Link power control method and system for coherent optical wavelength division multiplexing transmission systems

Abstract

This invention provides a link power control method and system for a coherent optical wavelength division multiplexing (WDM) transmission system, comprising: building a coherent optical WDM transmission simulation system; configuring the transmit power of each channel in the WDM system; using the initial state of the transmit power as input to an execution neural network to obtain adjustment values ​​for the transmit power of different channels, and calculating the channel transmit power using the initial state and the transmit power adjustment values; configuring the attenuation of each channel of the wavelength selection switch in the WDM system; using the initial state of the wavelength selection switch as input to an execution neural network to obtain adjustment values ​​for the attenuation of the wavelength selection switch, and calculating the channel transmit power using the initial state and the attenuation adjustment values. This invention is based on a deep reinforcement learning algorithm, and by adjusting the transmit power of different channels in the coherent optical WDM system and the attenuation of different channels in the wavelength selection switch, it achieves optimized power configuration in the link, thereby optimizing the communication quality of each channel at the receiving end after transmission.

Description

Technical Field

[0001] This invention relates to the field of link power control technology, and more specifically, to a link power control method and system for a coherent optical wavelength division multiplexing transmission system. Background Art

[0002] Coherent optical wavelength division multiplexing (CWDM) systems are the mainstream optical transmission systems. In CWDM systems, the signals in each channel are affected by fiber nonlinearity, resulting in inconsistent transmission quality and impacting the overall system transmission level. Link power control is used to balance the communication quality of each channel. Current deep learning-based methods for controlling link power rely on accurate models of link transmission.

[0003] Patent application CN116938383A discloses a nonlinear compensation method and system based on a module multiplexing system, including the following steps: selecting any one mode as the second mode and the remaining modes as the first mode; applying a frequency shift to the wavelength division multiplexed optical signal of the second mode using a first optical frequency shifter; inputting the wavelength division multiplexed optical signal of the first mode and the shifted wavelength division multiplexed optical signal of the second mode into the same module multiplexer, and outputting them from the same module demultiplexer after passing through the same multimode fiber; applying a reverse frequency shift to the wavelength division multiplexed optical signal of the second mode output from the module demultiplexer using a second optical frequency shifter; and performing wavelength demultiplexing on the wavelength division multiplexed optical signal of the first mode output from the module demultiplexer and the wavelength division multiplexed optical signal of the second mode with the applied reverse frequency shift to complete the nonlinear compensation. However, this patent cannot completely solve the existing technical problems, nor can it meet the needs of this invention. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a link power control method and system for a coherent optical wavelength division multiplexing transmission system.

[0005] The link power control method for a coherent optical wavelength division multiplexing transmission system provided by the present invention includes:

[0006] Step 1: Build a coherent optical wavelength division multiplexing transmission simulation system, including setting the length of the optical fiber in the link, the parameters of the optical amplifier, and the overall transmission distance;

[0007] Step 2: Train the deep neural network using a simulation environment and configure the transmit power of each channel in the wavelength division multiplexing system;

[0008] Step 3: Input the initial state of the transmission power into the trained execution neural network to obtain the adjustment values ​​of the transmission power of different channels, and calculate the channel transmission power using the initial state and the transmission power adjustment values;

[0009] Step 4: Train the deep neural network using a simulation environment and configure the attenuation of each channel of the wavelength selection switch in the wavelength division multiplexing system;

[0010] Step 5: Input the initial state of the wavelength selective switch into the trained execution neural network to obtain the adjustment value of the wavelength selective switch attenuation, and calculate the channel transmit power using the initial state and the attenuation adjustment value.

[0011] Preferably, step 2 includes:

[0012] Step 2.1: Use the initial value of the transmit power as the starting state, use the starting state as the input to the execution neural network, and use the output as the adjustment value of the transmit power. Calculate the channel transmit power using the starting state and the transmit power adjustment value, then transmit the signal, and obtain the result at the receiving end. At the receiving end, obtain the signal-to-noise ratio (SNR) of each channel, calculate the corresponding reward using the SNR obtained at the receiving end and the target SNR, and update the starting state to the next state. Set the starting state, transmit power adjustment value, next state, and reward into a data pair, store it in the data pool, and generate the data in the data pool through continuous interaction between the execution neural network and the simulated transmission system.

