power system, the Thevenin equivalent method is used to simplify the AC system connected to the new energy source into an ideal voltage source in series with an equivalent impedance. According to the definition of the new energy short-circuit ratio index, the new energy short-circuit ratio index data at the new energy grid-connected bus is calculated, which is expressed as a continuous analog value.

[0080] Step S2: pre-processing the historical and online new energy short-circuit ratio data to obtain a binary discrete pulse sequence of the historical and online new energy short-circuit ratio data.

[0081] S2 specifically includes:

[0082] Step S21: discretize the historical and historical and online new energy short-circuit ratio data to obtain first discrete new energy short-circuit ratio data.

[0083] S21 specifically includes:

[0084] Step S211: Select a time step and a sampling interval, sample the continuous simulated new energy short-circuit ratio data, and obtain first discrete new energy short-circuit ratio data.

[0085] Step S22: Processing the first discrete new energy short-circuit ratio data through a Poisson sequence generator to obtain second discrete new energy short-circuit ratio data that is compatible with the neuromorphic event-driven characteristics.

[0086] S22 specifically includes:

[0087] Step S221: applying a sliding window according to the first discrete new energy short-circuit ratio data to obtain a plurality of processing window data matrices.

[0088] Step S222: Apply a Poisson pulse sequence generator to generate corresponding random numbers that obey the Poisson distribution for each processing window data matrix.

[0089] Step S223: When the value of the element in the processing window data matrix is ​​greater than the value of the corresponding random number, the value of the element in the processing window data matrix is ​​set to 1, and when the value of the element in the processing window data matrix is ​​less than the value of the corresponding random number, the value of the element in the processing window data matrix is ​​set to 0, multiple processing window data matrices are updated to obtain multiple updated processing window data matrices.

[0090] Step S224: Obtain a binary discrete pulse sequence of historical and online new energy short-circuit ratio data according to a plurality of updated processing window data matrices.

[0091] In practical applications, the Poisson pulse sequence generator is used to block and reprocess the first discrete new energy short-circuit ratio data, that is, to design the block length, the time step of each block and the window size setting. Here, corrections can be made according to the specific power system situation. If the power grid indicators fluctuate smoothly, it is necessary to involve the sampling number of longer historical data, so a larger sliding window is set; if the power grid indicators fluctuate violently, it is necessary to reduce the sampling number of historical data, so a smaller sliding window is set. Assuming that the sliding window size is 5×5 and the step size is 2, for each processing window, a random number that obeys the Poisson distribution is first randomly generated, and it is compared with the input information. If the input new energy short-circuit ratio data is greater than this random number at this point, a pulse output is generated, and the new energy short-circuit ratio value at the corresponding position of the window is set to 1, otherwise it is 0. After processing the input information matrix, a binary output discrete pulse sequence is obtained, which is the input signal of the prefrontal cortex network and the hippocampal network.

[0092] Step S3: construct a brain-like neural morphology network; the brain-like neural morphology network includes a hippocampal neural network and a prefrontal cortex neural network; the hippocampal neural network and the prefrontal cortex neural network both include multiple LIF neurons and multiple ALIF neurons.

[0093] Among them, the hippocampal network is used for experience storage. It is used in the human brain to store prior knowledge learned from history and provide guidance in new cognitive learning tasks. In the dual network model, the hippocampal network will serve as an auxiliary learning network to send guidance learning signals to the main network. The prefrontal cortex network is the main network mentioned above, which is used to extract the characteristics of the input new energy short-circuit ratio data and receive the learning signals transmitted by the auxiliary network; at the same time, as the main network of this cognitive activity, the prefrontal cortex network also needs to transmit its own state information to the auxiliary network; the internal information exchange is based on the guidance function of the dopamine learning signal, but it does not affect the main network to generate a new energy short-circuit ratio prediction output after a specific time period after the input of a new energy short-circuit ratio indicator.

[0094] Specifically, the input of the hippocampal network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the state information of the prefrontal cortex network. The learning signal generated by the hippocampal network will be transmitted to the prefrontal cortex network affected by the synaptic plasticity mechanism to adjust the discharge state of the prefrontal cortex network. The output of the hippocampal network is only one, that is, the learning signal transmitted to the prefrontal cortex network, which guides the adjustment of synaptic weights for specific neurons in the prefrontal cortex.

