In recent years, the level of computerisation in society has increased significantly, resulting in a surge in demand for data processing. Consequently, traditional computing technologies based on Von Neumann architectures are encountering limitations. The separation of processors and storage units in traditional computers makes them less efficient and power-efficient when processing large amounts of data. Biological neural network computing presents a potential alternative solution to the Von Neumann bottleneck. Neuromorphic computing at the physical level relies on synaptic devices that mimic the function of biological synapses in neural networks. Significant progress has been made in artificial synaptic devices, such as phase-change memories, amnesia, and field-effect transistors. Organic thin-film transistors are currently a popular research topic due to their simple manufacturing process and low cost. In contrast to inorganic thin-film transistors, which require fabrication under high vacuum and high temperature conditions, organic electronic devices only require the formulation of organic materials into a solution for thin-film processing and annealing below 200 degrees Celsius. This reduces equipment requirements and manufacturing costs. To improve the simulation of synaptic function in thin film transistors, researchers have added a functional layer between the insulation and semiconductor layers. Various synaptic transistors have been prepared by combining materials with different properties. The three most common types of synaptic transistors are electret, ferroelectric, and floating gate, classified based on their functional layer. Ferroelectric synaptic transistors are a popular research topic among artificial synaptic devices due to the plasticity of their polarisation, which closely resembles that of biological synapses. However, the retention time of these ferroelectric synapses is short due to the gradual decrease of the ferroelectric polarization state over time. The rough ferroelectric film results in poor interface characteristics, including high leakage current and easy mutual diffusion between the ferroelectric film and the semiconductor layer. This is because the ferroelectric layer is in direct contact with the semiconductor. These issues negatively impact the device's basic performance, limiting the potential applications and development of ferroelectric synapses. This paper proposes adding an electret functional layer to ferroelectric synaptic transistors. The electret PVN can help passivate the rough surface of the ferroelectric material and improve the interface quality. The root-mean-square roughness of the film surface has been reduced from 2.95 nm to 0.66 nm. The film's homogeneity facilitates smooth carrier transportation at its interface. Additionally, the P(VDF-TrFE) layer has a higher dielectric constant than the PVN layer, resulting in a stronger electric field applied to the PVN layer. This allows for better electron trapping, resulting in a larger postsynaptic current and longer retention time after stimulation. These findings suggest a transition from short-term to long-term plasticity. Compared to conventional ferroelectric synaptic transistors, the switching ratio has been increased by an order of magnitude, and the threshold voltage has been reduced by 10 V. This leads to lower operating voltage and energy consumption. Photoelectrical synapses have become a popularresearch topic in recent years due to their advantages over traditional electrical synapses. These advantages include low crosstalk, low energy consumption, high bandwidth, and high transmission speed. Unlike traditional electrical synapses, which use a single electrical stimulus as the input signal, photoelectrical synapses offer a more efficient alternative. The combination of light and electricity in the input mode is more similar to the operational mechanism of biological synapses than a single electrical input. This is because a single device simulating a synapse has an insignificant impact on an entire biological neural network. However, constructing a complete artificial neural network requires a vast number of components and intricate circuits. Multilevel optoelectronic synapses hold great potential in reducing the number of devices and simplifying neural networks. This paper demonstrates that the electret layer of the device can capture different numbers of electrons depending on the colour of light. As a result, the device exhibits four distinct synaptic states under different colours of light, highlighting the potential of multi-stage photoelectric synapses. The hybrid input mode of light and electricity brings it closer to biological synapses, enabling more complex synaptic behaviours. The function of multi-stage photoelectric synapses gives them unlimited potential in the field of neural networks.