[0013] Step 2.2: Train the critical neural network using data from the data pool. Use the current state and action of the data pair as input to the critical neural network to obtain its output. Take the difference between the output and the reward function and use the absolute value as the loss function of the critical neural network. Update the parameters of the critical neural network using gradient descent.

[0014] Step 2.3: Train the execution neural network using the data in the data pool and the critique neural network. Use the current state in the data pair as the input to the execution neural network to obtain its output. Use this output and the current state as the input to the critique neural network to obtain its output. Use the negative of the critique neural network's output as the loss function and update the parameters of the execution neural network using gradient descent.

[0015] Preferably, the channel transmit power is calculated using the initial state and the transmit power adjustment value, expressed as:

[0016] P ch =P0+P Δ

[0017] Among them, P ch P is the channel transmit power, P0 is the initial state, P Δ This is the power adjustment value;

[0018] The reward is calculated using the signal-to-noise ratio (SNR) obtained from the receiver and the target SNR, expressed as:

[0019] Ri =||SNR i -SNR target || 2

[0020] Among them, R i As a reward, SNR i For the receiver signal-to-noise ratio, SNR target The target signal-to-noise ratio.

[0021] Preferably, step 4 includes:

[0022] Step 4.1: Use the initial attenuation value as the starting state, and use the starting state as the input to the neural network. Obtain the output as the attenuation adjustment value. Calculate the attenuation of each channel using the starting state and the attenuation adjustment value, then transmit the data and obtain the result at the receiving end. Finally, obtain the signal-to-noise ratio (SNR) of each channel at the receiving end. Calculate the corresponding reward using the SNR obtained at the receiving end and the target SNR, and update the starting state to the next state. Set the starting state, attenuation adjustment value, next state, and reward into a data pair and store it in the data pool. Generate the data in the data pool through continuous interaction between the neural network and the simulated transmission system.

[0023] Step 4.2: Train a critical neural network using the data in the data pool;

[0024] Step 4.3: Use the data in the data pool and the critical neural network to train the execution neural network.

[0025] Preferably, the attenuation of each channel is calculated using the initial state and the attenuation adjustment value, as expressed by:

[0026] A ch =A0+A Δ

[0027] Among them, A ch For channel attenuation, A0 is the initial state, A Δ This is the attenuation adjustment value.

[0028] The link power control system for a coherent optical wavelength division multiplexing transmission system provided by the present invention includes:

[0029] Module M1: Build a coherent optical wavelength division multiplexing transmission simulation system, including setting the length of the optical fiber in the link, the parameters of the optical amplifier, and the overall transmission distance;

[0030] Module M2: Uses a simulation environment to train deep neural networks and configures the transmit power of each channel in the wavelength division multiplexing system;

[0031] Module M3: Takes the initial state of the transmit power as input and feeds it into the trained execution neural network to obtain the adjustment values ​​of the transmit power of different channels. It then calculates the channel transmit power using the initial state and the transmit power adjustment values.

[0032] Module M4: Uses a simulation environment to train deep neural networks and configures the attenuation of each channel of the wavelength selection switch in the wavelength division multiplexing system;

[0033] Module M5: The initial state of the wavelength selective switch is taken as input and fed into the trained execution neural network to obtain the adjustment value of the wavelength selective switch attenuation. The channel transmit power is calculated using the initial state and the attenuation adjustment value.