[0095] The input signal of the prefrontal cortex network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the learning signal transmitted by the hippocampal network. The output signal of the prefrontal cortex network is one, that is, the output pulse sequence, which is transmitted to the network output end and the hippocampal network at the same time, completing the results output and state feedback respectively. Both the hippocampal network and the prefrontal cortex network use LIF neuron and ALIF neuron models to construct a multi-layer brain-like neural morphological network based on the STDP rule.

[0096] Furthermore, LIF neurons characterize their states through membrane potential at a specific moment, which is affected not only by the input of presynaptic neurons but also by the stimulation of input information. Assuming V j (t) represents the membrane potential of neuron j at time t, and the synaptic weight between presynaptic neuron i and neuron j is The synaptic weight of the input information into the network is Membrane time constant τ m Controls the decay of membrane potential. We set α is used to characterize the effect of the membrane time constant on the membrane potential decay. Whenever the membrane potential exceeds the threshold, the neuron generates a pulse output, which is expressed as Considering the delay of presynaptic pulse train transmission, the time constant d is set to represent the delay. The LIF neuron model is shown in formula (1).

[0097]

[0098] Unlike LIF, ALIF neurons have two hidden variables, namely, its membrane potential V at time t and j (t) and the adaptive threshold V th is a state representation. The adaptive threshold is affected by the discharge of neurons. Specifically, the more excited the neuron is, the higher its discharge rate is, and the discharge threshold increases accordingly. Similar to the membrane potential update, the membrane time constant τ a It also controls the attenuation of the adaptive threshold. We set The effect of the membrane time constant on the adaptive threshold attenuation is characterized by ρ. Whenever a neuron generates a pulse output, the adaptive threshold of the neuron generates an update amount a j (t-1), considering the influence of membrane time constant, the update amount is adjusted and added to the default threshold voltage to complete the adaptive threshold learning. The ALIF neuron model is shown in formula (2).

[0099]

[0100] According to the jump of the new energy short-circuit ratio index of the new power system, the percentage setting of neuron categories is changed. If the jump of the new energy short-circuit ratio index is active, the system needs to maintain a strong adaptive ability to ensure the robustness of the output prediction results, so a larger percentage of neuron categories is set. Select a suitable network model to build the model of the hippocampal neural network and the prefrontal cortex network; build a dual network model with learning signals and realize its online learning channel.

[0101] Step S4: Input the binary output discrete pulse sequence within the first preset time period of the new energy short-circuit ratio index (the first preset time period can be half a month, a month, etc., which is not specifically limited here and is set according to the situation) into the trained brain-like neuromorphic network, and output the prediction result of the new energy short-circuit ratio index at a specific time point after the set time period; the prediction result is a discrete signal.

[0102] S4 specifically includes:

[0103] Step S41: Inputting the binary output discrete pulse sequence into the hippocampal neural network and the prefrontal cortex neural network.

[0104] Step S42: The prefrontal cortex neural network extracts features of the binary output discrete pulse sequence and inputs the features into the hippocampal neural network.

[0105] Step S43: The hippocampal neural network generates a learning signal for the synaptic weights of different neurons of the prefrontal cortex neural network according to the binary output discrete pulse sequence and the feature, and feeds the learning signal back to the prefrontal cortex neural network.

[0106] Step S44: The prefrontal cortex neural network adjusts the synaptic weights of different neurons according to the learning signal, and outputs the prediction result of the new energy short-circuit ratio index at a specific time point after the set time period according to the adjusted synaptic weights of different neurons.

[0107] Step S5: The prediction result of the new energy short-circuit ratio index at a specific time point after the set time period is processed by the retainer and converted into a continuous prediction result, thereby ensuring that the output predicted new energy short-circuit ratio index is not zero at any time point.

[0108] A Poisson encoder is used to extract features from the output binary discrete pulse sequence, and neuron membrane potential and threshold variables are used to characterize the features. The neuron synaptic weights of the dual network model are updated through the learning signal of the hippocampal neural network and the back propagation mechanism of the prefrontal cortex neural network based on the synaptic plasticity rule, and the new energy short circuit ratio prediction result at a specific time point after the output time period is generated.