[0034] Preferably, the module M2 includes:

[0035] Module M2.1: The initial transmit power value is used as the starting state, and the starting state is used as the input to the execution neural network. The output is used as the adjustment value for the transmit power. The channel transmit power is calculated using the starting state and the transmit power adjustment value, then transmitted, and the result is obtained at the receiving end. At the receiving end, the signal-to-noise ratio (SNR) of each channel is obtained. The corresponding reward is calculated using the SNR obtained at the receiving end and the target SNR, and the starting state is updated to the next state. The starting state, transmit power adjustment value, next state, and reward are set as data pairs and stored in the data pool. Through continuous interaction between the execution neural network and the simulated transmission system, the data in the data pool is generated.

[0036] Module M2.2: Trains a critical neural network using data from the data pool, taking the current state and action of the data pair as input to the critical neural network to obtain its output; the difference between this output and the reward function is taken as the loss function of the critical neural network, and the parameters of the critical neural network are updated using gradient descent.

[0037] Module M2.3: Train the execution neural network using data from the data pool and the critical neural network. Use the current state of the data pair as the input to the execution neural network to obtain its output. Use this output and the current state as the input to the critical neural network to obtain its output. Use the negative of the critical neural network's output as the loss function and update the parameters of the execution neural network using gradient descent.

[0038] Preferably, the channel transmit power is calculated using the initial state and the transmit power adjustment value, expressed as:

[0039] P ch =P0+P Δ

[0040] Among them, P ch P is the channel transmit power, P0 is the initial state, PΔ This is the power adjustment value;

[0041] The reward is calculated using the signal-to-noise ratio (SNR) obtained from the receiver and the target SNR, expressed as:

[0042] R i =||SNR i -SNR target || 2

[0043] Among them, R i As a reward, SNR i For the receiver signal-to-noise ratio, SNR target The target signal-to-noise ratio.

[0044] Preferably, the module M4 includes:

[0045] Module M4.1: The initial attenuation value is used as the starting state, which is then used as the input to the neural network. The output is used as the attenuation adjustment value. Using the starting state and the attenuation adjustment value, the attenuation of each channel is calculated, then transmitted, and the result is obtained at the receiving end. Finally, the signal-to-noise ratio (SNR) of each channel is obtained at the receiving end. Using the SNR obtained at the receiving end and the target SNR, the corresponding reward is calculated, and the starting state is updated to the next state. The starting state, attenuation adjustment value, next state, and reward are set as data pairs and stored in the data pool. Through continuous interaction between the neural network and the simulated transmission system, the data in the data pool is generated.

[0046] Module M4.2: Trains a critical neural network using data from the data pool;

[0047] Module M4.3: Trains the execution neural network using data from the data pool and a critical neural network.

[0048] Preferably, the attenuation of each channel is calculated using the initial state and the attenuation adjustment value, as expressed by:

[0049] A ch =A0+A Δ

[0050] Among them, A ch For channel attenuation, A0 is the initial state, A Δ This is the attenuation adjustment value.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] (1) Based on deep reinforcement learning algorithm, this invention optimizes the power configuration in the link by adjusting the transmission power of different channels in the coherent optical wavelength division multiplexing system and the attenuation of different channels in the wavelength selection switch, thereby optimizing the communication quality of each channel at the receiving end after transmission.

[0053] (2) This invention enables neural networks to be trained without relying on accurate differentiable models through deep reinforcement learning, thereby solving the problem of different communication quality of different channels in coherent optical communication systems and enabling different channels to obtain the same signal-to-noise ratio. Attached Figure Description

[0054] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0055] Figure 1 For the generation of training data and the construction of the data pool;

[0056] Figure 2 To evaluate the network training flowchart;

[0057] Figure 3 To execute the network training flowchart;

[0058] Figure 4 A flowchart for setting up the transmitter power and wavelength selection switch;

[0059] Figure 5 To set the power optimization results at the transmitter. Detailed Implementation

[0060] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0061] Example 1

[0062] This invention provides a link power control method for a coherent optical wavelength division multiplexing transmission system based on deep reinforcement learning, comprising:

[0063] Step 1: Build a coherent optical wavelength division multiplexing transmission simulation system, including setting the length of the optical fiber in the link, the parameters of the optical amplifier, and the overall transmission distance;

[0064] Step 2: Train the deep neural network using a simulation environment so that the execution neural network can configure the transmit power of each channel in the wavelength division multiplexing system;

[0065] Step 2.1: Select a reasonable initial value of the transmission power as the starting state, use the starting state as the input of the neural network, and obtain the output as the adjustment value of the transmission power. Then, calculate the channel transmission power according to Equation (1) based on the starting state and the transmission power adjustment value, and then transmit the data and obtain the result at the receiving end.

[0066] P ch =P0+P Δ (1)

[0067] Among them, P ch P is the channel transmit power, P0 is the initial state, P Δ This is the power adjustment value.

[0068] Finally, the signal-to-noise ratio of each channel is obtained at the receiving end. Based on the signal-to-noise ratio obtained at the receiving end and the target signal-to-noise ratio, the corresponding reward is calculated according to Equation (2), and the starting state is updated to the next state.

[0069] R i =||SNR i -SNR target || 2 (2)

[0070] Among them, R i As a reward, SNR i For the receiver signal-to-noise ratio, SNR target The target signal-to-noise ratio.

[0071] The initial state, transmit power adjustment value, next state, and reward are set as data pairs and stored in a data pool. Through continuous interaction between the neural network and the simulated transmission system, the data in the data pool is generated, such as... Figure 1 .

[0072] Step 2.2: Train a critical neural network using data from the data pool, such as... Figure 2 The current state and action in the data pair are used as input to the critical neural network to obtain its output. The difference between this output and the reward function is taken as the absolute value, which is used as the loss function of the critical neural network. The parameters of the critical neural network are then updated using gradient descent.

[0073] Step 2.3: Train the execution neural network using the data in the data pool and the critical neural network, such as... Figure 3 The current state of the data pair is used as the input to the execution neural network to obtain its output; this output and the current state are used as the input to the crit neural network to obtain its output; the negative of this output is used as the loss function, and then the parameters of the execution neural network are updated using gradient descent.

[0074] Step 3: Input the initial state of the transmit power into the trained execution neural network to obtain the adjustment values ​​of the transmit power for different channels. Then, using the initial state and the transmit power adjustment values, calculate the transmit power of each channel according to equation (1).

[0075] Step 4: Train the deep neural network using a simulation environment, enabling the neural network to configure the attenuation of each channel of the wavelength selection switch in the wavelength division multiplexing system, such as... Figure 4 .

[0076] Step 4.1: Select a reasonable initial attenuation value as the starting state, use the starting state as the input to the neural network, and obtain the output as the attenuation adjustment value. Then, using the starting state and the attenuation adjustment value, calculate the attenuation of each channel according to equation (3), then transmit the data, and obtain the result at the receiving end, such as... Figure 5 .

[0077] A ch =A0+A Δ (3)

[0078] Among them, A ch For channel attenuation, A0 is the initial state, A Δ This is the attenuation adjustment value.

[0079] Finally, the signal-to-noise ratio (SNR) of each channel is obtained at the receiving end. Based on the SNR obtained at the receiving end and the target SNR, the corresponding reward is calculated according to equation (2), and the initial state is updated to the next state.

[0080] The initial state, decay adjustment value, next state, and reward are set as data pairs and stored in a data pool. Data in the data pool is generated through continuous interaction between the neural network and the simulation transmission system.

[0081] Step 4.2: Train the critical neural network using the data in the data pool, following the same process as in Step 2.2.

[0082] Step 4.3: Using the same process as in Step 2.2, train the execution neural network with the data in the data pool and the critical neural network.

[0083] Step 5: Input the initial state of the wavelength selective switch into the trained execution neural network to obtain the adjustment value of the wavelength selective switch attenuation. Then, calculate the channel transmit power according to equation (3) using the initial state and the attenuation adjustment value.