[0109] The prediction result of the short circuit ratio of new energy at a specific time point after the set time period is simulated, and the simulation processing includes: a first-order retainer, a zero-order retainer, etc. The zero-order retainer is to maintain the prediction result of the short circuit ratio index of new energy at the specific time point, that is, within a specific time interval, the prediction value of the short circuit ratio of new energy is set to a constant value before the moment. The first-order retainer refers to connecting the prediction result of the short circuit ratio index of new energy at the specific time point through multiple first-order equations, that is, the prediction value of the short circuit ratio of new energy within a specific time interval is an increasing value.

[0110] Specifically, Figure 2 As shown, a hippocampal network and a prefrontal cortex network simulating the hippocampus and prefrontal cortex of the human brain are established through LIF neurons and ALIF neurons. The hippocampal network serves as an auxiliary network, and the prefrontal cortex network serves as the main network for cognitive learning interaction.

[0111] The prefrontal cortex network is composed of LIF neurons and ALIF neurons, which are distributed in different layers and are fully connected by synapses. The input signal is connected to the input layer neurons of the prefrontal cortex network through synapses, and the hippocampal network will also adjust the synaptic information of specific neurons in the prefrontal cortex network through specific synapses. The hippocampal network is also composed of LIF and ALIF neurons, and is still fully connected in the form of layers. As an integral auxiliary network, the hippocampal network undertakes the task of inner-loop self-learning. It is affected by the input signal and the state signal of the prefrontal cortex network, and specifically generates learning signals for the synaptic weights of different neurons in the hidden layer of the prefrontal cortex network, and transmits them to the neurons of the prefrontal cortex network. Among them, the learning signal is inspired by the synaptic plasticity mechanism, such as Figure 3 As shown. The connection strength between a neuron and the presynaptic neuron will be affected by the activity of the neuron itself. Assume that the presynaptic neuron is i, the postsynaptic neuron is j, and s represents the neuron sequence. refers to the f-th spike train generated by neuron i, It can represent the time difference between the time when neuron i sends a pulse and the time when neuron j receives the pulse for the fth pulse, t i and t j They represent the time when the presynaptic neuron releases the pulse f and the postsynaptic neuron generates the corresponding pulse f. The parameter τ represents the time interval before and after the synapse. The significance of A is to ensure that when the time when the presynaptic and postsynaptic neurons generate pulses is too close and the time difference is close to 0, the synaptic weight changes to the maximum extent, that is, A + and A - Represents the maximum value of the synaptic weight change, and formula (3) and formula (4) are valid:

[0112]

[0113] The prefrontal cortex network adjusts the synaptic strength of specific neurons according to the learning signals transmitted by the hippocampal network. Here, the synaptic plasticity rule is applied, which is the biological basis of learning and memory activities at the cellular level. As shown in formula (5) and formula (6), represents the membrane potential information of neuron i at time t, V th represents the threshold voltage, which is adaptive in the ALIF model. Represents the pulse train transmitted by the presynaptic neuron.

[0114]

[0115] If V>V th ,v←V reset

[0116]

[0117] Through iterative learning, the prefrontal cortex network and hippocampal network adjust local synaptic plasticity and store memory of learning experience, complete self-learning of the prefrontal cortex network and hippocampal network, and realize the learning of the entire brain-like pulse recursive neuromorphic network.

[0118] Figure 4 Schematic diagram of determining the hyperparameters and initialization weights for secondary optimization in an embodiment of the present invention, as shown in FIG. Figure 4 As shown, the steps for determining the hyperparameters and initialization weights for secondary optimization are as follows:

[0119] In the first step, after the network is trained with a small amount of new energy short-circuit ratio index calibration data set, the new energy short-circuit ratio historical data of a specific time period is selected to conduct a secondary evaluation of the network performance. It is converted into a binary discrete pulse sequence and transmitted to the hippocampal network and the prefrontal cortex network.

[0120] In the second step, through the self-learning process of the hippocampal network and the prefrontal cortex network, the network synaptic weights, the influencing factors of the membrane potential time constant, the influencing factors of the qualification trace time constant of the LIF and ALIF neurons, etc. are adjusted.

[0121] The third step is to integrate the adjusted dual network model and use it as the network parameter for future data prediction. At the same time, the hyperparameters are also updated twice.