[0084] Example 2

[0085] The present invention also provides a link power control system for a coherent optical wavelength division multiplexing (WDM) transmission system. The link power control system of the coherent optical wavelength division multiplexing (WDM) transmission system can be implemented by executing the process steps of the link power control method of the coherent optical wavelength division multiplexing (WDM) transmission system. That is, those skilled in the art can understand the link power control method of the coherent optical wavelength division multiplexing (WDM) transmission system as a preferred embodiment of the link power control system of the coherent optical wavelength division multiplexing (WDM) transmission system.

[0086] The link power control system for a coherent optical wavelength division multiplexing (WDM) transmission system provided by the present invention includes: Module M1: building a coherent optical wavelength division multiplexing transmission simulation system, including setting the length of the optical fiber in the link, the parameters of the optical amplifier, and setting the overall transmission distance; Module M2: training a deep neural network using the simulation environment and configuring the transmit power of each channel in the WDM system; Module M3: taking the initial state of the transmit power as input and inputting it into the trained execution neural network to obtain the adjustment value of the transmit power of different channels, and calculating the channel transmit power using the initial state and the transmit power adjustment value; Module M4: training a deep neural network using the simulation environment and configuring the attenuation of each channel of the wavelength selection switch in the WDM system; Module M5: taking the initial state of the wavelength selection switch as input and inputting it into the trained execution neural network to obtain the adjustment value of the wavelength selection switch attenuation, and calculating the channel transmit power using the initial state and the attenuation adjustment value.

[0087] The module M2 includes: Module M2.1: Taking the initial value of the transmit power as the starting state, using the starting state as the input to the execution neural network, and obtaining the output as the adjustment value of the transmit power, calculating the channel transmit power using the starting state and the transmit power adjustment value, then transmitting, and obtaining the result at the receiving end; obtaining the signal-to-noise ratio (SNR) of each channel at the receiving end, calculating the corresponding reward using the SNR obtained at the receiving end and the target SNR, and updating the starting state to the next state; setting the starting state, transmit power adjustment value, next state, and reward into a data pair, storing it in the data pool, and generating data in the data pool through continuous interaction between the execution neural network and the simulation transmission system; Module M2.2: using data The data in the data pool is used to train the critical neural network. The current state and the action performed in the data pair are used as inputs to the critical neural network to obtain its output. The difference between the output and the reward function is taken as the absolute value and used as the loss function of the critical neural network. The parameters of the critical neural network are updated using gradient descent. Module M2.3: The data in the data pool and the critical neural network are used to train the execution neural network. The current state in the data pair is used as inputs to the execution neural network to obtain its output. The output and the current state are used as inputs to the critical neural network to obtain its output. The negative of the critical neural network's output is used as the loss function and the parameters of the execution neural network are updated using gradient descent.

[0088] The channel transmit power is calculated using the initial state and transmit power adjustment value, expressed as follows:

[0089] P ch =P0+P Δ

[0090] Among them, P ch P is the channel transmit power, P0 is the initial state, P Δ This is the power adjustment value;

[0091] The reward is calculated using the signal-to-noise ratio (SNR) obtained from the receiver and the target SNR, expressed as:

[0092] R i =||SNR i -SNR target || 2

[0093] Among them, R i As a reward, SNR i For the receiver signal-to-noise ratio, SNR target The target signal-to-noise ratio.

[0094] The module M4 includes: Module M4.1: Taking the initial attenuation value as the starting state, using the starting state as the input to the execution neural network, and obtaining the output as the attenuation adjustment value, calculating the attenuation of each channel using the starting state and the attenuation adjustment value, then transmitting the data and obtaining the result at the receiving end; finally, obtaining the signal-to-noise ratio (SNR) of each channel at the receiving end, calculating the corresponding reward using the SNR obtained at the receiving end and the target SNR, and updating the starting state to the next state; combining the starting state, attenuation adjustment value, next state, and reward into a data pair and storing it in the data pool, generating the data in the data pool through continuous interaction between the execution neural network and the simulation transmission system; Module M4.2: Training the critical neural network using the data in the data pool; Module M4.3: Training the execution neural network using the data in the data pool and the critical neural network.

[0095] The attenuation of each channel is calculated using the initial state and attenuation adjustment value, as expressed by:

[0096] A ch =A0+A Δ

[0097] Among them, A ch For channel attenuation, A0 is the initial state, A Δ This is the attenuation adjustment value.

[0098] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0099] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A link power control method for a coherent optical wavelength division multiplexing transmission system, characterized in that, include: Step 1: Build a coherent optical wavelength division multiplexing transmission simulation system, including setting the length of the optical fiber in the link, the parameters of the optical amplifier, and the overall transmission distance; Step 2: Train the deep neural network using a simulation environment and configure the transmit power of each channel in the wavelength division multiplexing system; Step 3: Input the initial state of the transmission power into the trained execution neural network to obtain the adjustment values ​​of the transmission power of different channels, and calculate the channel transmission power using the initial state and the transmission power adjustment values; Step 4: Train the deep neural network using a simulation environment and configure the attenuation of each channel of the wavelength selection switch in the wavelength division multiplexing system; Step 5: Input the initial state of the wavelength selective switch into the trained execution neural network to obtain the adjustment value of the wavelength selective switch attenuation, and calculate the channel transmit power using the initial state and the attenuation adjustment value; Step 2 includes: Step 2.1: Use the initial value of the transmit power as the starting state, use the starting state as the input to the execution neural network, and use the output as the adjustment value of the transmit power. Calculate the channel transmit power using the starting state and the transmit power adjustment value, then transmit the signal, and obtain the result at the receiving end. At the receiving end, obtain the signal-to-noise ratio (SNR) of each channel, calculate the corresponding reward using the SNR obtained at the receiving end and the target SNR, and update the starting state to the next state. Set the starting state, transmit power adjustment value, next state, and reward into a data pair, store it in the data pool, and generate the data in the data pool through continuous interaction between the execution neural network and the simulated transmission system. Step 2.2: Train the critical neural network using data from the data pool. Use the current state and action of the data pair as input to the critical neural network to obtain its output. Take the difference between the output and the reward function and use the absolute value as the loss function of the critical neural network. Update the parameters of the critical neural network using gradient descent. Step 2.3: Train the execution neural network using the data in the data pool and the critique neural network. Use the current state in the data pair as the input to the execution neural network to obtain its output. Use this output and the current state as the input to the critique neural network to obtain its output. Use the negative of the critique neural network's output as the loss function and update the parameters of the execution neural network using gradient descent. Step 4 includes: Step 4.1: Use the initial attenuation value as the starting state, and use the starting state as the input to the neural network. Obtain the output as the attenuation adjustment value. Calculate the attenuation of each channel using the starting state and the attenuation adjustment value, then transmit the data and obtain the result at the receiving end. Finally, obtain the signal-to-noise ratio (SNR) of each channel at the receiving end. Calculate the corresponding reward using the SNR obtained at the receiving end and the target SNR, and update the starting state to the next state. Set the starting state, attenuation adjustment value, next state, and reward into a data pair and store it in the data pool. Generate the data in the data pool through continuous interaction between the neural network and the simulated transmission system. Step 4.2: Train a critical neural network using the data in the data pool; Step 4.3: Use the data in the data pool and the critical neural network to train the execution neural network.

2. The link power control method for a coherent optical wavelength division multiplexing transmission system according to claim 1, characterized in that, The channel transmit power is calculated using the initial state and transmit power adjustment value, expressed as follows: in, For channel transmit power, This is the initial state. This is the power adjustment value; The reward is calculated using the signal-to-noise ratio (SNR) obtained from the receiver and the target SNR, expressed as: in, As a reward, For the signal-to-noise ratio at the receiving end, The target signal-to-noise ratio.