[0122] Figure 5 Schematic diagram of the learning algorithm of the dual network model under any new energy short-circuit ratio prediction task in the embodiment of the present invention. Figure 5As shown, according to the specific learning task ID, that is, the future prediction time and the historical data reading period, the prediction input signal is read and transmitted to the hippocampal network and the prefrontal cortex network. The incoming information is discretized to obtain the first discrete new energy short-circuit ratio data. Through the Poisson pulse sequence generator, the first discrete new energy short-circuit ratio data is converted into the second discrete new energy short-circuit ratio data, that is, the binary pulse output signal. Through the collaborative working mechanism of the hippocampal network and the prefrontal cortex network, the new energy short-circuit ratio prediction result in the future prediction time period corresponding to the task ID is output.

[0123] Figure 6 This is a schematic diagram of the neuron implementation of the hippocampal network and the prefrontal cortex network under the network architecture, such as Figure 6 As shown in the figure, the network architecture of the hippocampal network and the prefrontal cortex network includes a pulse encoder, an input feature extractor, and a feature classifier, and the basic structure is such as VGG11, VGG19, ResNet, etc. According to the percentage of initialized neuron categories, including setting the proportion of LIF neurons and ALIF neurons, they are randomly distributed in the activation layers of different hidden layers.

[0124] The beneficial effects of the present invention are as follows:

[0125] 1. The present invention adopts the LIF model and the ALIF model, and sets up the network according to the percentage of neuron categories, which is more biologically interpretable. The superposition of the ALIF model enables the adaptive characteristics of the triggering threshold of biological neurons to play an adjustment role in the network.

[0126] 2. The present invention establishes a dual network structure that simulates the human brain hippocampus and prefrontal cortex, so that the learning process of the prediction system is not only affected by the synaptic plasticity rules, but also regulated by the learning signal of the hippocampal network. This is similar to the dopamine signal function of the human brain nervous system during learning and cognitive activities.

[0127] 3. The present invention establishes a learning method for the hippocampal network and the prefrontal cortex network, and realizes the continuous learning ability of the network under a small number of calibration samples through the collaboration of the two networks. This enables the network to obtain rapid and accurate historical data in a new power system with a limited time window, and through powerful online learning capabilities, it can provide more accurate and robust prediction results of the short-circuit ratio of new energy. This training method greatly reduces the network's dependence on label data and improves the generalization ability and robustness of the network.

[0128] Exemplary Devices

[0129] Figure 7 1 is a schematic diagram of a structure of a transient stability index prediction device for a new power system under high-proportion access of new energy provided by an exemplary embodiment of the present invention. Figure 7As shown, the apparatus 700 includes:

[0130] The first acquisition module 710 is used to acquire the continuous short-circuit ratio data of the power grid of the power system, and the continuous short-circuit ratio data of the power grid of the power system includes: historical short-circuit ratio data and real-time short-circuit ratio data;

[0131] The second acquisition module 720 is used to pre-process the continuous renewable energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence, where the binary discrete pulse sequence includes: a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence;

[0132] A generation module 730 is used to train a pre-built pulse-driven brain-like neuromorphic network according to a historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, wherein the brain-like neuromorphic network includes a hippocampal neural network and a prefrontal cortex neural network, and the hippocampal neural network and the prefrontal cortex neural network both include a plurality of leakage integration triggering neurons and a plurality of artificial leakage integration triggering neurons;

[0133] The output module 740 is used to output the prediction result of the binary new energy short circuit ratio after the second preset time period according to the historical binary discrete pulse sequence and the real-time binary discrete pulse sequence within the first preset time period by using the new energy short circuit ratio data prediction model;

[0134] The third acquisition module 750 is used to convert the binary new energy short circuit ratio prediction result into a continuous signal through a retainer, and obtain the new energy short circuit ratio prediction result after a second preset time period.

[0135] Optionally, the first acquisition module 710 includes:

[0136] The first acquisition submodule is used to discretize the continuous new energy short-circuit ratio data of the power grid according to the preset time and the adoption interval, and obtain the discretized new energy short-circuit ratio data;

[0137] The second acquisition submodule is used to process the discretized new energy short-circuit ratio data through a Poisson sequence generator to obtain a binary discrete pulse sequence that is compatible with the neuromorphic event-driven characteristics.