3. The link power control method for a coherent optical wavelength division multiplexing transmission system according to claim 1, characterized in that, The attenuation of each channel is calculated using the initial state and attenuation adjustment value, as expressed by: in, For channel attenuation, This is the initial state. This is the attenuation adjustment value.

4. A link power control system for a coherent optical wavelength division multiplexing transmission system, characterized in that, include: Module M1: Build a coherent optical wavelength division multiplexing transmission simulation system, including setting the length of the optical fiber in the link, the parameters of the optical amplifier, and the overall transmission distance; Module M2: Uses a simulation environment to train deep neural networks and configures the transmit power of each channel in the wavelength division multiplexing system; Module M3: Takes the initial state of the transmit power as input and feeds it into the trained execution neural network to obtain the adjustment values ​​of the transmit power of different channels. It then calculates the channel transmit power using the initial state and the transmit power adjustment values. Module M4: Uses a simulation environment to train deep neural networks and configures the attenuation of each channel of the wavelength selection switch in the wavelength division multiplexing system; Module M5: The initial state of the wavelength selective switch is taken as input and fed into the trained execution neural network to obtain the adjustment value of the wavelength selective switch attenuation. The channel transmit power is calculated using the initial state and the attenuation adjustment value. The module M2 includes: Module M2.1: The initial transmit power value is used as the starting state, and the starting state is used as the input to the execution neural network. The output is used as the adjustment value for the transmit power. The channel transmit power is calculated using the starting state and the transmit power adjustment value, then transmitted, and the result is obtained at the receiving end. At the receiving end, the signal-to-noise ratio (SNR) of each channel is obtained. The corresponding reward is calculated using the SNR obtained at the receiving end and the target SNR, and the starting state is updated to the next state. The starting state, transmit power adjustment value, next state, and reward are set as data pairs and stored in the data pool. Through continuous interaction between the execution neural network and the simulated transmission system, the data in the data pool is generated. Module M2.2: Trains a critical neural network using data from the data pool, taking the current state and action of the data pair as input to the critical neural network to obtain its output; the difference between this output and the reward function is taken as the loss function of the critical neural network, and the parameters of the critical neural network are updated using gradient descent. Module M2.3: Train the execution neural network using data from the data pool and the critical neural network. Use the current state of the data pair as the input to the execution neural network to obtain its output. Use this output and the current state as the input to the critical neural network to obtain its output. Use the negative of the critical neural network's output as the loss function and update the parameters of the execution neural network using gradient descent. The module M4 includes: Module M4.1: The initial attenuation value is used as the starting state, which is then used as the input to the neural network. The output is used as the attenuation adjustment value. Using the starting state and the attenuation adjustment value, the attenuation of each channel is calculated, then transmitted, and the result is obtained at the receiving end. Finally, the signal-to-noise ratio (SNR) of each channel is obtained at the receiving end. Using the SNR obtained at the receiving end and the target SNR, the corresponding reward is calculated, and the starting state is updated to the next state. The starting state, attenuation adjustment value, next state, and reward are set as data pairs and stored in the data pool. Through continuous interaction between the neural network and the simulated transmission system, the data in the data pool is generated. Module M4.2: Trains a critical neural network using data from the data pool; Module M4.3: Trains the execution neural network using data from the data pool and a critical neural network.

5. The link power control system of the coherent optical wavelength division multiplexing transmission system according to claim 4, characterized in that, The channel transmit power is calculated using the initial state and transmit power adjustment value, expressed as follows: in, For channel transmit power, This is the initial state. This is the power adjustment value; The reward is calculated using the signal-to-noise ratio (SNR) obtained from the receiver and the target SNR, expressed as: in, As a reward, For the signal-to-noise ratio at the receiving end, The target signal-to-noise ratio.

6. The link power control system of the coherent optical wavelength division multiplexing transmission system according to claim 4, characterized in that, The attenuation of each channel is calculated using the initial state and attenuation adjustment value, as expressed by: in, For channel attenuation, This is the initial state. This is the attenuation adjustment value.

Images