[0138] Optionally, the second acquisition submodule includes:

[0139] A first acquisition unit is used to process the discrete new energy short-circuit ratio data using a sliding window to obtain a plurality of processing window data matrices;

[0140] A generating unit, for generating a random number obeying Poisson distribution corresponding to each processing window data matrix by using a Poisson pulse sequence generator according to a plurality of processing window data matrices;

[0141] A second acquisition unit is used to compare the value of the element in each processing window data matrix with the corresponding random number, update the processing window data matrix according to the comparison result, and obtain multiple updated processing window data matrices;

[0142] The third acquisition unit is used to acquire a binary discrete pulse sequence according to a plurality of update processing window data matrices.

[0143] Optionally, the second acquisition unit updates the processing window data matrix according to the comparison result to acquire multiple updated processing window data matrices, including:

[0144] A first setting subunit is used to set the value of the element in the processing window data matrix to 1 when the value of the element in the processing window data matrix is ​​greater than the value of the corresponding random number in the comparison result;

[0145] The second setting subunit is used to set the value of the element in the processing window data matrix to 0 when the value of the element in the processing window data matrix is ​​smaller than the value of the corresponding random number;

[0146] The acquisition subunit is used to update multiple processing window data matrices and acquire multiple updated processing window data matrices.

[0147] Optionally, the input of the hippocampal network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the state information of the prefrontal cortex network; the input signal of the prefrontal cortex network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the learning signal transmitted by the hippocampal network; both the hippocampal network and the prefrontal cortex network use leaky integration triggering neurons and artificial leaky integration triggering neurons models to construct a multi-layer brain-like neural morphological network based on the synaptic time-dependent plasticity rule, in which

[0148] The leaky integration triggering neuron model is as follows:

[0149]

[0150] Where V j (t) represents the membrane potential of neuron j at time t; is the synaptic weight between presynaptic neuron i and neuron j; The synaptic weights for input information into the network; The effect of membrane time constant on membrane potential decay is characterized by α, τ m is the membrane time constant, which controls the decay of membrane potential; is the pulse output generated by the neuron when the membrane potential exceeds the threshold; d represents the delay of the presynaptic pulse train transmission;

[0151] The artificial leaky integration triggering neuron model is as follows:

[0152] a j (t) = ρa j (t-1)+S j (t-1)

[0153] A j (t) = βa j (t)+A j (t)

[0154]

[0155] Where V j (t) is the membrane potential at time t; V th is the adaptive threshold; The effect of the membrane time constant on the adaptive threshold decay is characterized by ρ, τ a is the membrane time constant; a j (t-1) is the update amount generated by the adaptive threshold of the neuron when the neuron generates a pulse output.

[0156] Optionally, the generating module 730 includes:

[0157] An input submodule, for inputting the historical binary output discrete pulse train into the hippocampal neural network and the prefrontal cortex neural network;

[0158] An extraction submodule, for extracting features of the historical binary output discrete pulse sequence through a prefrontal cortex neural network, and inputting the features into a hippocampal neural network;

[0159] The generating submodule is used for generating learning signals for synaptic weights of different neurons of the prefrontal cortex neural network according to historical binary output discrete pulse sequences and features through the hippocampal neural network, and feeding back the learning signals to the prefrontal cortex neural network.

[0160] An adjustment submodule, used for adjusting the synaptic weights of different neurons according to the learning signal through the prefrontal cortex neural network, and outputting the binary new energy short-circuit ratio training prediction result at a specific time point after the second set time period according to the adjusted synaptic weights of different neurons;

[0161] An output submodule, used for processing the binary new energy short circuit ratio indicator training prediction results through a retainer, converting them into continuous prediction results, and outputting continuous new energy short circuit ratio training prediction results;

[0162] The generation submodule is used to generate a new energy short circuit ratio data prediction model based on the new energy short circuit ratio training prediction results and the corresponding new energy short circuit ratio training actual results.

[0163] Exemplary Electronic Devices

[0164] Figure 8 This is a structure of an electronic device provided by an exemplary embodiment of the present invention. Figure 8 As shown, the electronic device 80 includes one or more processors 81 and a memory 82 .

[0165] The processor 81 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0166] The memory 82 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 81 may run the program instructions to implement the methods of the software programs of the various embodiments of the present invention described above and / or other desired functions. In one example, the electronic device may also include: an input device 83 and an output device 84, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0167] In addition, the input device 83 may also include, for example, a keyboard, a mouse, etc.

[0168] The output device 84 can output various information to the outside, and can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto.

[0169] Of course, to simplify, Figure 8 Only some of the components related to the present invention in the electronic device are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device may further include any other appropriate components according to specific application conditions.

[0170] Exemplary computer program products and computer-readable storage media

[0171] In addition to the above-mentioned methods and devices, an embodiment of the present invention may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the method according to various embodiments of the present invention described in the above-mentioned "Exemplary Method" section of this specification.

[0172] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the embodiments of the present invention, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0173] In addition, an embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enable the processor to execute the steps of the method according to various embodiments of the present invention described in the above “Exemplary Method” section of this specification.

[0174] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, system or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0175] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details disclosed above are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0176] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0177] The block diagrams of the devices, systems, equipment, and systems involved in the present invention are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, systems, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The word "such as" used here refers to the phrase "such as but not limited to", and can be used interchangeably with it.

[0178] The method and system of the present invention may be implemented in many ways. For example, the method and system of the present invention may be implemented by software, hardware, firmware or any combination of software, hardware, firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present invention are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present invention may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present invention. Thus, the present invention also covers a recording medium storing a program for executing the method according to the present invention.

[0179] It should also be noted that in the system, device and method of the present invention, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. The above description of the disclosed aspects is provided to enable any technician in the field to make or use the present invention. Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined here can be applied to other aspects without departing from the scope of the present invention. Therefore, the present invention is not intended to be limited to the aspects shown here, but in accordance with the widest range consistent with the principles and novel features disclosed here.

[0180] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for predicting transient stability indicators of a new power system under high-proportion access of new energy, characterized in that: include: Acquire continuous short-circuit ratio data of renewable energy sources in the power grid of the power system, wherein the continuous short-circuit ratio data of renewable energy sources in the power grid includes historical short-circuit ratio data and real-time short-circuit ratio data; Preprocessing the continuous renewable energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence, wherein the binary discrete pulse sequence includes: a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence; According to the historical binary discrete pulse sequence, a pre-constructed pulse-driven brain-like neuromorphic network is trained to generate a new energy short-circuit ratio data prediction model, wherein the brain-like neuromorphic network includes a hippocampal neural network and a prefrontal cortex neural network, and the hippocampal neural network and the prefrontal cortex neural network both include a plurality of leakage integration triggering neurons and a plurality of artificial leakage integration triggering neurons; Using the new energy short-circuit ratio data prediction model, according to the historical binary discrete pulse sequence and the real-time binary discrete pulse sequence within the first preset time period, output a binary new energy short-circuit ratio prediction result after a second preset time period; The binary new energy short-circuit ratio prediction result is converted into a continuous signal through a retainer to obtain the new energy short-circuit ratio prediction result after a second preset time period.

2. The method according to claim 1, characterized in that Preprocessing the continuous renewable energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence includes: According to the preset time and the adoption interval, the continuous new energy short-circuit ratio data of the power grid is discretized to obtain the discretized new energy short-circuit ratio data; The discrete new energy short-circuit ratio data is processed by a Poisson sequence generator to obtain the binary discrete pulse sequence adapted to the neuromorphic event-driven characteristics.

3. The method according to claim 2, characterized in that The discrete new energy short-circuit ratio data is processed by a Poisson sequence generator to obtain the binary discrete pulse sequence adapted to the neuromorphic event-driven characteristics, including: Applying a sliding window to process the discrete new energy short-circuit ratio data to obtain a plurality of processing window data matrices; A Poisson pulse sequence generator is used to generate a random number that obeys Poisson distribution and corresponds to each processing window data matrix according to a plurality of processing window data matrices; Compare the value of each element in the processing window data matrix with the corresponding random number, update the processing window data matrix according to the comparison result, and obtain multiple updated processing window data matrices; The binary discrete pulse sequence is obtained according to a plurality of update processing window data matrices.

4. The method according to claim 3, characterized in that The processing window data matrix is ​​updated according to the comparison result to obtain multiple updated processing window data matrices, including: In the comparison result, when the value of an element in the processing window data matrix is ​​greater than the corresponding value of the random number, the value of the element in the processing window data matrix is ​​set to 1; When the value of an element in the processing window data matrix is ​​smaller than the corresponding value of the random number, the value of the element in the processing window data matrix is ​​set to 0; A plurality of the processing window data matrices are updated to obtain a plurality of the updated processing window data matrices.

5. The method according to claim 1, characterized in that The input of the hippocampal network consists of two parts, one of which is the input new energy short-circuit ratio data, and the other is the state information of the prefrontal cortex network; the input signal of the prefrontal cortex network consists of two parts, one of which is the input new energy short-circuit ratio data, and the other is the learning signal transmitted by the hippocampal network; the hippocampal network and the prefrontal cortex network both use leaky integrated triggering neurons and artificial leaky integrated triggering neurons models to construct a multi-layer brain-like neural morphological network based on the synaptic time-dependent plasticity rule, wherein The leaky integration triggering neuron model is as follows: Where V j (t) represents the membrane potential of neuron j at time t; is the synaptic weight between presynaptic neuron i and neuron j; The synaptic weights for input information into the network; The effect of membrane time constant on membrane potential decay is characterized by α, τ m is the membrane time constant, which controls the decay of membrane potential; is the pulse output generated by the neuron when the membrane potential exceeds the threshold; d represents the delay of the presynaptic pulse train transmission; The artificial leakage integration triggering neuron model is as follows: a j (t)=ρa j (t-1)+S j (t-1) And j (t)=βa j (t)+A j (t) Where V j (t) is the membrane potential at time t; V th is the adaptive threshold; The effect of the membrane time constant on the adaptive threshold decay is characterized by ρ, τ a is the membrane time constant; a j (t-1) is the update amount generated by the adaptive threshold of the neuron when the neuron generates a pulse output.

6. The method according to claim 1, characterized in that The pre-built pulse-driven brain-like neuromorphic network is trained according to the historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, including: Inputting the historical binary output discrete pulse sequence into the hippocampal neural network and the prefrontal cortex neural network; Extracting features of the historical binary output discrete pulse sequence through the prefrontal cortex neural network, and inputting the features into the hippocampal neural network; The hippocampal neural network generates a learning signal for the synaptic weights of different neurons in the prefrontal cortex neural network according to the historical binary output discrete pulse sequence and the feature, and feeds the learning signal back to the prefrontal cortex neural network. The synaptic weights of the different neurons are adjusted according to the learning signal by the prefrontal cortex neural network, and the binary new energy short-circuit ratio training prediction result at a specific time point after the second set time period is output according to the adjusted synaptic weights of the different neurons; Processing the binary new energy short circuit ratio indicator training prediction result through the retainer, converting it into a continuous prediction result, and outputting a continuous new energy short circuit ratio training prediction result; The new energy short circuit ratio data prediction model is generated according to the new energy short circuit ratio training prediction results and the corresponding new energy short circuit ratio training actual results.

7. A transient stability index prediction device for a new power system with a high proportion of new energy access, characterized in that: include: A first acquisition module is used to acquire continuous short-circuit ratio data of renewable energy sources in a power grid of a power system, wherein the continuous short-circuit ratio data of renewable energy sources in a power grid includes historical short-circuit ratio data and real-time short-circuit ratio data; A second acquisition module is used to pre-process the continuous new energy short-circuit ratio data of the power grid to obtain a binary discrete pulse sequence, wherein the binary discrete pulse sequence includes: a historical binary discrete pulse sequence and a real-time binary discrete pulse sequence; A generation module, used for training a pre-built pulse-driven brain-like neuromorphic network according to the historical binary discrete pulse sequence to generate a new energy short-circuit ratio data prediction model, wherein the brain-like neuromorphic network includes a hippocampal neural network and a prefrontal cortex neural network, and both the hippocampal neural network and the prefrontal cortex neural network include a plurality of leakage integration triggering neurons and a plurality of artificial leakage integration triggering neurons; An output module, used to output a prediction result of the short-circuit ratio of a binary new energy source after a second preset time period according to the historical binary discrete pulse sequence and the real-time binary discrete pulse sequence within a first preset time period by using the new energy short-circuit ratio data prediction model; The third acquisition module is used to convert the binary new energy short-circuit ratio prediction result into a continuous signal through a retainer, and obtain the new energy short-circuit ratio prediction result after a second preset time period.

8. The device according to claim 7, characterized in that The first acquisition module includes: The first acquisition submodule is used to discretize the continuous new energy short-circuit ratio data of the power grid according to a preset time and an adoption interval to obtain discretized new energy short-circuit ratio data; The second acquisition submodule is used to process the discretized new energy short-circuit ratio data through a Poisson sequence generator to obtain the binary discrete pulse sequence that is compatible with the neuromorphic event-driven characteristics.

9. The device according to claim 8, characterized in that The second acquisition submodule includes: A first acquisition unit is used to process the discrete new energy short-circuit ratio data using a sliding window to obtain a plurality of processing window data matrices; A generating unit, configured to use a Poisson pulse sequence generator to generate a random number obeying Poisson distribution corresponding to each processing window data matrix according to a plurality of processing window data matrices; A second acquisition unit is used to compare the value of each element in the processing window data matrix with the corresponding random number, update the processing window data matrix according to the comparison result, and obtain multiple updated processing window data matrices; The third acquisition unit is used to acquire the binary discrete pulse sequence according to the multiple update processing window data matrices.

10. The device according to claim 9, characterized in that The second acquisition unit updates the processing window data matrix according to the comparison result, and acquires a plurality of updated processing window data matrices, including: A first setting subunit, configured to set the value of an element in the processing window data matrix to 1 when the value of an element in the processing window data matrix is ​​greater than the corresponding value of the random number in the comparison result; A second setting subunit, configured to set the value of an element in the processing window data matrix to 0 when the value of an element in the processing window data matrix is ​​smaller than the corresponding value of the random number; The acquisition subunit is used to update the multiple processing window data matrices and acquire the multiple updated processing window data matrices.

11. The device according to claim 7, characterized in that The input of the hippocampal network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the state information of the prefrontal cortex network; the input signal of the prefrontal cortex network consists of two parts, one is the input new energy short-circuit ratio data, and the other is the learning signal transmitted by the hippocampal network; the hippocampal network and the prefrontal cortex network both use leaky integrated triggering neurons and artificial leaky integrated triggering metamodels to construct a multi-layer brain-like neural morphological network based on the synaptic time-dependent plasticity rule, wherein The leaky integration triggering neuron model is as follows: Where V j (t) represents the membrane potential of neuron j at time t; is the synaptic weight between presynaptic neuron i and neuron j; The synaptic weights for input information into the network; The effect of membrane time constant on membrane potential decay is characterized by α, τ m is the membrane time constant, which controls the decay of membrane potential; is the pulse output generated by the neuron when the membrane potential exceeds the threshold; d represents the delay of the presynaptic pulse train transmission; The artificial leakage integration triggering neuron model is as follows: a j (t)=ρa j (t-1)+S j (t-1) And j (t)=βa j (t)+A j (t) Where V j (t) is the membrane potential at time t; V th is the adaptive threshold; The effect of the membrane time constant on the adaptive threshold decay is characterized by ρ, τ a is the membrane time constant; a j (t-1) is the update amount generated by the adaptive threshold of the neuron when the neuron generates a pulse output.

12. The device according to claim 7, characterized in that Generate modules, including: An input submodule, used for inputting the historical binary output discrete pulse sequence into the hippocampal neural network and the prefrontal cortex neural network; An extraction submodule, for extracting features of the historical binary output discrete pulse sequence through the prefrontal cortex neural network, and inputting the features into the hippocampal neural network; A generating submodule is used to generate a learning signal for the synaptic weights of different neurons of the prefrontal cortex neural network according to the historical binary output discrete pulse sequence and the feature through the hippocampal neural network, and feed the learning signal back to the prefrontal cortex neural network. an adjustment submodule, configured to adjust the synaptic weights of the different neurons according to the learning signal through the prefrontal cortex neural network, and output a binary new energy short-circuit ratio training prediction result at a specific time point after the second set time period according to the adjusted synaptic weights of the different neurons; An output submodule, used for processing the binary new energy short circuit ratio indicator training prediction result through the retainer, converting it into a continuous prediction result, and outputting a continuous new energy short circuit ratio training prediction result; The generation submodule is used to generate the new energy short circuit ratio data prediction model according to the new energy short circuit ratio training prediction result and the corresponding new energy short circuit ratio training actual result.

13. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to execute the method according to any one of claims 1 to 6.

14. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing instructions executable by the processor; The processor is used to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1 to 6.

Images